Site Projections Posted (1 Viewer)

[Chase Stuart]

A couple things I've noticed:1. Jacobs has had a 5.0 ypc both of the last 2 years on over 200 carries in both. You both have him projected to drop almost half a yard to 4.6 ypc. Granted, 4.6 ypc is still excellent, but is this more of a "5.0 ypc is hard to maintain" or do you see something that is going to cause them to be less effective?2. To David--AJ has been an absolute PPR machine the last 3 years. In 2 of the last 3 years, he's had 103 and 115 receptions. In 2007, when he was amazing but only played 9 games, he was on pace for over 106 receptions. Projecting him for only 85 receptions seems like a pretty substantial drop. Looking at the Houston team, you're projecting the QB's to throw 24 fewer attempts compared to 2008. Yet, you projected 9 fewer RB receptions, 7 fewer TE receptions, and 30 fewer WR receptions. You're also projecting a substantial drop in y/a for Schaub and the Texans passing game in general. Just curious what the thinking behind this is.

1. Projecting any RB to get 5.0 YPC is, IMO, misguided. I don't think any RB with over 150 carries -- Jacobs, DeAngelo Williams, ADP, etc. -- should be projected for that. 4.7 or 4.8 is probably the top range you could give, and very few guys should ever be projected to average such a high number.2. This is a very valid criticism, though. AJ has gone from 6.4 to 6.7 to 7.2 receptions per game the last three seasons. I can state with pretty good certainty that Dodds' final projections won't have AJ averaging 5.3 receptions per game.

 

[Biabreakable]

I looked at the Patriots projections, and essentially NE gains a 7 TD improvement from last year. I have no idea what the right amount of change should be, but I'm wondering if Brady's return might net the Pats more than 7 TD. By comparison, the total number of offensive TD in 2007 was 64 and projected now at 47. That's also only 2 more total TD than 2006 when the Pats fielded such notables as Reche Caldwell, Doug Gabriel, and 90-year-old Troy Brown at receiver. That seems too low to me . . . is there some rationale to this?

Hey David,I'm not sure if you were directing that just Dodds, or to me also.I currently have NE throwing 29 TDs and scoring 16 rushing TD [45 TDs]. While I understand your point that New England produced 67 TDs in 2007, I also think it's important to recognize that for what it was, a major historical outlier.In my projections workbook, I maintain per team offensive stats in all the major categories as well as running averages [currently using 6-year averages b/c I built this workbook that many seasons ago]. From 2003-2008, the Patriots have averaged 29 passing TDs and, you guessed it, 16 rushing TDs. And that includes the 50 TD passing seasons in question.So with Brady coming back from injury and the relative uncertainty we have in the middle of May given how mum the Pats organization is about injuries; I'm comfortable expecting New England to produce their normal TD production, as opposed to either last year [when the passing game was below plan while the rushing TDs were high] or their record-breaking and likely never repeatable 2007 campaign.

What were your projections for Brady last year before he got injured?

 

[Biabreakable]

I looked at the Patriots projections, and essentially NE gains a 7 TD improvement from last year. I have no idea what the right amount of change should be, but I'm wondering if Brady's return might net the Pats more than 7 TD. By comparison, the total number of offensive TD in 2007 was 64 and projected now at 47. That's also only 2 more total TD than 2006 when the Pats fielded such notables as Reche Caldwell, Doug Gabriel, and 90-year-old Troy Brown at receiver. That seems too low to me . . . is there some rationale to this?

Hey David,I'm not sure if you were directing that just Dodds, or to me also.I currently have NE throwing 29 TDs and scoring 16 rushing TD [45 TDs]. While I understand your point that New England produced 67 TDs in 2007, I also think it's important to recognize that for what it was, a major historical outlier.In my projections workbook, I maintain per team offensive stats in all the major categories as well as running averages [currently using 6-year averages b/c I built this workbook that many seasons ago]. From 2003-2008, the Patriots have averaged 29 passing TDs and, you guessed it, 16 rushing TDs. And that includes the 50 TD passing seasons in question.So with Brady coming back from injury and the relative uncertainty we have in the middle of May given how mum the Pats organization is about injuries; I'm comfortable expecting New England to produce their normal TD production, as opposed to either last year [when the passing game was below plan while the rushing TDs were high] or their record-breaking and likely never repeatable 2007 campaign.

I have been arguing in other threads how people should NOT be expecting Brady to throw for 50 TD, and there are a number of people thinking he will come closer to that than not.As for the methodology to coming up with YOUR projections, I believe this is an apples and oranges proposition. The Pats, as consituted, are not even close to rostering the same players vs. several years ago, especially at the skill positions (save for Brady and Faulk) than the team that was rolled out 5 years ago.They have 0 of the same receivers (and the ones they have now are a HUGE upgrade to what they had), they have only one similar RB (Faulk as I mentioned), their current OL is an upgrade vs what they fielded back in the day, and they've since opened up the playbook to better utilize all the toys in the arsenal.I'm certainly not going to tell you or anyone else how to do their projections. And if I wanted to I certainly could do some on my own (I would but don't have the time available to constantly update them).But if I were to make my own for the Pats, I would suggest using the last 4 years with Brady excluding the 2007 season as half the baseline and then taking the 2007 numbers and using that for the other half. Basically, setting the bar this year for NE at half what they did between their other years and what they did in 2007.Utilizing that approach, Brady would be projected at 4271/37. Admittedly, that TD total still seems a little high, but the point was to show that the more current numbers better reflect what the current team has accomplished and should be factored in more heavily IMO.Looking at your Pats projections, you have them scoring the same total offensive TD as 2006. Personally, I just don't see how this version of the Pats offense does not score more than that team did.

I am going to start working on NE projections today as it keeps nibbling at me and won't go away.I think your half and half idea is an interesting one and really not a bad way to a projection anyways. The most recent season is what is most relevant. So while you don't want to look only at last year having the most recent season weighted 50% of a 3 or 4 year average isn't too bad for a starting point.I will look at more than 3 years when personel is consistent for that long. But it rarely is.

 

[Dr. Octopus]

Here is the part you simply cannot project but it is the core of what makes these projections so dangerous for redrafters...INJURIES.

What a revelation! NFL players get injured! Hey, wait- another revelation! You can't predict or project injuries by and large! Damn those projections!!!!!!!1

Really? It's a perfectly fair question or angle on the stats. 18 QB with 20+ TDs is too high. You did nothing to dispute it, you just decided to mock. Got it, thanks.

I think when you do projections you have to take a "best case scenario" approach. Its much more pointless to try and predict injuries than to project stats under the "best case scenario". Obviously at the end of the day some of those QBs will get injured and therefore not approach those numbers - but how can we know which ones?I think you can use projections as long as you realize they are not "rankings" necessarily - risk and ADP have to come into play on draft day.

 

[gianmarco]

A couple things I've noticed:1. Jacobs has had a 5.0 ypc both of the last 2 years on over 200 carries in both. You both have him projected to drop almost half a yard to 4.6 ypc. Granted, 4.6 ypc is still excellent, but is this more of a "5.0 ypc is hard to maintain" or do you see something that is going to cause them to be less effective?2. To David--AJ has been an absolute PPR machine the last 3 years. In 2 of the last 3 years, he's had 103 and 115 receptions. In 2007, when he was amazing but only played 9 games, he was on pace for over 106 receptions. Projecting him for only 85 receptions seems like a pretty substantial drop. Looking at the Houston team, you're projecting the QB's to throw 24 fewer attempts compared to 2008. Yet, you projected 9 fewer RB receptions, 7 fewer TE receptions, and 30 fewer WR receptions. You're also projecting a substantial drop in y/a for Schaub and the Texans passing game in general. Just curious what the thinking behind this is.

1. Projecting any RB to get 5.0 YPC is, IMO, misguided. I don't think any RB with over 150 carries -- Jacobs, DeAngelo Williams, ADP, etc. -- should be projected for that. 4.7 or 4.8 is probably the top range you could give, and very few guys should ever be projected to average such a high number.2. This is a very valid criticism, though. AJ has gone from 6.4 to 6.7 to 7.2 receptions per game the last three seasons. I can state with pretty good certainty that Dodds' final projections won't have AJ averaging 5.3 receptions per game.

I don't think Jacobs should be projected for 5.0 either. But, given how effective he has been pretty consistently over the last 2 years on a significant # of carries, I think he deserves the benefit of the doubt to be projected in the 4.7-4.8 range. Splitting hairs, yes, but if he's willing to project AP for 4.7, CJ3 for 4.7, and DeAngelo for 5.0, then bumping up Jacobs closer to what he's done just makes more sense to me unless there's a reason to downgrade him.

 

[T Bell]

Here is the part you simply cannot project but it is the core of what makes these projections so dangerous for redrafters...INJURIES.

What a revelation! NFL players get injured!

Hey, wait- another revelation! You can't predict or project injuries by and large!

Damn those projections!!!!!!!1

Really? It's a perfectly fair question or angle on the stats. 18 QB with 20+ TDs is too high. You did nothing to dispute it, you just decided to mock.

Got it, thanks.

Yes, you're right. I decided to mock a tired argument that couldn't be more flawed from someone notorious in the SP for making ridiculous predictions and arguments. How flawed is the criticism? Well, it's so flawed that you failed to offer any alternative method for making projections other than to whine that Dodds' projections are "too high".

Projections of which players are "too high"? How much should they be reduced? Should every player be reduced? You don't address any of this. You just complain that they're "too high".

Why? Because there aren't good alternatives for predicting injuries and putting them into hard statistical projections.

So yeah, I mocked you. And I'd do it again too.

 

[renesauz]

Here is the part you simply cannot project but it is the core of what makes these projections so dangerous for redrafters...INJURIES.

What a revelation! NFL players get injured!

Hey, wait- another revelation! You can't predict or project injuries by and large!

Damn those projections!!!!!!!1

Really? It's a perfectly fair question or angle on the stats. 18 QB with 20+ TDs is too high. You did nothing to dispute it, you just decided to mock.

Got it, thanks.

But you're asking the impossible. To accomadate you, they would have to project the actual injuries and time missed for them. By nature, projections have to steer clear of significant injuries, or they don't work at all. It's one thing to predict every starting RB to miss one game....it's another to predict which specific QBs will miss 5 or more games.WE ALL UNDERSTAND that many players will fail to live up to projection because of injury. We also understand that many players will greatly exceed their projections because of unexpected opportunity (due to a starter's injury).

