players to pay attention to (1 Viewer)

[moleculo]

Instead of doing full blown projections this year like I have in the past, my intention is to leverage other peoples projections as much as I can and do what I can to improve where I can. My intent is to basically scour the web for at least five free, independent sets of projections, average them, and use that to generate my ranking list. Additionally, I will be tweaking the aggregate projections as I see fit.

The tweaking, IMO, is most appropriate for players that there is a high degree of disagreement amongst the various projectors. If 5 independant people think player X will score 290-310 FF points, I don't see any reason for me to spend much time thinking about player X. However, if these same 5 people think player Y will score between 150 and 250 FF points, this is a player that I should be thinking about and generating my own projections.

I have found five sites that offer free projections:

[*]CBSsports.com

[*]ESPN.com

[*]NFL.com

[*]FFToday.com

[*]FantasySharks.com

(note to mods: if you don't want competition web sites mentioned, feel free to edit or let me know and I will remove).

Projections from these sites have all been compiled @ fantasyrundown.com, and placed into a nice excel file. I have created a process whereby I can and evaluate projected scoring fairly easy, and continually re-evaluate as the pre-season progresses. I have calculated average scoring and standard deviations between the above 5 projection sites.

To determine who to keep an eye on, if you simply use stdev, you end up with lots of bottom tier guys. IMO it's more important to pay attention to guys at the top of the pecking order, but stdev is really what I want to know about. So, I created a variable called "interest score", which is basically average*stdev^2: standard deviation is more important than average, but average is important to consider. Again, this is just a way to make a list top heavy so I don't waste time focusing on Ryan Fitzpatrick and his ilk. My intent is to pay close attention to the top most interesting players at each position: 4 QB's, 8 RB's, 12 WR's, and 4 TE's.

The top players to watch are as follows, with some comments

7/26/2010

QB1 Orton, Kyle DEN - unclear how to project the Denver O.

QB2 Rodgers, Aaron GB - he's actually fairly stable, but his average score is so high that he ends up here.

QB3 Delhomme, Jake CLE

QB4 Hasselbeck, Matt SEA

RB1 Young, Albert MIN - he's high because of ESPN.com. Haven't read anything about him. obviously they are thinking he fills Chester Taylors shoes.

RB2 Rice, Ray BAL

RB3 Gore, Frank SF

RB4 Washington, Leon SEA positional battle

RB5 Forsett, Justin SEA positional battle

RB6 Forte, Matt CHI positional battle

RB7 Jackson, Fred BUF

RB8 Taylor, Chester CHI positional battle

WR1 Welker, Wes NE - injury?

WR2 Jackson, Vincent SD - suspension/holdout/trade?

WR3 Owens, Terrell FA - team?

WR4 Williams, Mike SEA

WR5 Gonzalez, Anthony IND

WR6 Randle El, Antwaan PIT

WR7 Price, Taylor NE

WR8 Knox, Johnny CHI

WR9 Robinson, Laurent STL

WR10 Holmes, Santonio NYJ

WR11 Caldwell, Andre CIN

WR12 Hester, Devin CHI

TE1 Clark, Dallas IND

TE2 Daniels, Owen HOU

TE3 Davis, Vernon SF

TE4 Winslow, Kellen TB

I expect there to be some convergence as we move closer to the start of the season - I will be updating my watch list accordingly. Again, this really should be considered an attention directing tool. These represent guys where there seems to be significant differences in opinion of 2010 production.

 

[hook014]

Great premise and good list. Black dot for draft prep.

 

[Captain Hook]

I appreciate the research and work and think this is a very worthwhile effort

That said, I don't think Rodgers should be on the list - he is universally rated as one of if not the top QB everyplace, so to me whatever the averages are don't matter - it's not like you are going to change your QB list

And I think Albert Young is pretty far off the radar unless people are playing in 32 team leagues....not even clear he beats out Ryan Moats for whatever relief duty the Vikings might have and ESPN has 0 TD projected for him, so not sure he would make a watch list

 

[moleculo]

I appreciate the research and work and think this is a very worthwhile effortThat said, I don't think Rodgers should be on the list - he is universally rated as one of if not the top QB everyplace, so to me whatever the averages are don't matter - it's not like you are going to change your QB listAnd I think Albert Young is pretty far off the radar unless people are playing in 32 team leagues....not even clear he beats out Ryan Moats for whatever relief duty the Vikings might have and ESPN has 0 TD projected for him, so not sure he would make a watch list

yeah, I think I need to tweak some stuff to get Rodgers off of the list. The thing to note is that i'm not interested in rankings; I'm interested in projected scores. Rankings are useless when comparing across positions - is it better to draft QB1 or RB6 towards the end of the first round, for example. For that reason, it's pretty good to have a good estimation of projected scoring, and that's why I focus on projections only.Regarding Albert young - as we get into training camp and roles become defined, I expect most projectors to fall into line and projections should start to converge. Albert Young is high because ESPN projects him w/ 79 points (PPR), and CBS has him @ 7 points - a huge swing. The difference is that ESPN has him w/ 40 receptions - maybe a typo, maybe they expect him to be a 3rd down back and catch a lot of dumpoffs - either way, it's worth noting.

