Fantasy Football - Footballguys Forums
This is a sample guest message. Register a free account today to become a member! Once signed in, you'll be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!
Projection Dominator (1 Viewer)
- Thread starter tytyty
- Start date
What does it do?
Mike Herman
Footballguy
"With the Projections Dominator you will be able to weight each staff member's numbers plus actuals to generate a unique set of numbers (to import into the Draft Dominator). All players can also be manually changed.
Additionally, this program will allow you to print customized cheatsheets and Top 200 lists.
No longer will a person using the Draft Dominator be tipping his hand. Because the underlying numbers could be Dodds, Wood, part Dodds, Wood and Henry, etc.
Good times are a coming here at Footballguys.com"
- Mr. Dodds
Additionally, this program will allow you to print customized cheatsheets and Top 200 lists.
No longer will a person using the Draft Dominator be tipping his hand. Because the underlying numbers could be Dodds, Wood, part Dodds, Wood and Henry, etc.
Good times are a coming here at Footballguys.com"
- Mr. Dodds
Art_VanDalay
Footballguy
Maybe it is me but I don't think this is working correctly right now. When I put in weights for the years 2003-05 (15%, 25% and 60% respectively) it seems to be adding together the totals for receptions, yards, etc. from these years instead of averaging them. Anyone else try this out? The ones for Dodds, Smith, etc. seem to work correctly as they are being averaged. Am I missing something?
SammyJankis
Footballguy
I've always thought that FF projections are done incorrectly. That's not exclusive to FBGs. When I heard the term "weighted projections", my thought was that the problem was perhaps addressed, but it sounds like its something completely different that I had hoped. Let me explain.....
Projections are really a measure of expected value. In statistics, Expected Value = the sum of all probabilities each multiplied by their value. For example, if I told you that we were going to gamble on a flip of a coin. If I won, you'd pay me two dollars, but if I lost I'd only owe you a dollar. Then the expected value of each flip is (.50)(-$1) + (.50)($2) = -.50 + $1 = .50 In other words, each flip of the coin is expected to earn me fifty cents in the long run.
Projections are just a measure of expected value. The problem with the way that they're done is that they are absolute. Take Ron Dayne for example. FBGs has Dayne projected at RB25. Does anyone out there think this where Dayne likely ends up at #25, either Dayne fans or Dayne doubters? If Dayne were to stay healthy and stay the starter (which is entirely possible) I'd guess that he would finish significantly higher than #25, while if Dayne were to struggle and lose his job, get injured, and/or split time (all also entirely possible) there's a good chance he wouldn't be worth much of anything.
The bottom line is that projections should be presented as a simple expected value statement. With Dayne you might say there is a 25% chance he finishes top 5, a 50% chance he splits time and finishes middle of the pack and a 25% chance he is worthless. This information is more valueable than the absolute prediction without the EV formula.
So what the projections dominator should do is present these EV formulas, then allow the user to modify the percentages for each expected outcome based on their opinion. To continue with the Dayne example, a big Dayne supporter would greatly increase the "chance for top 5" potential which would greatly increase Dayne's projection. Someone that thought Dayne stands no chance to hold the job would also adjust the percentages accordingly dropping Dayne's projection.
Adjusting Dodd's vs. Wood's projections doesn't buy you much IMO.
Of course, for many players, both the volatility in their projections wouldn't fluctuate nearly as much as Dayne. (I chose him as an example becuase he's particuarilly volatile). However, I would maintain that the volatile players that you have a "gut feel" about are the players that help you win your league, so being able to tweak projections accordingly goes a long way toward adding value to the tool.
My two cents.
Projections are really a measure of expected value. In statistics, Expected Value = the sum of all probabilities each multiplied by their value. For example, if I told you that we were going to gamble on a flip of a coin. If I won, you'd pay me two dollars, but if I lost I'd only owe you a dollar. Then the expected value of each flip is (.50)(-$1) + (.50)($2) = -.50 + $1 = .50 In other words, each flip of the coin is expected to earn me fifty cents in the long run.
Projections are just a measure of expected value. The problem with the way that they're done is that they are absolute. Take Ron Dayne for example. FBGs has Dayne projected at RB25. Does anyone out there think this where Dayne likely ends up at #25, either Dayne fans or Dayne doubters? If Dayne were to stay healthy and stay the starter (which is entirely possible) I'd guess that he would finish significantly higher than #25, while if Dayne were to struggle and lose his job, get injured, and/or split time (all also entirely possible) there's a good chance he wouldn't be worth much of anything.
The bottom line is that projections should be presented as a simple expected value statement. With Dayne you might say there is a 25% chance he finishes top 5, a 50% chance he splits time and finishes middle of the pack and a 25% chance he is worthless. This information is more valueable than the absolute prediction without the EV formula.
So what the projections dominator should do is present these EV formulas, then allow the user to modify the percentages for each expected outcome based on their opinion. To continue with the Dayne example, a big Dayne supporter would greatly increase the "chance for top 5" potential which would greatly increase Dayne's projection. Someone that thought Dayne stands no chance to hold the job would also adjust the percentages accordingly dropping Dayne's projection.