You understand this as well, which might be why he was being sarcastic to you....because it seems you were sarcastic first. If the real thing is that you don't like using projections...then stay out of the conversation

 

[smotherhook]

i see felix jones getting more than 10 rec. for the year

 

[BuckeyeArt]

I think when you do projections you have to take a "best case scenario" approach. Its much more pointless to try and predict injuries than to project stats under the "best case scenario". Obviously at the end of the day some of those QBs will get injured and therefore not approach those numbers - but how can we know which ones?I think you can use projections as long as you realize they are not "rankings" necessarily - risk and ADP have to come into play on draft day.

Agree completely with this. I want to see what a player can do if remains the uninjured starter. It is far more useful to have a per game projection than a yearly projection that might account for an injury.This is why I wanted to hear a little more about their Cincy RB projections. The projected distribution of carries is about like the end of year distribution last year. But, they should look much different if you look at how the distribution actually occurred.Last year, the lead RB (each game) in Cincy got 85.5% of the RB carries (highest in the NFL). The year before, 78.3%. So, while the end of year numbers look a little like a RBBC, it really is not even remotely like one. Both Wood and Dodds are projecting Benson for only 56-60% of the carries which suggests that they are either projecting an injury, a RBBC, or a new starter during the year. I'm wondering which it is.

 

[Joe Bryant]

In the 10 years I've been following you and Joe, I've yet to figure out why you refuse to make your QB projections to mirror anything close to reality.

Just to follow up a bit more on this, because it gets mentioned every year (and also every week when the weekly projections start rolling out).I hope you didn't think my previous response was too glib. You are aboslutely 100% right that Dodds' projections do not mirror anything close to reality on issues like "how many QBs will throw for 30 TDs?" or "how many receiving yards will the league's leading receiver have?" But Dodds is not trying to answer those questions.

I don't want to speak for Dodds, but I guess I will anyway. If you ask him these questions, these are the answers I think he'd give:

Who do you think is most likely to lead the league in receiving yards? Calvin Johnson.

Do you think the league leader in receiving yards will have more than 1400 yards? Yes.

Do you think Calvin Johnson will have more than 1400 yards? No.

There is nothing inconsistent about this.

Here's some data:

in the last four years, there have been 41 WRs who Dodds has projected over 1000 yards and who have ended up playing 16 games in that season. On average, Dodds has underprojected those players by 9 yards. That's pretty darn close. There have been 16 QBs who he has projected to throw 25 or more TD passes, and who have ended up playing 16 games in that season. On average, he has overprojected those players by 1.9 TDs per season. Again, pretty close and, more importantly, no evidence that Dodds should systematically increase his projections for the top players.

data snipped

Thanks Doug. And thanks vandy for asking the question. It was a good point. And Doug did a good job of drawing out the reasons behind the answer. That's the beauty of this Shark Pool. Thanks.J

 

[Jason Wood]

I looked at the Patriots projections, and essentially NE gains a 7 TD improvement from last year. I have no idea what the right amount of change should be, but I'm wondering if Brady's return might net the Pats more than 7 TD. By comparison, the total number of offensive TD in 2007 was 64 and projected now at 47. That's also only 2 more total TD than 2006 when the Pats fielded such notables as Reche Caldwell, Doug Gabriel, and 90-year-old Troy Brown at receiver. That seems too low to me . . . is there some rationale to this?

Hey David,I'm not sure if you were directing that just Dodds, or to me also.I currently have NE throwing 29 TDs and scoring 16 rushing TD [45 TDs]. While I understand your point that New England produced 67 TDs in 2007, I also think it's important to recognize that for what it was, a major historical outlier.In my projections workbook, I maintain per team offensive stats in all the major categories as well as running averages [currently using 6-year averages b/c I built this workbook that many seasons ago]. From 2003-2008, the Patriots have averaged 29 passing TDs and, you guessed it, 16 rushing TDs. And that includes the 50 TD passing seasons in question.So with Brady coming back from injury and the relative uncertainty we have in the middle of May given how mum the Pats organization is about injuries; I'm comfortable expecting New England to produce their normal TD production, as opposed to either last year [when the passing game was below plan while the rushing TDs were high] or their record-breaking and likely never repeatable 2007 campaign.

I have been arguing in other threads how people should NOT be expecting Brady to throw for 50 TD, and there are a number of people thinking he will come closer to that than not.As for the methodology to coming up with YOUR projections, I believe this is an apples and oranges proposition. The Pats, as consituted, are not even close to rostering the same players vs. several years ago, especially at the skill positions (save for Brady and Faulk) than the team that was rolled out 5 years ago.They have 0 of the same receivers (and the ones they have now are a HUGE upgrade to what they had), they have only one similar RB (Faulk as I mentioned), their current OL is an upgrade vs what they fielded back in the day, and they've since opened up the playbook to better utilize all the toys in the arsenal.I'm certainly not going to tell you or anyone else how to do their projections. And if I wanted to I certainly could do some on my own (I would but don't have the time available to constantly update them).But if I were to make my own for the Pats, I would suggest using the last 4 years with Brady excluding the 2007 season as half the baseline and then taking the 2007 numbers and using that for the other half. Basically, setting the bar this year for NE at half what they did between their other years and what they did in 2007.Utilizing that approach, Brady would be projected at 4271/37. Admittedly, that TD total still seems a little high, but the point was to show that the more current numbers better reflect what the current team has accomplished and should be factored in more heavily IMO.Looking at your Pats projections, you have them scoring the same total offensive TD as 2006. Personally, I just don't see how this version of the Pats offense does not score more than that team did.

As you know David, I tweak my projections aggressively from mid April through Week One kickoff, and it's discussions like this that help me flesh it out. I don't disagree that my bias is to RAISE my production. But I need to see Brady on the practice field looking like, well, Brady, first. I also need to get a better idea of how the RB situation will be handled; right now there are too many fish in the sea.

What were your projections for Brady last year before he got injured?

Great question. I believe I had him at 34 TDs when we released May projections and bumped that up to 36 for much of the preseason.

 

[David Yudkin]

I also need to get a better idea of how the RB situation will be handled; right now there are too many fish in the sea.

Pats last year had no lead dog, used everyone, and still were Top 5-10 in almost every rushing category. I think they will utilize a similar approach, will keep people fresh, and each week will go with the guys that are being productive. So IMO, this will help keep their rushing totals high. The issue to me is HOW OFTEN they will run.

 

[Jason Wood]

I also need to get a better idea of how the RB situation will be handled; right now there are too many fish in the sea.

Pats last year had no lead dog, used everyone, and still were Top 5-10 in almost every rushing category. I think they will utilize a similar approach, will keep people fresh, and each week will go with the guys that are being productive. So IMO, this will help keep their rushing totals high. The issue to me is HOW OFTEN they will run.

We have no disagreement there.I have them at:490 rushes2,035 yards rushing4.2 yards per rush16 rushing TDs

 

[Biabreakable]

I also need to get a better idea of how the RB situation will be handled; right now there are too many fish in the sea.

Pats last year had no lead dog, used everyone, and still were Top 5-10 in almost every rushing category. I think they will utilize a similar approach, will keep people fresh, and each week will go with the guys that are being productive. So IMO, this will help keep their rushing totals high. The issue to me is HOW OFTEN they will run.

We have no disagreement there.I have them at:490 rushes2,035 yards rushing4.2 yards per rush16 rushing TDs

I am just going over this now and there are a couple other things I want to consider. Here is where I am at right now though.I am using 2007 as the highest weighted average and looked back 5 seasons. Normaly you would use the most recent season. But I think Brady going out after one pass attempt isn't representative of the kind of team they are with him on the field.Rushing attempts2008 513 - 4.42007 451 - 4.12006 499 - 3.92005 439 - 3.42004 524 - 4.1 I didn't use 2004 because I do not think NE has a RB like Dillon on their roster. Maybe Maroney could be the back with the highest carries this year. Not at 3.3 ypc he won't though. Maroneys best season so far was in 2007 so that is kind of already baking in his past performance. Maybe he will suprise. I doubt it.So average total runs for 2005/2006/2008 is 483.666 I then added 2007 and split them for 467.333Pretty close to what I expect end projection for number of runs.

 

[moderated]

After seeing these it's confirmed Pierre Thomas is the most over-rated fantasy player in the league for 09'.

This is eerily similar to the Samkon Gado love a few years back.

 

[Liquid Tension]

In the 10 years I've been following you and Joe, I've yet to figure out why you refuse to make your QB projections to mirror anything close to reality.

I do not understand what you are even saying in your statement, but I will use this to speak to historical norms. I feel these are critical to make sure a set of projections is based in reality. Most other sites do not do this check and those numbers are usually outside expectations. Here are the league numbers I craft my projections around.Pass Completions:

2008 - 10,089

2007 - 10,425

2006 - 9,796

2009 Projected - 10,163

Pass Attempts:

2008 - 16,536

2007 - 17,045

2006 - 16,389

2009 Projected - 16,673

Pass Percentage:

2008 - 61.0%

2007 - 61.2%

2006 - 59.8%

2009 Projected - 60.0%

Passing Average:

2008 - 6.54

2007 - 6.86

2006 - 6.85

2009 Projected - 6.73

Passing TDs:

2008 - 646

2007 - 720

2006 - 648

2009 Projected - 695

Interceptions:

2008 - 465

2007 - 534

2006 - 520

2009 Projected - 496

Rush Attempts:

2008 - 14,110

2007 - 13,804

2006 - 14,324

2009 Projected - 14,153

Rush Yards:

2008 - 59,340

2007 - 56,768

2006 - 60,094

2009 Projected - 58,553

Rushing Average:

2008 - 4.21

2007 - 4.11

2006 - 4.20

2009 Projected - 4.14

Rushing TDs:

2008 - 476

2007 - 386

2006 - 423

2009 Projected - 454

Well that seems very much reasonable. Assuming the projections add up then what we need to do is tweak the individual teams. I would think a team like Miami's offense would be one of the harder ones to project because of the wildcat.I have not dug into all of the projections yet, but I think Dom Hixon will do better than you have listed. he dropped one wide open bomb, but outside of that showed to be a solid #2 WR. Right now, he is the best they have.