 

[moleculo]

updated list (data per 8/21)

QB1 Tebow, Tim DEN

QB2 Croyle, Brodie KC

QB3 Schaub, Matt HOU

QB4 Rivers, Philip SD

RB1 Rice, Ray BAL

RB2 Jackson, Fred BUF

RB3 Forsett, Justin SEA

RB4 Forte, Matt CHI

RB5 Bradshaw, Ahmad NYG

RB6 Gore, Frank SF

RB7 Charles, Jamaal KC

RB8 Bush, Reggie NO

WR1 Collie, Austin IND

WR2 Jackson, Vincent SD

WR3 Smith, Steve NYG

WR4 Marshall, Brandon MIA

WR5 Welker, Wes NE

WR6 Bryant, Antonio CIN

WR7 Rice, Sidney MIN

WR8 Knox, Johnny CHI

WR9 Galloway, Joey WAS

WR10 Johnson, Andre HOU

WR11 Moss, Randy NE

WR12 Boldin, Anquan BAL

TE1 Daniels, Owen HOU

TE2 Gates, Antonio SD

TE3 Cooley, Chris WAS

TE4 Davis, Vernon SF

edit: modigied list. One set of projections (fantasysharks) was so off it was gumming everything up, especially at the top end. I tossed that data out, and I'm liking this a little better.

 

[TWP]

Just becuase a projection is 'consensus' doesnt mean it's right. I like the idea, but I'd also look over, say, the top 50 guys from your consensus list and see if there are people that you feel are way out of whack. For instance, I was way down on Forte last year even though the consensus was that he was a top pick. There's a ton of group think on internet sites...

 

[gheemony]

Moleculo, you and I are thinking along the same lines. I take a bunch of projections and the use excel to plot them using the stock graph type of chart. Great way to see volatility in projections.

For your project, I think you should take standard deviation/mean. Find which players have then biggest variation. To decide who makes the list, the players mean+standard deviation should put them above some relevant cut off (starter, first bench, rosterable). This should cut out Albert Young because he's probably never roster-worthy.

Questions (PM me if easier): how do you import the stats (I found pulling some data is hard on formatting), how do you match up for each player (sites may use different names (Chris Wells v Beanie Wells), do you format in a way that allows you to change scoring or do you just import fantasy points, and do you have a way to do live updating (I don't have enough time to update as frequently as some sites do)?

If there was an easy Import tool, I know I could have done a lot more analysis. But spending too much time on data clean up and formatting.

 

[moleculo]

Moleculo, you and I are thinking along the same lines. I take a bunch of projections and the use excel to plot them using the stock graph type of chart. Great way to see volatility in projections.For your project, I think you should take standard deviation/mean. Find which players have then biggest variation. To decide who makes the list, the players mean+standard deviation should put them above some relevant cut off (starter, first bench, rosterable). This should cut out Albert Young because he's probably never roster-worthy.Questions (PM me if easier): how do you import the stats (I found pulling some data is hard on formatting), how do you match up for each player (sites may use different names (Chris Wells v Beanie Wells), do you format in a way that allows you to change scoring or do you just import fantasy points, and do you have a way to do live updating (I don't have enough time to update as frequently as some sites do)?If there was an easy Import tool, I know I could have done a lot more analysis. But spending too much time on data clean up and formatting.

gheemoney - let me get back to you this evening - I want to spend some time on this, and I'm blocked from FBG @ work.

 

[Ministry of Pain]

 

[gheemony]

Moleculo - take your time, we have lots of it. Would be happy to trade email about what I do and share formatted stats (though I use pay content: FBG and KUBIAK). I do import from CBS. I site I would like to import from is fantasydc.com. Mike Clay appears to do good projections. At least good process and he shows his work.

 

[Tigerbot Hesh]

 

[moleculo]

So the night before the draft, I downloaded the compiled data from fantasyrundown.com and discovered that his data was practically worthless - projections that were supposed to be updated weren't, there were typos all over, etc. pure garbage. I had to go through the effort of manually downloading all of the various projections, formatting them so I could read them in, compiling, and then finally tweaking as I see fit. Not a small exercise to do the night before a draft. Major PITA. Had I realized that fantasyrundown.com would be useless, I would have set my stuff up differently and it would have been soo much easier. Beyond that, fantsysharks.com's projections were garbage, as were NFL.com. So, the data set I used included Eisenberg (CBS.com), Richards (CBS.com), ESPN.com, FFToday.com, and unlucky (FBG member who makes his projections public).