Adjusting Dodd's vs. Wood's projections doesn't buy you much IMO.
Of course, for many players, both the volatility in their projections wouldn't fluctuate nearly as much as Dayne. (I chose him as an example becuase he's particuarilly volatile). However, I would maintain that the volatile players that you have a "gut feel" about are the players that help you win your league, so being able to tweak projections accordingly goes a long way toward adding value to the tool.
My two cents.
Pretty useless? You should let every business with more then 20 employees know that then. Every day people read reports, projections, marketing data that someone else collected and analyzed. Favoring employees who you trust more (more experience, proven track record) is an essential skill in business.Adjusting one guy's projections vs another's is pretty useless when I'm hoping you odn't use ANYONE'S projections but your own to do your draft.
If you look at any medium to large business, the CEO doesn't do his own projections, for anything. Learn who you can trust, adjust how you weigh them, and then make your call.
FBG is simply preparing reports for us. FBG does the grunt work, and as the CEO, we decide who we trust, and how much. None of this has anything to do with being "useless". It's pretty much how the world runs. And while your FF team might be a small operation, if you can’t find value in FREE expert projections, you’re the one missing out.
This should also be how injury projections work (and I think it is often done that way) in that all players should be projected with a 16 game schedule in mind and then have a % of games expected to start calculated in.So if I think someone would likely rush for 1500 yards and 18 TD's if they start all 16 games, but I think they'll miss three games to injury, then I think they'll make 81% of those totals. Thats 1296 yards and 14.58 TD's. What? Half a TD??? Yup, that's the way the projection should come out... and that should give this guy the edge over someone who will only score 14 TD's.
Sigfawn, I agree with you in the "usefulness" of the projections. But there is a good point here that I never really looked at like this. There probably is NO WAY Dayne will be a RB25 based on the above. Actually it is a great point now I think of it. Again Dayne is an extreme example, but that doesn't help during draft time. Understood I do my projection tweaking, but that is why I pay for this site, to get a big head start so I ONLY have to tweak.Again, I disagree with "useless" but I think this would be a huge added feature to be able to have an area to customize these players on some of the great tools we have here.
SammyJankis
Footballguy
My post wasn't meant as a slam at all. Just trying to add some constructive criticism.
Additionally, if you express projections as a measure of expected value, you can calculate a measure of volatility for each player's projection as well.
For example, say that in the middle rounds you are considering two players. One is a deep sleeper. You give him only a 1 in 4 chance of winning a starting job. However, you project that if we wins that job, he'll score 200 points, but if he doesn't win it, he'll be completely worthless and score zero.
Another player that you're considering is a lock to start all year, but is a guy that has been steady but unspectatuclar, you don't see his value fluctuating much at all, and give him a 100% chance to score 50 points.
In the end, you've got two guys that you'd project to score 50 points. Clearly though, they're not the same. The first player is a "shoot the moon" hit a home-run player, while the other is a sure thing, but the return is much less. In the long-run, meaning if you entered many many fantasy drafts, each selecting these two players, the return on the investment expects to be about the same.
Depending on the situation though, most owners will have a preference one way or the other. Sometimes you want to give up value to after the sure thing, other times you're willing to take a guy that is a "lottery ticket" figuring that if he doesn't pay off, you'll just cut him.
Again, I've picked a rather extreme example to emphasize my point. Most players are rarely an "all or nothing" proposition. Regardless, the amount of volatility associated with a projection is useful to a drafter when comparing guys with similar projections: safer play or higher upside? The projections as they are currently done don't provide this insight.
Again, just one for the suggestion box, certainlly not a slam.
Additionally, if you express projections as a measure of expected value, you can calculate a measure of volatility for each player's projection as well.
For example, say that in the middle rounds you are considering two players. One is a deep sleeper. You give him only a 1 in 4 chance of winning a starting job. However, you project that if we wins that job, he'll score 200 points, but if he doesn't win it, he'll be completely worthless and score zero.
Another player that you're considering is a lock to start all year, but is a guy that has been steady but unspectatuclar, you don't see his value fluctuating much at all, and give him a 100% chance to score 50 points.
In the end, you've got two guys that you'd project to score 50 points. Clearly though, they're not the same. The first player is a "shoot the moon" hit a home-run player, while the other is a sure thing, but the return is much less. In the long-run, meaning if you entered many many fantasy drafts, each selecting these two players, the return on the investment expects to be about the same.
Depending on the situation though, most owners will have a preference one way or the other. Sometimes you want to give up value to after the sure thing, other times you're willing to take a guy that is a "lottery ticket" figuring that if he doesn't pay off, you'll just cut him.
Again, I've picked a rather extreme example to emphasize my point. Most players are rarely an "all or nothing" proposition. Regardless, the amount of volatility associated with a projection is useful to a drafter when comparing guys with similar projections: safer play or higher upside? The projections as they are currently done don't provide this insight.
Again, just one for the suggestion box, certainlly not a slam.