 

[Biabreakable]

I also need to get a better idea of how the RB situation will be handled; right now there are too many fish in the sea.

Pats last year had no lead dog, used everyone, and still were Top 5-10 in almost every rushing category. I think they will utilize a similar approach, will keep people fresh, and each week will go with the guys that are being productive. So IMO, this will help keep their rushing totals high. The issue to me is HOW OFTEN they will run.

We have no disagreement there.I have them at:490 rushes2,035 yards rushing4.2 yards per rush16 rushing TDs

I am just going over this now and there are a couple other things I want to consider. Here is where I am at right now though.I am using 2007 as the highest weighted average and looked back 5 seasons. Normaly you would use the most recent season. But I think Brady going out after one pass attempt isn't representative of the kind of team they are with him on the field.Rushing attempts2008 513 - 4.42007 451 - 4.12006 499 - 3.92005 439 - 3.42004 524 - 4.1 I didn't use 2004 because I do not think NE has a RB like Dillon on their roster. Maybe Maroney could be the back with the highest carries this year. Not at 3.3 ypc he won't though. Maroneys best season so far was in 2007 so that is kind of already baking in his past performance. Maybe he will suprise. I doubt it.So average total runs for 2005/2006/2008 is 483.666 I then added 2007 and split them for 467.333Pretty close to what I expect end projection for number of runs.

Moving on-For NE using 2007 as the highest weighted average.513 - 4.4451 - 4.1499 - 3.9439 - 3.4524 - 4.1 I didn't use 2004 because Dillon483.666467.3331095 - 5.3 356 1st downs1058 - 6.2 393 1st downs1055 - 5.1 330 1st downs1031 - 5.5 334 1st downs1060593.333 PA 467.333 RA Used 2005 for Brady's rushing attempts 27 or 1.7/game this is the fewest times he has run the ball. But coming off injury he may not take off as quickly until he gets more comfortable. Brady has averaged 30.4 RA over his career. 3 less isn't a big deal.2007 372006 492005 27Matt Cassel 73 - 37 = 362008 513RA - 36 = 477 10 more than the weighted average 467Cassell still ran the ball 46 more times than last year than I expect Brady to. So who picks up those carries? does anyone? Sammy Morris 32 years old2008 156 727 4.7 7 17catch 161 9.5 42 0 1 1 2007 85 384 4.5 3 6catch 35 5.8 18 0 I didn't use 2006 or earlier because Morris was with other teams. He has a career average 4.2 ypc so the situation seems to be helping him. 32 now maybe he falls off.2009 120.5 carriesKevin Faulk - about to be 33 in june 2008 83 507 6.1 3 58catch 486 8.4 22 3 2007 62 265 4.3 0 47catch 383 8.1 23 1 1 0 2006 25 123 4.9 1 43catch 356 8.3 43T 2 2 1 Kevin Faulk is aging like fine wine. Having Moss and Welker demanding so much attention doesen't hurt. I give him the average of the last 2 seasons 72.5 carries. Of course on receptions is where he does the most damage. Faulk might be a good example of a player with low carries who is performing well at an older age. Warrick Dunn is another. These guys do seem to last a long time2008 LaMont Jordan 80 363 4.5 4 - They didn't keep Jordan so who gets those carries?Laurence Maroney2008 New England Patriots 3 3 28 93 3.3 02007 New England Patriots 13 6 185 835 6 4catch 116 29.0 43 0 2006 New England Patriots 14 0 175 745 4.3 6 22catch 194 8.8 31 1 avg 180 in 2006-07 and played in 13.5 games before only playing 3 last season. In the 3 games he did play averaged 9.3 carries 3.3 ypc is bad.13.5 times 9.3/game = 125.55 carries so I averaged that with 180 = 152.77 thats probably too many unless he eats into Morris slice. Morris had 156 carries last year which is close to this 152.77So at this time I am going to give Maroney the lead back at 153 carries. Morris will move to the 3rd RB and take Jordans role. BenJarvus Green-Ellis 74 275 3.7 15 5 I think they still use him but 3.7 is not a high percentage play. With Brady back and the other guys healthy I would start cutting his carries 1st. Heath Evans 11 23 2.1 4 0 Wes Welker 3 26 8.7 19 0 Kevin O'Connell 3 -6 -2.0 -2 0 Randy Moss 2 0 0.0 2 0 These are other runs. I ushualy just count this as 20.So 467 - 20 = 447 - 27 for Brady = 420Maroney 153 3.85 or 4.3? hmm 4.3 because otherwise Morris will be the leader here.Morris 120 4.2Faulk 73 5.1 Thats 346 leaving 74 carries for Green-Ellis or some of the other RB getting more.Like I said I think having the ball in Brady's hands is a lot more effective play than giving it to Green-Ellis. Ellis can also block and that might be something they want him to do more of protecting Brady. I am only giving him 34 carries. That leaves 40 plays that could be passes or runs with the other RB. And this is where ranges start coming into play. These 40 carries could be attributed to the difference in QB totals. I expect Brady to throw where Cassell would run.Maroney 153 4.3 658 yardsMorris 120 4.2 504 yardsFaulk 73 5.1 327 yardsGreen-Ellis 34 3.8 129So that is 380 carries for 1663 yards 4.377 ypc with 40 carries still floating. That could be 175 yards 1838 total yards rushing or they could be pass attempts. I am still setting them aside.2008 2278 yards 4.4 ypc2007 1849 yards 4.1 ypc2006 1969 yards 3.9That's where I am so far. Will try to do passing tomorow.

 

[David Yudkin]

Maroney 153 4.3 658 yardsMorris 120 4.2 504 yardsFaulk 73 5.1 327 yardsGreen-Ellis 34 3.8 129So that is 380 carries for 1663 yards 4.377 ypc with 40 carries still floating. That could be 175 yards 1838 total yards rushing or they could be pass attempts. I am still setting them aside.2008 2278 yards 4.4 ypc2007 1849 yards 4.1 ypc2006 1969 yards 3.9That's where I am so far. Will try to do passing tomorow.

And Fred Taylor holds a clip board and watches from the sidelines?

 

[Biabreakable]

Maroney 153 4.3 658 yardsMorris 120 4.2 504 yardsFaulk 73 5.1 327 yardsGreen-Ellis 34 3.8 129So that is 380 carries for 1663 yards 4.377 ypc with 40 carries still floating. That could be 175 yards 1838 total yards rushing or they could be pass attempts. I am still setting them aside.2008 2278 yards 4.4 ypc2007 1849 yards 4.1 ypc2006 1969 yards 3.9That's where I am so far. Will try to do passing tomorow.

And Fred Taylor holds a clip board and watches from the sidelines?

LOL good point. Knew I was missing something.What is the pecking order here? Fred is 33 years old but he is still Fred.:crumplespaper:

 

[David Yudkin]

Maroney 153 4.3 658 yardsMorris 120 4.2 504 yardsFaulk 73 5.1 327 yardsGreen-Ellis 34 3.8 129So that is 380 carries for 1663 yards 4.377 ypc with 40 carries still floating. That could be 175 yards 1838 total yards rushing or they could be pass attempts. I am still setting them aside.2008 2278 yards 4.4 ypc2007 1849 yards 4.1 ypc2006 1969 yards 3.9That's where I am so far. Will try to do passing tomorow.

And Fred Taylor holds a clip board and watches from the sidelines?

LOL good point. Knew I was missing something.

At some point I will try to compile my best guess . . . but my version would have the carries dispersed in most to fewest would be Morris, Taylor, Maroney, Faulk, BJGE.

 

[Biabreakable]

Maroney 153 4.3 658 yardsMorris 120 4.2 504 yardsFaulk 73 5.1 327 yardsGreen-Ellis 34 3.8 129So that is 380 carries for 1663 yards 4.377 ypc with 40 carries still floating. That could be 175 yards 1838 total yards rushing or they could be pass attempts. I am still setting them aside.2008 2278 yards 4.4 ypc2007 1849 yards 4.1 ypc2006 1969 yards 3.9That's where I am so far. Will try to do passing tomorow.

And Fred Taylor holds a clip board and watches from the sidelines?

LOL good point. Knew I was missing something.

At some point I will try to compile my best guess . . . but my version would have the carries dispersed in most to fewest would be Morris, Taylor, Maroney, Faulk, BJGE.

Ok thanks. I knew that 40 was going some where. Probably more than that for Fred.So in that case Fred gets Lamont Jordans role from last year 80 carries2008 Jacksonville Jaguars 13 13 143 556 3.9 1 2007 Jacksonville Jaguars 15 15 223 1,202 5.4 5 2006 Jacksonville Jaguars 15 15 231 1,146 5.0 5Fred clearly fell back at the age of 32 and that probably continues.Morris 120 4.2 504 yardsFred Taylor 80 3.9 312 yardsFaulk 73 5.1 327 yardsMaroney 74 3.85 weighted average 285 yardsGreen-Ellis 28 3.8 129this is 375 carries 1557 yards with 45 carries floating the ypc may be too low for Taylor and Maroney. Hard for me to see Maroney doing better than this and not getting more carries. Team average is 4.152 here which isn't that off. If this is correct then Morris and Fred will probably be the guys who split the 45 166 yards that would be 1725 total. Again I think my averages may be a bit low here. The RB and oline did better than this in 2008. Maybe the team average should be 4.2

 

[Ariakis]

I think my biggest take on projections from FBG's is that the averages work out but the elites might be under-estimated and the fodder gets over-estimated as you try to project injuries and platoons. There is always a huge middle class with very little to separate them.

You've got Romo far worse than last year yet still in the top 10 at QB's. Despite missing games you've got him with less passing yards, far less td's, and more INT's than last season.

 

[Maurile Tremblay]

Mathematically speaking, the issue here is that the expected value of a maximum of a bunch of random variables is not the same as the maximum of the expected values of those same variables. It is in no way inconsistent to believe that someone will have 1500 receiving yards this year while not projecting any particular individual to do so.

An example: every single week of NFL play for the last three years, some team has won a game by 20 or more points. Every single week. But Vegas almost never sets a line as high as 20. And that's because the questions "do you think some team will win by at least 20 this week?" and "is there a particular team you think will win by 20 this week?" usually have different answers.

Very good explanation.

 

[Maurile Tremblay]

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

Even though each projection comes down to a single number, that number really represents a range.I'm projecting that Jake Delhomme will throw for 3400 yards this season (based on playing 15 games).