The purpose of this whole exercise was to cut down on the work to get good projections, not complicate things - scrambling was not what I wanted to do the night before the draft. But, that's neither here nor there...more venting than anything else.

For "interest score" - that is, to determine which players I needed to look at closer, I siimply multiplied the mean with the standard deviation. I don't know how or why I came up with that, but it seemed to do a good job of weighing out the high variability / high-scoring players that I wanted to focus on. My final list (data current as of 8/28) was as follows:

Leinart, Matt

Brees, Drew

Favre, Brett

Delhomme, Jake

Johnson, Chris

Peterson, Adrian

Rice, Ray

Washington, Leon

Forsett, Justin

Peterson, Adrian

Forte, Matt

Mathews, Ryan

Williams, Mike

Jackson, Vincent

Boldin, Anquan

Austin, Miles

Welker, Wes

Marshall, Brandon

Holmes, Santonio

Murphy, Louis

Moss, Santana

Crabtree, Michael

Gibson, Brandon

Doucet, Early

Cooley, Chris

Clark, Dallas

Gates, Antonio

Daniels, Owen

again, this list represents the highest scoring players with the least amount of agreement according to the projections I considered. If course, there were other players I considered too...those were basically guys where my ranking differed the most as compared to the most recently published ADP.

For the rest of your questions (gheemoney), I think I'll have to e-mail you the data so you can see how I went about doing it - the format is pretty rough right now because I didn't put it together for public use and got sloppy, so hopefully you can work through it. I'm pretty sure I still have your e-mail address. In short, I start with a set list of players and then use some vlookup to fill in the data I'm looking for. I had to create a unique identifier to do so - for that, I use the format last, firstTEAM - example: Brees, DrewNO. This keeps the Mike Williams/Steve Smith/Zach Miller from getting confused. To get all the various data sources straight, I had to do some fancy excel text manipulation: trim(), find(), left(), concatenate(), etc. For team names, I had to define a "rosetta" where a team format (NWE vs NE) is replaced w/ vlookup(). To get Chris Wells vs beanie wells straight - well, that was done manually. I haven't found a good way to format that correctly. One other trick is to break it all up into multiple files. One master file with all of the compiled projections gets to be too big and really slows down my computer.

Lastly, I converted projections to FF points before compiling, just to cut down on the # of computations.

I know I can do this better next year. If you would like to collaborate or at least bounce around some ideas, I'm all ears, but I don't want to get into all of this again until Feb or so ( Kick-off is in about 10 min!)

 

[dickey moe]

CBS projections for certain studs look really low this week. Low numbers predicted for Ray Rice this week, for example. Granted he has a tough matchup, but they were predicting close to single digit point production from him (at least the last time I checked).

 

[pack66]

Cool deal. Wish I would have ran across your thread prior to my draft. Like most here, I'm kind of a stats geek and love to see how the #'s play out v. the actual results. Please continue to post updates. Would also like to see how you went about picking your team.

 

[supersecretid]

So the night before the draft, I downloaded the compiled data from fantasyrundown.com and discovered that his data was practically worthless - projections that were supposed to be updated weren't, there were typos all over, etc. pure garbage. I had to go through the effort of manually downloading all of the various projections, formatting them so I could read them in, compiling, and then finally tweaking as I see fit. Not a small exercise to do the night before a draft. Major PITA. Had I realized that fantasyrundown.com would be useless, I would have set my stuff up differently and it would have been soo much easier. Beyond that, fantsysharks.com's projections were garbage, as were NFL.com. So, the data set I used included Eisenberg (CBS.com), Richards (CBS.com), ESPN.com, FFToday.com, and unlucky (FBG member who makes his projections public).The purpose of this whole exercise was to cut down on the work to get good projections, not complicate things - scrambling was not what I wanted to do the night before the draft. But, that's neither here nor there...more venting than anything else. For "interest score" - that is, to determine which players I needed to look at closer, I siimply multiplied the mean with the standard deviation. I don't know how or why I came up with that, but it seemed to do a good job of weighing out the high variability / high-scoring players that I wanted to focus on. My final list (data current as of 8/28) was as follows:Leinart, MattBrees, DrewFavre, BrettDelhomme, JakeJohnson, ChrisPeterson, AdrianRice, RayWashington, LeonForsett, JustinPeterson, AdrianForte, MattMathews, RyanWilliams, MikeJackson, VincentBoldin, AnquanAustin, MilesWelker, WesMarshall, BrandonHolmes, SantonioMurphy, LouisMoss, SantanaCrabtree, MichaelGibson, BrandonDoucet, EarlyCooley, ChrisClark, DallasGates, AntonioDaniels, Owenagain, this list represents the highest scoring players with the least amount of agreement according to the projections I considered. If course, there were other players I considered too...those were basically guys where my ranking differed the most as compared to the most recently published ADP.For the rest of your questions (gheemoney), I think I'll have to e-mail you the data so you can see how I went about doing it - the format is pretty rough right now because I didn't put it together for public use and got sloppy, so hopefully you can work through it. I'm pretty sure I still have your e-mail address. In short, I start with a set list of players and then use some vlookup to fill in the data I'm looking for. I had to create a unique identifier to do so - for that, I use the format last, firstTEAM - example: Brees, DrewNO. This keeps the Mike Williams/Steve Smith/Zach Miller from getting confused. To get all the various data sources straight, I had to do some fancy excel text manipulation: trim(), find(), left(), concatenate(), etc. For team names, I had to define a "rosetta" where a team format (NWE vs NE) is replaced w/ vlookup(). To get Chris Wells vs beanie wells straight - well, that was done manually. I haven't found a good way to format that correctly. One other trick is to break it all up into multiple files. One master file with all of the compiled projections gets to be too big and really slows down my computer. Lastly, I converted projections to FF points before compiling, just to cut down on the # of computations. I know I can do this better next year. If you would like to collaborate or at least bounce around some ideas, I'm all ears, but I don't want to get into all of this again until Feb or so ( Kick-off is in about 10 min!)