I agree. My main point was, this is far from useless. It's like saying, a week before the NFL draft, 5 other teams want to share with you their scouting/grading reports. You response with "nah, that's useless, I do my own scouting".
So there's value to it, and I'm sure it can be improved. But someone claiming it was useless was pretty funny.
Pots
Footballguy
Sorry, but no. It is useless because these arne't NFL teams sharing their info that you don't know about. It's just a bunch of guys who put a lot of work simlar to what you do every year in your own rankings. If you are hardcore in this game, 99% of the time, they don't know anyhting more than you already do. Sure if you have a favorite projecitonist they might come in handy by weighting their's higher than others, but people who win go by their own projections, not someone else's. Weighting a bunch of other player's projections is useless if you can't weight your own against them. I'm not knokcing anyone at FBG and their projections, just people who decide to use other projections as the only weights in rankings. Now if they incorporated your own inputed projections vs. the average FBG projectionist rankings, then you might have something. Or even more prefably, your projections with ADP.
My two cents:
To have EV formulas in DD wouldn't you need EV formulas for each stat, ie carries, ypc, receptions, yds, etc. because you want your rankins to be adjustable to different scoring leagues?
You can"t use unique "player value"(fantasy points) in the formula for more that one league because that # depends on league scoring.
To have EV formulas in DD wouldn't you need EV formulas for each stat, ie carries, ypc, receptions, yds, etc. because you want your rankins to be adjustable to different scoring leagues?
You can"t use unique "player value"(fantasy points) in the formula for more that one league because that # depends on league scoring.
SammyJankis
Footballguy
There are a lot of ways you could do it. Most scream for some kind of program or spreadsheet to make things easiser. (Like the dominator. Which is exactly why when I saw that you could "weight" the projections, my first thought was that this is where this was going....)The thing is that in situations like coin flips or die rolls, its easy to see all the outcomes and their probabilties. Not so easy here, to truly list every possible outcome wouldn't really be possible. What I would submit is that you categorize the set of possible outcomes, give default probabilities of each, then let the users adjust the probablities accordingly.
As far as grouping the situations, there are several ways this could be done:
- You could do three sets of projections for every guy. Maybe something like middle of the road, high-side, low-side, then asign perecentages to each.
- You could do as many sets as made sense for each player. For example, maybe you look at Shaun Alexander and make only one projection as you consider the outcomes that he slumps badly to be so unlikely as to not affect your projection much. (I would generally say to not make any kind of injury projection for most players. All players have some chance of injury and predicting it is impossible. I would only downgrade certain players that have higher than average injury risks such as Lee Suggs, chris Brown, etc.) But particuarilly for guys that there is debate about, you project each outcome. For example, folks are split on whether Kevin Jones will have a great leap in receptions, or whether he'll not be much of a pass catcher as he was last year.
Last edited by a moderator:
SammyJankis
Footballguy
Sorry. I did a bad job of answering your question. The process goes:
1) Set each set of projections. Not projecting total fantasy points, but stats for each. (Example: Dayne wins starting job = 1200 yards 14TDs, Dayne splits time = 800 yards 6 TDs, Dayne is worthless = 300 yards, 2 TDs)
2) Calculate the value of each by applying the input scoring system. (Example: In the input fantasy scoring system the resepctive values for those projections are: 200, 85, 30)
3) Apply probablities (possibly user adjusted) to each possible projection and sum is your expected value. (Example: I think Dayne is a long shot to start all year, but think he'll have some games when he's the man and others where Bell steps up and Dayne languishes, so I think the "middle of the road projection" is far and away most likely. I asign values of 10%, 80% and 10% to the three projections. This gives me a projection of 91 for him. This affirms my middle of the road projection, but accounts for the fact that he could explode or be worthless. When I read a camp report where Shanny says "Dayne is the man" then I can adjust my percentages and get an accurate representation of how it affects his value.)
The "user adjusted" part of the projections is the key thing here. If you really believe that Dayne wins the job, he gets a big bump on your sheet. If you think he is a bum with no shot of starting all year, he drops way below RB25.
Another thing this gives you is that you can compute a measure of volatility for each player. So that when I look at Dayne's projections, I can see that they are much more likely to fluctuate one way when compared to other backs in the same range.
Last edited by a moderator:
SammyJankis
Footballguy
One more thought: Another big advantage of this approach is that I've seen many on this board state that they like to create their own projections but they like to use FBGs as a starting point.
There is an "art vs. a science" to good projections. The art is figuring out what the numbers really look like if Dayne wins the job in Denver. The art, is using your gut and knowledge to determine just how likely that is. Designing the tool in this manner allows FBGs to take care of the "science" part leaving the "Art" (the fun part) in the hands of those using the tool.
There is an "art vs. a science" to good projections. The art is figuring out what the numbers really look like if Dayne wins the job in Denver. The art, is using your gut and knowledge to determine just how likely that is. Designing the tool in this manner allows FBGs to take care of the "science" part leaving the "Art" (the fun part) in the hands of those using the tool.
Similar threads