But that 3400 really represents a range -- it means I think he's pretty likely to throw for between 3200 and 3600 yards.

Most of the time, when a QB throws for over 4000 yards, it means that he hit the high end of any reasonable projected range for him. When I project a QB to throw for 3900 yards, it doesn't mean I don't think he can hit 4000. In fact, if I project a guy to throw for 3900 yards, it means I think he has a good chance to throw for over 4000 -- and roughly as good a chance to throw for fewer than 3800.

If some QBs hit the high ends of their expected ranges while others hit the low ends, the best-performing QB will likely outperform the midpoint of his range (i.e., his projection) even if the projections correctly identified the midpoint of each QB's range.

In other words, if actual results have greater variance than projected results, it doesn't mean that the projections were flawed.

If you roll four six-sided dice, it is more likely than not that at least one of the dice will land on 6. But the best projection for each in individual die is nonetheless 3.5. The variance of the actual results is greater than the variance of the mathematically correct projections.

Football players aren't dice, but the same general principle applies. If the projections are done right, expect the projections to have less variance than the eventual results.

 

[Ministry of Pain]

Here is the part you simply cannot project but it is the core of what makes these projections so dangerous for redrafters...INJURIES.

What a revelation! NFL players get injured!

Hey, wait- another revelation! You can't predict or project injuries by and large!

Damn those projections!!!!!!!1

Really? It's a perfectly fair question or angle on the stats. 18 QB with 20+ TDs is too high. You did nothing to dispute it, you just decided to mock.

Got it, thanks.

Yes, you're right. I decided to mock a tired argument that couldn't be more flawed from someone notorious in the SP for making ridiculous predictions and arguments. How flawed is the criticism? Well, it's so flawed that you failed to offer any alternative method for making projections other than to whine that Dodds' projections are "too high".

Projections of which players are "too high"? How much should they be reduced? Should every player be reduced? You don't address any of this. You just complain that they're "too high".

Why? Because there aren't good alternatives for predicting injuries and putting them into hard statistical projections.

So yeah, I mocked you. And I'd do it again too.

You are completely off the ranch. I posted quite a few thoughts and I picked a couple teams at random to begin with. You simply came behind me and started using the emoticons which is one of the lower forms of posting and requires very little thought. If you like projections more power to you but don't spin this into the idea that I had no alternatives. You have been posting here for years so I am sure you have read what I usually post sometime in July/Aug, most of the board regs have seen me in great deatail show what I feel is stronger with an overall grade for each RB instead of just shotgun blasting the projections. Have some class T please, this whole "I'd do it again"...I guess you're the man here int he SP, eh? Can we dial that down a notch? It's not productive and I was never attacking DD...certainly he can defend himself but a couple times I mentioned his projections if those are your thing are amongst the best. (Maybe you should go back and read it again). Not sure why you felt the need to spill FFA mantra over to the SP but whatever. Good Luck

 

[LTsharks]

After seeing these it's confirmed Pierre Thomas is the most over-rated fantasy player in the league for 09'.This is eerily similar to the Samkon Gado love a few years back.

Back on track here a little bit....I wonder about Thomas too. He looked so good closing out last season, it is hard not to imagine him having a bigger role in 2009. That said, the rumblings of the Saints looking at RBs and the presence of Bush make me wonder what his ceiling really is.Both sets of projections seem high at first, but when you look at the breakdown for Saints RBs they both seem within reason (carries seem high for Woods). Not that I really disagree (Thomas just confuses me), but I would love some input on why Dodds and Wood feel he warrants those projections.

 

[H8fulHuck]

Sorry you felt I was unclear. Your projections for QBs (and WRs/TEs) every year are Glenn Beck-levels of crazy conservative. They are useless in gauging actual performance because you generally understate your top performers by 15% or more than what the top performers of that year will end up at. To illustrate this point using your 2009 #s:

# of QBs in 2008 over or equal to 7.7 YPA: 10

2007: 7

2006: 6

# you're projecting in 2009: 1

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

# of Qbs over 30 Tds in 2008: 3

2007: 4

2006: 1

# you're projecting in 2009: 1

# of QBs over 575 attempts in 2008: 3

2007: 3

2006: 3

# you're projecting in 2009: 1

Mathematically speaking, the issue here is that the expected value of a maximum of a bunch of random variables is not the same as the maximum of the expected values of those same variables. It is in no way inconsistent to believe that someone will have 1500 receiving yards this year while not projecting any particular individual to do so.An example: every single week of NFL play for the last three years, some team has won a game by 20 or more points. Every single week. But Vegas almost never sets a line as high as 20. And that's because the questions "do you think some team will win by at least 20 this week?" and "is there a particular team you think will win by 20 this week?" usually have different answers.

very There are several players with higher ceilings than others and lower floors. What if you changed the FF projections on it's ear by applying a deviation variable to the stats to reflect the risk/reward for each player.

Example:

Pierre Thomas and Darren McFadden are ranked next to each other on DD projections. McFadden might be a higher risk/reward player.

Name					 Probability   Rush		  YD		 TD		REC	   YD	  TD16 Pierre Thomas								  1-20%	 175		   749		6		  7		 24	 0							 21-30%	 190		   850		8		  9		 30	 0							 31-70%	 200		   910		10		10		 35	  1							 71-93%	 215		   972		11		11		 39	  1							 94-99%	 240		  1080	   12		15		 44	  217 Darren McFadden 							  1-15%	145			616		4		 25		 200   0							 16-25%	 160			702		5		 33		265	1							 26-65%	 215			946		7		 40		320	1							 66-79%	 240		   1050	   7		 47		375	2							 80-95%	 260		   1144	   9		 48		382	2							 96-99%	 290		   1280	  12		50		 400	2

 

[Biabreakable]

Sorry you felt I was unclear. Your projections for QBs (and WRs/TEs) every year are Glenn Beck-levels of crazy conservative. They are useless in gauging actual performance because you generally understate your top performers by 15% or more than what the top performers of that year will end up at. To illustrate this point using your 2009 #s:

# of QBs in 2008 over or equal to 7.7 YPA: 10

2007: 7

2006: 6

# you're projecting in 2009: 1

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

# of Qbs over 30 Tds in 2008: 3

2007: 4

2006: 1

# you're projecting in 2009: 1

# of QBs over 575 attempts in 2008: 3

2007: 3

2006: 3

# you're projecting in 2009: 1

Mathematically speaking, the issue here is that the expected value of a maximum of a bunch of random variables is not the same as the maximum of the expected values of those same variables. It is in no way inconsistent to believe that someone will have 1500 receiving yards this year while not projecting any particular individual to do so.An example: every single week of NFL play for the last three years, some team has won a game by 20 or more points. Every single week. But Vegas almost never sets a line as high as 20. And that's because the questions "do you think some team will win by at least 20 this week?" and "is there a particular team you think will win by 20 this week?" usually have different answers.

very There are several players with higher ceilings than others and lower floors. What if you changed the FF projections on it's ear by applying a deviation variable to the stats to reflect the risk/reward for each player.

Example:

Pierre Thomas and Darren McFadden are ranked next to each other on DD projections. McFadden might be a higher risk/reward player.

Name Probability Rush YD TD REC YD TD16 Pierre Thomas 1-20% 175 749 6 7 24 0 21-30% 190 850 8 9 30 0 31-70% 200 910 10 10 35 1 71-93% 215 972 11 11 39 1 94-99% 240 1080 12 15 44 217 Darren McFadden 1-15% 145 616 4 25 200 0 16-25% 160 702 5 33 265 1 26-65% 215 946 7 40 320 1 66-79% 240 1050 7 47 375 2 80-95% 260 1144 9 48 382 2 96-99% 290 1280 12 50 400 2

Pretty interesting stuff here. Have you tested this?After I get a decent team then individual player projections I then start breaking the individual players into ranges. Similar to what MT was talking about of the static projection being a mid range of the projection. When I do this I can take into account if players at same position competing for playing time possible injuries or beating the other players out for more playing time. So a floor and ceiling. I don't have any percentages as far as what I think the probability for those ranges would be. I am not sure how I could answer that. How did you determine yours?

I think the important thing with projections is always making sure that each players piece of the pie stays consistent within the whole.

For example if McFadden gets 40 more carries then those carries need to be taken away from someone else. Or you need to consider increasing the total number of plays run which could happen if McFadden is performing that well.

 

[Tanner9919]

My first cut of these things

All criticism welcome here on these. I have about 40 hours into these numbers, but I am sure I have some things that don't make any sense at all.

Jason Wood has also released his numbers.

PK and Def projections will be coming within days and then we will move on to the VBD and DRaft Dominator (likely by the middle of next week)

wondering how you came to the conclusion that Portis is only good for the 8th best RB in the NFL this season?!with games against Rams, Lions, Bucs, Panthers, Chiefs, Falcons, Broncos, Saints, Raiders, I'm trying to understand how you could rank Portis any lower than MJD @ #2?!

the guy should easily rack up at least 140 rush yards against each of the Rams, Lions, Chiefs, Saints, Broncos..700 yards.

another 100 ( minimum) against Bucs, Panthers, Falcons..1000 yards so far..then at least 80 per game vs. Cowboys/Giants/Eagles. another 480 for a total of 1480 yards. and that is the low end of the stick. take a look at his averages against these three teams since he was traded to Washington...he does exteremely well in the NFC East..

the only way he finishes with less that 1500 rush yards this season, is if he gets hurt , something he never does. the guy is a rock..

I think his totals are closer to 1700 rushing yards in what could be his finest season in the NFL.

Stevey Jackson at #4 is way too high, IMO..you're asking Sjax to do some things he's only done once during his career: rush for more than 1200 yards in a single season, catch more than 43 balls, play a full 16-game schedule, injury free.

I highly doubt it happens. The guy runs upright, loves to deliver hits to opposing defenders, has a lousy O-line, and he plays for a bad team..not many RB's have finished a season as the 4th best RB while playing for a terrible, losing team..

winning NFL teams and RB's with great stats, usually go hand-in-hand.