You still have that spreadsheet? I'd love to get a copy, would save me some work.

 

[stuckinthemuck]

This is great work and clearly there are several people who want this type of hard data. I used a less statistically rigorous approach with the new fantasy pros website. They're doing a lot of the work for you. I used them extensively while drafting to figure out which player would incur more risk and which might be a more stable pick. Although they're not formally calculating a standard deviation you can easily eyeball the number to see which player has more or less consensus.

I used this method in two drafts, one a league with 14 managers and 6 payout slots and one with 12 managers and only 2 payout slots. In both leagues I was deliberating between Jamal Charles and Ryan Grant. I took Grant in the 14 team league because I was looking for a more solid, less risky pick to get into the payout slots, while I took Charles in the 12 team league because I need to incur more risk to get into the top two slots.

Their data were as follows:

Best rank Worst rank ADP Expert rank

Grant 8 18 12 12

Charles 9 23 13 14

It's easy to see that there was more variability in ranks even though I didn't formally crunch the numbers. Pair that with the common knowledge of the RB's respective situations on their teams and it's even easier to know who's got a safe situation and who's primed for a huge (or crappy) year.

Keep working on developing this system. It would be very useful and, dare I say it, something that diehards might even shell out a few bucks for if it's done right.

 

[moleculo]

This is great work and clearly there are several people who want this type of hard data. I used a less statistically rigorous approach with the new fantasy pros website. They're doing a lot of the work for you. I used them extensively while drafting to figure out which player would incur more risk and which might be a more stable pick. Although they're not formally calculating a standard deviation you can easily eyeball the number to see which player has more or less consensus.

I used this method in two drafts, one a league with 14 managers and 6 payout slots and one with 12 managers and only 2 payout slots. In both leagues I was deliberating between Jamal Charles and Ryan Grant. I took Grant in the 14 team league because I was looking for a more solid, less risky pick to get into the payout slots, while I took Charles in the 12 team league because I need to incur more risk to get into the top two slots.

Their data were as follows:

Best rank Worst rank ADP Expert rank

Grant 8 18 12 12

Charles 9 23 13 14

It's easy to see that there was more variability in ranks even though I didn't formally crunch the numbers. Pair that with the common knowledge of the RB's respective situations on their teams and it's even easier to know who's got a safe situation and who's primed for a huge (or crappy) year.

Keep working on developing this system. It would be very useful and, dare I say it, something that diehards might even shell out a few bucks for if it's done right.

It's a very cool site there - but I gotta say - I think a ranking list is almost worthless as compared to a set of projections. Ranking is fine to comapre two RB's, but how do you compare a RB to a WR or a QB? Moreso, if you have 5 RB's all projected to score between 15.5 and 16.4 fantasy points, I'd call that even - i.e. virtually no difference between them. A ranking methodology has them in order 1,2,3,4,and 5. If someone things that two guys are about the same, tell me. Don't tell me that one guys is ranked 8th and the other is 13th.I'm happy to see that people are interested. I'll try to spend some time cleaning up the data so y'all can see what I'm trying to do. If you are interested, drop me a note @ fbgmoleculo _at_ gmail_dot_com and I'll be happy to send out what I came up with. Maybe next year I'll be able to tease out some more interesting info.