Love the Slaton ranking although he's still probably too low! this guy is going to be a stud for Houston!

you might be too low on Brandon Jacobs, Forte, Ronnie Brown, Pierre Thomas, D. Ward.

and finally, I'd be shocked if Moreno finished with the statline you're expecting, 225/1013/8?

with 7 RB's on that Denver roster, a HC who is in love with RBBC, no QB to speak of, no defense, and a new coaching staff, I wonder how Moreno will even get 1/2 of those stats you're projecting him for, especially with games against Philly,NY, Balt, NE, Giants, SD (2x), Washington,Dallas, Pitt, Chicago...just how is he going to get 1,000 yards when he plays these defenses, not to mention the 8-man fronts he'll see all year long..

Denver is the one team most likely to finish in the bottom 5 in time of possession, rush yards, total yards from scrimmage,

wins, etc. They should easily be one of the 5 worst teams in the NFL.

took a quick look at your QB rankings, they look dead-on! you're a believer in Cutler in Chicago!

the ranking of Romo is perfect at #9..he might be even lower than that without T.O.

WR's look great, Bowe might be a tad low..he should finish with double-digit TDs.

 

[Liquid Tension]

In the 10 years I've been following you and Joe, I've yet to figure out why you refuse to make your QB projections to mirror anything close to reality.

I do not understand what you are even saying in your statement, but I will use this to speak to historical norms. I feel these are critical to make sure a set of projections is based in reality. Most other sites do not do this check and those numbers are usually outside expectations. Here are the league numbers I craft my projections around.Pass Completions:

2008 - 10,089

2007 - 10,425

2006 - 9,796

2009 Projected - 10,163

Pass Attempts:

2008 - 16,536

2007 - 17,045

2006 - 16,389

2009 Projected - 16,673

Pass Percentage:

2008 - 61.0%

2007 - 61.2%

2006 - 59.8%

2009 Projected - 60.0%

Passing Average:

2008 - 6.54

2007 - 6.86

2006 - 6.85

2009 Projected - 6.73

Passing TDs:

2008 - 646

2007 - 720

2006 - 648

2009 Projected - 695

Interceptions:

2008 - 465

2007 - 534

2006 - 520

2009 Projected - 496

Rush Attempts:

2008 - 14,110

2007 - 13,804

2006 - 14,324

2009 Projected - 14,153

Rush Yards:

2008 - 59,340

2007 - 56,768

2006 - 60,094

2009 Projected - 58,553

Rushing Average:

2008 - 4.21

2007 - 4.11

2006 - 4.20

2009 Projected - 4.14

Rushing TDs:

2008 - 476

2007 - 386

2006 - 423

2009 Projected - 454

Sorry you felt I was unclear. Your projections for QBs (and WRs/TEs) every year are Glenn Beck-levels of crazy conservative. They are useless in gauging actual performance because you generally understate your top performers by 15% or more than what the top performers of that year will end up at. To illustrate this point using your 2009 #s: # of QBs in 2008 over or equal to 7.7 YPA: 10

2007: 7

2006: 6

# you're projecting in 2009: 1

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

# of Qbs over 30 Tds in 2008: 3

2007: 4

2006: 1

# you're projecting in 2009: 1

# of QBs over 575 attempts in 2008: 3

2007: 3

2006: 3

# you're projecting in 2009: 1

Do you believe in regression to the mean?

Chase, this was not a productive answer. if you are not using the last years a regression to the mean is not statisdtically relevant? I am not sure where you are going here?

 

[Liquid Tension]

Do you believe in regression to the mean?

To which mean?

http://www.pro-football-reference.com/blog/?p=370I wrote that a couple of years ago; if you're not familiar with the concept, it might be worth a read. Here's a key passage:

The reasoning behind “regression to the mean” is iron-clad: when an impressive feat is hit, there’s a good bit of luck involved. Sometimes, it’s hit by someone who is actually as good as his stats (although this becomes less likely the more impressive the feat is). But other times it’s by a player who is a little lucky, and sometimes it’s by a player who’s really lucky.

Now NFL players aren’t computer programs or dice, but the same theory applies. And we see these results every year in the NFL. No one projects LaDainian Tomlinson to rush for 28 TDs again, because we know his true ability isn’t 28 TDs per season. To reach such a ridiculous result, a good bit of luck had to be involved. And regression to the mean becomes more likely in the NFL than when flipping a coin, because of strength of schedule. Many impressive feats involve general luck, and also luck due to facing an easy schedule. Every year, some team plays the easiest schedule in the league, and as a result, will achieve results they couldn’t normally achieve without a ton of luck. But since strength of schedule is incredibly inconsistent from year to year, we see this effect ride on top of regression to the mean to push down the great seasons. Because if you’re going to throw for 49 TDs in a season, you’ve got to be: a) awesome; b) have lots of luck; and c) have a really easy schedule. And only one of those traits is likely to be there the next season.

The point is, when a QB throws 600 times in a season, a lot of things that are unlikely to happen, happen. We shouldn't project unlikely things to happen, even if in the aggregate, it will happen a few times.

yeah, but if 3 years in a row something happened 4 times, then that is something that would appear to BE the mean or at least something that increase (or reduces) the previous mean.

 

[Liquid Tension]

Here is the part you simply cannot project but it is the core of what makes these projections so dangerous for redrafters...INJURIES.

What a revelation! NFL players get injured! Hey, wait- another revelation! You can't predict or project injuries by and large! Damn those projections!!!!!!!1

Really? It's a perfectly fair question or angle on the stats. 18 QB with 20+ TDs is too high. You did nothing to dispute it, you just decided to mock. Got it, thanks.

Last year 17 TEAMS passed for 20+ TDs, and four others passed for 18-19 TDs. Obviously several of those involved more than one QB due to the starter getting injured.How do you predict QB injuries? If you cannot, you have to give those TDs to the starter, or split them between two QBs on most of the teams. Neither solution works perfectly, but splitting them is especially useless, imo.

Now this is the interesting question if you ask me (and nobody did). When faced with expectation that a team will throw for 22 TD's, do you give all 22 to the starter or do you project the average injury (maybe QB's average 14 games a year) and project the starter to have 19 TD's and the backup 3 TD's? This may be a style or preference question?The way I try to look at it is on a per game basis. What I mean, is IF a player is healthy what would I expect the player to put up on a game by game basis. I can then decide who is more of an injury risk (if I choose to play that angle) and then get quality backups for the higher risk personnel.There really isn't a perfect solution, but it is very important to be clear which way you did it and manage accordingly. Obviously, in this case it is important to understand the way David did it as well.

 

[Liquid Tension]

Who do you think is most likely to lead the league in receiving yards? Calvin Johnson.Do you think the league leader in receiving yards will have more than 1400 yards? Yes.Do you think Calvin Johnson will have more than 1400 yards? No.

Excellent example Doug!

 

[Liquid Tension]

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

Even though each projection comes down to a single number, that number really represents a range.I'm projecting that Jake Delhomme will throw for 3400 yards this season (based on playing 15 games).

But that 3400 really represents a range -- it means I think he's pretty likely to throw for between 3200 and 3600 yards.

Most of the time, when a QB throws for over 4000 yards, it means that he hit the high end of any reasonable projected range for him. When I project a QB to throw for 3900 yards, it doesn't mean I don't think he can hit 4000. In fact, if I project a guy to throw for 3900 yards, it means I think he has a good chance to throw for over 4000 -- and roughly as good a chance to throw for fewer than 3800.

If some QBs hit the high ends of their expected ranges while others hit the low ends, the best-performing QB will likely outperform the midpoint of his range (i.e., his projection) even if the projections correctly identified the midpoint of each QB's range.

In other words, if actual results have greater variance than projected results, it doesn't mean that the projections were flawed.

If you roll four six-sided dice, it is more likely than not that at least one of the dice will land on 6. But the best projection for each in individual die is nonetheless 3.5. The variance of the actual results is greater than the variance of the mathematically correct projections.

Football players aren't dice, but the same general principle applies. If the projections are done right, expect the projections to have less variance than the eventual results.

Sorry about this, but your projection for each individual die would be either 1-6 as they have equal chances. 3.5 is only the average expected total of 6 rolls.

 

[Biabreakable]

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

Even though each projection comes down to a single number, that number really represents a range.I'm projecting that Jake Delhomme will throw for 3400 yards this season (based on playing 15 games).

But that 3400 really represents a range -- it means I think he's pretty likely to throw for between 3200 and 3600 yards.

Most of the time, when a QB throws for over 4000 yards, it means that he hit the high end of any reasonable projected range for him. When I project a QB to throw for 3900 yards, it doesn't mean I don't think he can hit 4000. In fact, if I project a guy to throw for 3900 yards, it means I think he has a good chance to throw for over 4000 -- and roughly as good a chance to throw for fewer than 3800.

If some QBs hit the high ends of their expected ranges while others hit the low ends, the best-performing QB will likely outperform the midpoint of his range (i.e., his projection) even if the projections correctly identified the midpoint of each QB's range.

In other words, if actual results have greater variance than projected results, it doesn't mean that the projections were flawed.

If you roll four six-sided dice, it is more likely than not that at least one of the dice will land on 6. But the best projection for each in individual die is nonetheless 3.5. The variance of the actual results is greater than the variance of the mathematically correct projections.

Football players aren't dice, but the same general principle applies. If the projections are done right, expect the projections to have less variance than the eventual results.

Sorry about this, but your projection for each individual die would be either 1-6 as they have equal chances. 3.5 is only the average expected total of 6 rolls.

I agree with this and it is one of the reasons I break players out into ranges.Good stuff LT!

 

[moleculo]

Sorry you felt I was unclear. Your projections for QBs (and WRs/TEs) every year are Glenn Beck-levels of crazy conservative. They are useless in gauging actual performance because you generally understate your top performers by 15% or more than what the top performers of that year will end up at. To illustrate this point using your 2009 #s:

# of QBs in 2008 over or equal to 7.7 YPA: 10

2007: 7

2006: 6

# you're projecting in 2009: 1

# of QBs over 4000 yards in 2008: 6

2007: 4

2006: 5

# you're projecting in 2009: 1

# of Qbs over 30 Tds in 2008: 3

2007: 4

2006: 1

# you're projecting in 2009: 1

# of QBs over 575 attempts in 2008: 3

2007: 3

2006: 3

# you're projecting in 2009: 1

Mathematically speaking, the issue here is that the expected value of a maximum of a bunch of random variables is not the same as the maximum of the expected values of those same variables. It is in no way inconsistent to believe that someone will have 1500 receiving yards this year while not projecting any particular individual to do so.An example: every single week of NFL play for the last three years, some team has won a game by 20 or more points. Every single week. But Vegas almost never sets a line as high as 20. And that's because the questions "do you think some team will win by at least 20 this week?" and "is there a particular team you think will win by 20 this week?" usually have different answers.

very There are several players with higher ceilings than others and lower floors. What if you changed the FF projections on it's ear by applying a deviation variable to the stats to reflect the risk/reward for each player.

Example:

Pierre Thomas and Darren McFadden are ranked next to each other on DD projections. McFadden might be a higher risk/reward player.

Name Probability Rush YD TD REC YD TD16 Pierre Thomas 1-20% 175 749 6 7 24 0 21-30% 190 850 8 9 30 0 31-70% 200 910 10 10 35 1 71-93% 215 972 11 11 39 1 94-99% 240 1080 12 15 44 217 Darren McFadden 1-15% 145 616 4 25 200 0 16-25% 160 702 5 33 265 1 26-65% 215 946 7 40 320 1 66-79% 240 1050 7 47 375 2 80-95% 260 1144 9 48 382 2 96-99% 290 1280 12 50 400 2

Ideally, this is a great idea. In practice though, I don't think it's workable. Its pretty tough getting singular projection values for ~300 players - much less breaking each of these guys down to probability.Here's a thought - do two sets of projections.

Set A assumes no injuries; starters get a lions share of carries, and WR's get a appropriate distributions of catches. Back-ups have relatively little value.

Set B assumes that all starters miss a significant portion of time - the amount of time can vary from person to person. in this scenario, all back-ups see an increased work-load as they will see a larger role.

Now, for team starters, set A defines the ceiling and set B defines the floor. For back-ups, set B defines the ceiling and set A defines the floor.

It's quite a bit more work, but could be interesting.

 

[Badm]

OK, obscure question but I noticed Drew Carter from oak is not listed. Is he toast? I thought he was coming back full strength this year.

 

[fatkid]

Jason and David, Donald Driver is projected to be under 1,000 yards for the first time in many years. He's posted 1K yards each of the last five seasons and been a Top 30 WR that entire time - yet his projections (and ADP) say otherwise. Do you project the Green Bay passing game to go down, other WRs to steal targets, or another reason for the diminished numbers?I'll hang up and listen.

I'll put in my guess as to why this is.Driver hit 1048 and 1012 the last two years playing 15 and 16 games. His YPG has gone like this the last 3 years.2006 - 802007 - 702008 - 63He is going to be 34 this season.The current projection has him at 55 YPG. Considering he is getting old, and their are talented young WR in GB, I don't see that as a surprise.

 

[fatkid]

A couple things I've noticed:1. Jacobs has had a 5.0 ypc both of the last 2 years on over 200 carries in both. You both have him projected to drop almost half a yard to 4.6 ypc. Granted, 4.6 ypc is still excellent, but is this more of a "5.0 ypc is hard to maintain" or do you see something that is going to cause them to be less effective?2. To David--AJ has been an absolute PPR machine the last 3 years. In 2 of the last 3 years, he's had 103 and 115 receptions. In 2007, when he was amazing but only played 9 games, he was on pace for over 106 receptions. Projecting him for only 85 receptions seems like a pretty substantial drop. Looking at the Houston team, you're projecting the QB's to throw 24 fewer attempts compared to 2008. Yet, you projected 9 fewer RB receptions, 7 fewer TE receptions, and 30 fewer WR receptions. You're also projecting a substantial drop in y/a for Schaub and the Texans passing game in general. Just curious what the thinking behind this is.

Here is my shot at this also1. LT has never averaged 5 YPC in consecutive years, much less 3 straight. Neither did Emmit Smith. In fact I haven't found a modern player who has averaged 5 YPC 3 streight years. OJ didn't do it, Barry Sanders didn't do it. The ONLY player I found who did this was Jim Brown. For Running backs, the YPC depends on their health plus the health of their Offensive Line and QB. So for the last 2 years everythng went right for Jacobs, and he pulled it off. Doesn't mean it should be viewed as a given.2. He has AJ ranked 6th, a shade behind Steve Smith and Reggie Wayne. I would take AJ ahead of Smith, but still 5th behind Fitz, Moss, Wayne, and megatron. He seems to be in the right spot for me. The argument you made for AJ can be made for the projections for any of the top WR.

 

[Biabreakable]

1. LT has never averaged 5 YPC in consecutive years, much less 3 straight. Neither did Emmit Smith. In fact I haven't found a modern player who has averaged 5 YPC 3 streight years. OJ didn't do it, Barry Sanders didn't do it. The ONLY player I found who did this was Jim Brown. For Running backs, the YPC depends on their health plus the health of their Offensive Line and QB. So for the last 2 years everythng went right for Jacobs, and he pulled it off. Doesn't mean it should be viewed as a given.

Actualy was looking at this the other day because DeAngelo Williams is in line to have over 5 ypc for the 3rd year in a row if he does.Marsall Faulk did this 3 times in a row. Sanders did it 4 times but not consecutively. LT twice but not consecutively.I dunno if I would be in a rush to compare Jacobs to these guys either but ypc isn't neccessary to be a good RB. Brandon Jacobs:

Year Age Tm Pos G GS Att Yds TD Lng Y/A Y/G A/G Rec Yds Y/R TD Lng R/G Y/G YScm RRTD Fmb 2005 23 NYG  16 0 38 99 7 21 2.6 6.2 2.4		99 7 1 2006 24 NYG  15 0 96 423 9 16 4.4 28.2 6.4 11 149 13.5 0 43 0.7 9.9 572 9 2 2007 25 NYG RB 11 9 202 1009 4 43 5.0 91.7 18.4 23 174 7.6 2 34 2.1 15.8 1183 6 5 2008 26 NYG RB 13 13 219 1089 15 44 5.0 83.8 16.8 6 36 6.0 0 9 0.5 2.8 1125 15 3 Career	55 22 555 2620 35 44 4.7 47.6 10.1 40 359 9.0 2 43 0.7 6.5 2979 37 11

Sorry had wrong numbers in here before.

 

[gianmarco]

A couple things I've noticed:1. Jacobs has had a 5.0 ypc both of the last 2 years on over 200 carries in both. You both have him projected to drop almost half a yard to 4.6 ypc. Granted, 4.6 ypc is still excellent, but is this more of a "5.0 ypc is hard to maintain" or do you see something that is going to cause them to be less effective?2. To David--AJ has been an absolute PPR machine the last 3 years. In 2 of the last 3 years, he's had 103 and 115 receptions. In 2007, when he was amazing but only played 9 games, he was on pace for over 106 receptions. Projecting him for only 85 receptions seems like a pretty substantial drop. Looking at the Houston team, you're projecting the QB's to throw 24 fewer attempts compared to 2008. Yet, you projected 9 fewer RB receptions, 7 fewer TE receptions, and 30 fewer WR receptions. You're also projecting a substantial drop in y/a for Schaub and the Texans passing game in general. Just curious what the thinking behind this is.

Here is my shot at this also1. LT has never averaged 5 YPC in consecutive years, much less 3 straight. Neither did Emmit Smith. In fact I haven't found a modern player who has averaged 5 YPC 3 streight years. OJ didn't do it, Barry Sanders didn't do it. The ONLY player I found who did this was Jim Brown. For Running backs, the YPC depends on their health plus the health of their Offensive Line and QB. So for the last 2 years everythng went right for Jacobs, and he pulled it off. Doesn't mean it should be viewed as a given.2. He has AJ ranked 6th, a shade behind Steve Smith and Reggie Wayne. I would take AJ ahead of Smith, but still 5th behind Fitz, Moss, Wayne, and megatron. He seems to be in the right spot for me. The argument you made for AJ can be made for the projections for any of the top WR.

Again, I'm not saying he should be PROJECTED to have a 5.0 ypc. I'm simply stating that 4.6 is slightly on the lower side and think he deserves a bump up to the 4.7/4.8 range like a couple other backs. It's nitpicking, but his situation hasn't changed for the worse as far as I can tell and it seems he should be projected slightly higher.As to the 2nd, it's not about where he's ranked. But that's a SIGNIFICANT drop in receptions (almost 20%) for a guy who is a target/reception machine.

 

[Lott's Fingertip]

My first cut of these things

All criticism welcome here on these. I have about 40 hours into these numbers, but I am sure I have some things that don't make any sense at all.

Jason Wood has also released his numbers.

PK and Def projections will be coming within days and then we will move on to the VBD and DRaft Dominator (likely by the middle of next week)

wondering how you came to the conclusion that Portis is only good for the 8th best RB in the NFL this season?!with games against Rams, Lions, Bucs, Panthers, Chiefs, Falcons, Broncos, Saints, Raiders, I'm trying to understand how you could rank Portis any lower than MJD @ #2?!

the guy should easily rack up at least 140 rush yards against each of the Rams, Lions, Chiefs, Saints, Broncos..700 yards.

another 100 ( minimum) against Bucs, Panthers, Falcons..1000 yards so far..then at least 80 per game vs. Cowboys/Giants/Eagles. another 480 for a total of 1480 yards. and that is the low end of the stick. take a look at his averages against these three teams since he was traded to Washington...he does exteremely well in the NFC East..

the only way he finishes with less that 1500 rush yards this season, is if he gets hurt , something he never does. the guy is a rock..

I think his totals are closer to 1700 rushing yards in what could be his finest season in the NFL.

Stevey Jackson at #4 is way too high, IMO..you're asking Sjax to do some things he's only done once during his career: rush for more than 1200 yards in a single season, catch more than 43 balls, play a full 16-game schedule, injury free.

I highly doubt it happens. The guy runs upright, loves to deliver hits to opposing defenders, has a lousy O-line, and he plays for a bad team..not many RB's have finished a season as the 4th best RB while playing for a terrible, losing team..

winning NFL teams and RB's with great stats, usually go hand-in-hand.

Love the Slaton ranking although he's still probably too low! this guy is going to be a stud for Houston!

you might be too low on Brandon Jacobs, Forte, Ronnie Brown, Pierre Thomas, D. Ward.

and finally, I'd be shocked if Moreno finished with the statline you're expecting, 225/1013/8?

with 7 RB's on that Denver roster, a HC who is in love with RBBC, no QB to speak of, no defense, and a new coaching staff, I wonder how Moreno will even get 1/2 of those stats you're projecting him for, especially with games against Philly,NY, Balt, NE, Giants, SD (2x), Washington,Dallas, Pitt, Chicago...just how is he going to get 1,000 yards when he plays these defenses, not to mention the 8-man fronts he'll see all year long..

Denver is the one team most likely to finish in the bottom 5 in time of possession, rush yards, total yards from scrimmage,

wins, etc. They should easily be one of the 5 worst teams in the NFL.

took a quick look at your QB rankings, they look dead-on! you're a believer in Cutler in Chicago!

the ranking of Romo is perfect at #9..he might be even lower than that without T.O.

WR's look great, Bowe might be a tad low..he should finish with double-digit TDs.

Re: PortisHe hasn't rushed for 1500 yards since 2005, but he should be projected for a minimum of that now? Nothing wrong with a 1300 yard projection.

Re: S. Jackson

It boils down to this: nobody can predict injuries. Jackson has only played in 16 games once in 5 years, then 15, 14, 12, and 12 games. For comparison, let's look at your iron man, Portis, in his first 5 seasons: 16, 13, 15, 16, 8. Not that different, huh? Heck, Jackson has played in more games through 5 seasons than Portis did!

Jackson produced 1400 yards and 8 TDs in 12 games on a wretched team with a wretched line. That was good for #3 RB on a ppg basis... so if you can't predict injuries, and a guy produces at a very high level when he plays, where should he be ranked? Ranking Jackson out of the top 5 or 6 is just another way of saying he is definitely getting hurt. You may believe that, but there are MANY examples of players that had a few injuries that went on to have numerour injury-free seasons.

And your statement about his # of catches is absurd. Cherry pick stats much? What is magical about 43? He has caught "over" 37 every year as a starter.Jackson caught 40 and 38 balls in 12 games each of the last two years... but he won't catch 43 this year?

Heck, Jackson had 1400 combined yards in 12 games last year.... Dodds is only projecting about 1600 this year. Jackson may well reach that in 13-14 games, let alone 16 games.

All that aside, less than 2 ppg separate the #4 RB from the #11 RB. Don't get caught up in the rankings, pay attention to the projections.

"you might be too low on Brandon Jacobs, Forte, Ronnie Brown, Pierre Thomas, D. Ward."

Okay, so Jackson gets pummeled because he has had two different injuries in the last two years.... but Jacobs doesn't?!?!? Jackson gets a thumbs-down due to running style, but not Jacobs? They had almost identical ppg in 2007, and Jackson had a higher ppg last year.

It's a good thing Ronnie Brown doesn't miss games with injuries (1 full season out of 4).

It's a good thing Ward doesn't get injured. Ward's injuries

Apply the same logic to all the players and don't just give your personal favorites a pass for reasons that you condemn other players for.

 

[Maurile Tremblay]

1. LT has never averaged 5 YPC in consecutive years, much less 3 straight. Neither did Emmit Smith. In fact I haven't found a modern player who has averaged 5 YPC 3 streight years. OJ didn't do it, Barry Sanders didn't do it. The ONLY player I found who did this was Jim Brown. For Running backs, the YPC depends on their health plus the health of their Offensive Line and QB. So for the last 2 years everythng went right for Jacobs, and he pulled it off. Doesn't mean it should be viewed as a given.

Just as a matter of trivia -- Michael Turner averaged over 5 YPC for three straight years. Clinton Portis probably would have (IMO) if he'd stayed in Denver for his third year.(I've got Jacobs down for a 4.6 YPC this season. His career average is 4.7, and I think he's performed above his own personal long-term mean so far, not below it.)

 

[Maurile Tremblay]

Even though each projection comes down to a single number, that number really represents a range.

I'm projecting that Jake Delhomme will throw for 3400 yards this season (based on playing 15 games).

But that 3400 really represents a range -- it means I think he's pretty likely to throw for between 3200 and 3600 yards.

Most of the time, when a QB throws for over 4000 yards, it means that he hit the high end of any reasonable projected range for him. When I project a QB to throw for 3900 yards, it doesn't mean I don't think he can hit 4000. In fact, if I project a guy to throw for 3900 yards, it means I think he has a good chance to throw for over 4000 -- and roughly as good a chance to throw for fewer than 3800.

If some QBs hit the high ends of their expected ranges while others hit the low ends, the best-performing QB will likely outperform the midpoint of his range (i.e., his projection) even if the projections correctly identified the midpoint of each QB's range.

In other words, if actual results have greater variance than projected results, it doesn't mean that the projections were flawed.

If you roll four six-sided dice, it is more likely than not that at least one of the dice will land on 6. But the best projection for each in individual die is nonetheless 3.5. The variance of the actual results is greater than the variance of the mathematically correct projections.

Football players aren't dice, but the same general principle applies. If the projections are done right, expect the projections to have less variance than the eventual results.

Sorry about this, but your projection for each individual die would be either 1-6 as they have equal chances. 3.5 is only the average expected total of 6 rolls.

It depends on the purpose of my projection.The "expectation" for each individual die is 3.5.

If I were going to roll a six-sided die once and pay you a number of dollars equal to the result, how much would that offer be worth? The value of the offer would be $3.50. That is the mean roll.

If we play a different game -- say, you're going to guess what I roll and if you get it right, I'll pay you $6 but if you get it wrong, you pay me $1 -- you should never guess 3.5. That will never be right. You should guess 2, or any other specific number between 1 and 6.

Fantasy football is a bit like a dice game in that, for each player, there is a range of possible outcomes he may achieve during the season.

There's a discussion of Brandon Jacobs' YPC in this thread. I mentioned that I've got him projected for 4.6 YPC this season (actually 4.58). That's actually based on a range for what I think his 2009 YPC might be if he's given an arbitrarily large number of carries (i.e., his "true" 2009 YPC not affected by anomalies that might be caused by a small sample size).

5.1 - 0.4915%

5.0 - 1.9213%

4.9 - 6.0802%

4.8 - 11.6822%

4.7 - 16.1511%

4.6 - 20.3354%

4.5 - 18.4234%

4.4 - 13.3006%

4.3 - 7.3324%

4.2 - 3.2528%

4.1 - 0.8305%

4.0 - 0.1681%

3.9 - 0.0268%

3.8 - 0.0033%

If you multiply all those out, you get a mean expectation of 4.58 yards per carry. (And yes, I do that for every player. )

For fantasy football purposes -- should we do our projection more like the dice game where the answer was $3.5 (i.e., the mean of the range weighted by probability), or more like the dice game where we just picked some arbitrary number within the range?

I'd actually argue that drafting according to the mean expectation of each player's range of fantasy points is not optimal since it makes us less risk-averse than we should be in the early rounds and more risk-averse than we should be in the later rounds.

Be that as it may, when I do my projections, I am giving the mean expectation for each player's range of performance possibilities. The answer can be a fraction even if it's impossible to score 7.35 touchdowns in a season. And -- back to the original point of this tangent -- the variance of the mean expectations for each player will be less than the variance of the actual results.

A die can land from anywhere between 1 and 6, but the mean expectation is always 3.5.

I think Brandon Jacobs is reasonably likely to average anywhere between 4.3 and 4.9 YPC this season, but the mean expectation of my projections is 4.58.

 

[moleculo]

Even though each projection comes down to a single number, that number really represents a range.

I'm projecting that Jake Delhomme will throw for 3400 yards this season (based on playing 15 games).

But that 3400 really represents a range -- it means I think he's pretty likely to throw for between 3200 and 3600 yards.

Most of the time, when a QB throws for over 4000 yards, it means that he hit the high end of any reasonable projected range for him. When I project a QB to throw for 3900 yards, it doesn't mean I don't think he can hit 4000. In fact, if I project a guy to throw for 3900 yards, it means I think he has a good chance to throw for over 4000 -- and roughly as good a chance to throw for fewer than 3800.

If some QBs hit the high ends of their expected ranges while others hit the low ends, the best-performing QB will likely outperform the midpoint of his range (i.e., his projection) even if the projections correctly identified the midpoint of each QB's range.

In other words, if actual results have greater variance than projected results, it doesn't mean that the projections were flawed.

If you roll four six-sided dice, it is more likely than not that at least one of the dice will land on 6. But the best projection for each in individual die is nonetheless 3.5. The variance of the actual results is greater than the variance of the mathematically correct projections.

Football players aren't dice, but the same general principle applies. If the projections are done right, expect the projections to have less variance than the eventual results.

Sorry about this, but your projection for each individual die would be either 1-6 as they have equal chances. 3.5 is only the average expected total of 6 rolls.

It depends on the purpose of my projection.The "expectation" for each individual die is 3.5.

If I were going to roll a six-sided die once and pay you a number of dollars equal to the result, how much would that offer be worth? The value of the offer would be $3.50. That is the mean roll.

If we play a different game -- say, you're going to guess what I roll and if you get it right, I'll pay you $6 but if you get it wrong, you pay me $1 -- you should never guess 3.5. That will never be right. You should guess 2, or any other specific number between 1 and 6.

Fantasy football is a bit like a dice game in that, for each player, there is a range of possible outcomes he may achieve during the season.

There's a discussion of Brandon Jacobs' YPC in this thread. I mentioned that I've got him projected for 4.6 YPC this season (actually 4.58). That's actually based on a range for what I think his 2009 YPC might be if he's given an arbitrarily large number of carries (i.e., his "true" 2009 YPC not affected by anomalies that might be caused by a small sample size).

5.1 - 0.4915%

5.0 - 1.9213%

4.9 - 6.0802%

4.8 - 11.6822%

4.7 - 16.1511%

4.6 - 20.3354%

4.5 - 18.4234%

4.4 - 13.3006%

4.3 - 7.3324%

4.2 - 3.2528%

4.1 - 0.8305%

4.0 - 0.1681%

3.9 - 0.0268%

3.8 - 0.0033%

If you multiply all those out, you get a mean expectation of 4.58 yards per carry. (And yes, I do that for every player. )

For fantasy football purposes -- should we do our projection more like the dice game where the answer was $3.5 (i.e., the mean of the range weighted by probability), or more like the dice game where we just picked some arbitrary number within the range?

I'd actually argue that using the mean expectation of each player's range of fantasy points is not optimal since it makes us less risk-averse than we should be in the early rounds and more risk-averse than we should be in the later rounds.

Be that as it may, when I do my projections, I am giving the mean expectation for each player's range of performance possibilities. The answer can be a fraction even if it's impossible to score 7.35 touchdowns in a season. And -- back to the original point of this tangent -- the variance of the mean expectations for each player will be less than the variance of the actual results.

A die can land from anywhere between 1 and 6, but the mean expectation is always 3.5.

I think Brandon Jacobs is reasonably likely to average anywhere between 4.3 and 4.8 YPC this season, but the mean expectation of my projections is 4.58.

well, the next bit of critical info then is how wide the ranges are. What you have above is (hopefully) a normally distributed set of data centered around 4.58 YPC which is great. I'm sure that there is a corresponding sigma with this distribution as well. But, not everyone will have an equivalent sigma, which means that the effective floors and ceilings for all of the projections will vary. This variation could also be modeled and accounted for as a risk factor somehow.

 

[Maurile Tremblay]

well, the next bit of critical info then is how wide the ranges are. What you have above is (hopefully) a normally distributed set of data centered around 4.58 YPC which is great. I'm sure that there is a corresponding sigma with this distribution as well. But, not everyone will have an equivalent sigma, which means that the effective floors and ceilings for all of the projections will vary. This variation could also be modeled and accounted for as a risk factor somehow.

Yes, Brandon Jacobs has a wider range than LaDainian Tomlinson (projected mean = 4.50):4.9 - 0.0334%4.8 - 0.7751%4.7 - 6.6109%4.6 - 26.2005%4.5 - 38.1229%4.4 - 22.5529%4.3 - 5.1982%4.2 - 0.4919%4.1 - 0.0137%Tomlinson has 2657 career rushes, so we pretty much know what we're getting with him (if he maintains his health). Jacobs has 554 rushes, which is enough to conclude that he is pretty certainly not a complete fluke, but we still don't know exactly what we're getting with the same level of certainty (in both directions -- upside and downside). So he's got a wider range.(Note that those percentages aren't supposed to describe the possibilities for 2009's 300 or so carries -- that would have a larger range. Rather, they're supposed to describe his YPC in 2009 if he got a number of carries approaching infinity without wearing down, i.e., if we ran the 2009 season a few hundred thousand times, what would his overall average be?)

 

[gianmarco]

Even though each projection comes down to a single number, that number really represents a range.

I'm projecting that Jake Delhomme will throw for 3400 yards this season (based on playing 15 games).

But that 3400 really represents a range -- it means I think he's pretty likely to throw for between 3200 and 3600 yards.

Most of the time, when a QB throws for over 4000 yards, it means that he hit the high end of any reasonable projected range for him. When I project a QB to throw for 3900 yards, it doesn't mean I don't think he can hit 4000. In fact, if I project a guy to throw for 3900 yards, it means I think he has a good chance to throw for over 4000 -- and roughly as good a chance to throw for fewer than 3800.

If some QBs hit the high ends of their expected ranges while others hit the low ends, the best-performing QB will likely outperform the midpoint of his range (i.e., his projection) even if the projections correctly identified the midpoint of each QB's range.

In other words, if actual results have greater variance than projected results, it doesn't mean that the projections were flawed.

If you roll four six-sided dice, it is more likely than not that at least one of the dice will land on 6. But the best projection for each in individual die is nonetheless 3.5. The variance of the actual results is greater than the variance of the mathematically correct projections.

Football players aren't dice, but the same general principle applies. If the projections are done right, expect the projections to have less variance than the eventual results.

Sorry about this, but your projection for each individual die would be either 1-6 as they have equal chances. 3.5 is only the average expected total of 6 rolls.

It depends on the purpose of my projection.The "expectation" for each individual die is 3.5.

If I were going to roll a six-sided die once and pay you a number of dollars equal to the result, how much would that offer be worth? The value of the offer would be $3.50. That is the mean roll.

If we play a different game -- say, you're going to guess what I roll and if you get it right, I'll pay you $6 but if you get it wrong, you pay me $1 -- you should never guess 3.5. That will never be right. You should guess 2, or any other specific number between 1 and 6.

Fantasy football is a bit like a dice game in that, for each player, there is a range of possible outcomes he may achieve during the season.

There's a discussion of Brandon Jacobs' YPC in this thread. I mentioned that I've got him projected for 4.6 YPC this season (actually 4.58). That's actually based on a range for what I think his 2009 YPC might be if he's given an arbitrarily large number of carries (i.e., his "true" 2009 YPC not affected by anomalies that might be caused by a small sample size).

5.1 - 0.4915%

5.0 - 1.9213%

4.9 - 6.0802%

4.8 - 11.6822%

4.7 - 16.1511%

4.6 - 20.3354%

4.5 - 18.4234%

4.4 - 13.3006%

4.3 - 7.3324%

4.2 - 3.2528%

4.1 - 0.8305%

4.0 - 0.1681%

3.9 - 0.0268%

3.8 - 0.0033%

If you multiply all those out, you get a mean expectation of 4.58 yards per carry. (And yes, I do that for every player. )

For fantasy football purposes -- should we do our projection more like the dice game where the answer was $3.5 (i.e., the mean of the range weighted by probability), or more like the dice game where we just picked some arbitrary number within the range?

I'd actually argue that drafting according to the mean expectation of each player's range of fantasy points is not optimal since it makes us less risk-averse than we should be in the early rounds and more risk-averse than we should be in the later rounds.

Be that as it may, when I do my projections, I am giving the mean expectation for each player's range of performance possibilities. The answer can be a fraction even if it's impossible to score 7.35 touchdowns in a season. And -- back to the original point of this tangent -- the variance of the mean expectations for each player will be less than the variance of the actual results.

A die can land from anywhere between 1 and 6, but the mean expectation is always 3.5.

I think Brandon Jacobs is reasonably likely to average anywhere between 4.3 and 4.9 YPC this season, but the mean expectation of my projections is 4.58.

Interesting to note you have Jacobs at a higher % to have 4.3 ypc than 4.9 ypc (and likewise higher % for each corresponding drop from 4.6 ypc vs the same bump from 4.6 ypc) despite being an incredibly effective runner and no indication it should go down at this point. I guess it depends on how you view Jacobs and whether or not he outperformed his true talent/average the last 2 years or are they are a true indicator of what he's able to do. After just 1 year, I'd tend to agree and would even bump it down a little. After 2 years of continued consistent running, I think we may be closer to seeing what he's capable of doing behind the NYG O-line. I was nervous about him going into this last year. I'm not nearly as much heading into 2009 and think he presents one of the best values at RB this year.

 

[Maurile Tremblay]

Interesting to note you have Jacobs at a higher % to have 4.3 ypc than 4.9 ypc (and likewise higher % for each corresponding drop from 4.6 ypc vs the same bump from 4.6 ypc) despite being an incredibly effective runner and no indication it should go down at this point.

There are several indications that it should go down, IMO.1. He averaged fewer than 4.6 yards per carry in 2005 and 2006. I give 2007 and 2008 a lot more weight, but I don't think we should discount 2005 and 2006 entirely.

2. It is very rare for an NFL RB to average more than 4.6 yards per carry over an extended period. Here's the list of active players with the highest YPCs. Ignore the QBs. The RBs are:

Adrian Peterson - 601

DeAngelo Williams - 538

Michael Turner - 604

Tatum Bell - 569

Maurice Jones-Drew - 530

Frank Gore - 939

Brandon Jacobs - 555

I put their career rush attempts next to their names. Note that, with the exception of Gore, they are all down around 600 carries or less -- roughly two seasons' worth of a full-time load.

Exceptionally high YPCs can occur over a small sample size, but as the sample size gets larger it tends to get less anomalous (in terms of falling back in line with league-wide norms). A career YPC over 4.6 is somewhat anomalous, and for that reason alone is likely to regress to the NFL mean as the sample size increases.

If a running back goes for 13 yards on his first career carry, nobody starts thinking about how awesome his YPC is, as if it's not going to drop back down soon.

It's the same principle when a running back averages 4.7 yards per carry on his first 600-ish carries, although to a much lesser degree. He's still more likely to go down than up just based on the rarity of anyone whose true long-term YPC is above that, compared with the extreme commonness of RBs whose true YPC is below that.

 

[Liquid Tension]

1. LT has never averaged 5 YPC in consecutive years, much less 3 straight. Neither did Emmit Smith. In fact I haven't found a modern player who has averaged 5 YPC 3 streight years. OJ didn't do it, Barry Sanders didn't do it. The ONLY player I found who did this was Jim Brown. For Running backs, the YPC depends on their health plus the health of their Offensive Line and QB. So for the last 2 years everythng went right for Jacobs, and he pulled it off. Doesn't mean it should be viewed as a given.

Actualy was looking at this the other day because DeAngelo Williams is in line to have over 5 ypc for the 3rd year in a row if he does.Marsall Faulk did this 3 times in a row. Sanders did it 4 times but not consecutively. LT twice but not consecutively.I dunno if I would be in a rush to compare Jacobs to these guys either but ypc isn't neccessary to be a good RB. Brandon Jacobs:

Rushing Receiving   Year Age Tm Pos G GS Att Yds TD Lng Y/A Y/G A/G Rec Yds Y/R TD Lng R/G Y/G YScm RRTD Fmb 2007 22 GNB  11 3 75 267 1 46 3.6 24.3 6.8 16 130 8.1 0 16 1.5 11.8 397 1 0 2008 23 GNB  13 0 45 248 1 32 5.5 19.1 3.5 30 185 6.2 0 18 2.3 14.2 433 1 1

Not to Hijak, but it is pretty amazing that Jacobs YPC is so high when you consider that he gets most of the short yardage carries. He is a weapon for an NFL team and still good for fantasy.

 

[benbadman]

I counted 18 QBs you have slated for 20+ TDs...in 2008:11, 2007:13, 2006:10...I'll take the under as I am not sure there has ever been a season with 18 QBs tossing 20+ TDs, and along those lines I have to think this skews the WRs numbers a bit across the board. Sorry to be such a wet rag, again DD does a spot on job of these and I am more inclined to peek at his final stats than just about anyone else but projections are a very dangerous game IMO.

18 teams having the QB position pass for 20+ TDs seems reasonable. You cant predict 14 TD for Schaub and 6 TD for his backup.... injuries play a part.