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!
retrospective mock draft using 2009 data (2 Viewers)
- Thread starter moleculo
- Start date
No Way Jose
Footballguy
Mohammed Massaquooieeu, or however you spell it.i
moleculo
Footballguy
Final All-play standings (previous end of round all play in parenthesis)
moleculo (78) (72) (90) (91) (87) (83) (88) (78) (70) (75) 73
fubar (72) (79) (93) (92) (86) (94) (85) (89) (87) (89) 88
guppy D (109) (99) (90) (90) (85) (90) (84) (83) (78) (79) 76
gonzobill5 (83) (89) (101) (96) (86) (79) (83) (87) (88) (90) 91
No Way Jose (109) (101) (96) (97) (93) (90) (83) (82) (91) (96) 96
guppy E (111) (114) (87) (93) (95) (102) (98) (99) (93) (87) 86
Football Critic (104) (97) (92) (75) (74) (77) (77) (70) (76) (76) 73
Truman (75) (76) (66) (63) (81) (74) (79) (86) (84) (82) 80
guppy F (76) (79) (97) (97) (81) (100) (101) (98) (84) (89) 99
gheemoney (96) (97) (90) (96) (79) (72) (95) (108) (115) (115) 109
guppy C (64) (73) (76) (96) (95) (96) (89) (90) (100) (92) 96
guppy G (76) (79) (77) (70) (99) (98) (94) (86) (90) (84) 89
Gheemoney wins. Looks like my team came in last place. More analysis to follow.
moleculo (78) (72) (90) (91) (87) (83) (88) (78) (70) (75) 73
fubar (72) (79) (93) (92) (86) (94) (85) (89) (87) (89) 88
guppy D (109) (99) (90) (90) (85) (90) (84) (83) (78) (79) 76
gonzobill5 (83) (89) (101) (96) (86) (79) (83) (87) (88) (90) 91
No Way Jose (109) (101) (96) (97) (93) (90) (83) (82) (91) (96) 96
guppy E (111) (114) (87) (93) (95) (102) (98) (99) (93) (87) 86
Football Critic (104) (97) (92) (75) (74) (77) (77) (70) (76) (76) 73
Truman (75) (76) (66) (63) (81) (74) (79) (86) (84) (82) 80
guppy F (76) (79) (97) (97) (81) (100) (101) (98) (84) (89) 99
gheemoney (96) (97) (90) (96) (79) (72) (95) (108) (115) (115) 109
guppy C (64) (73) (76) (96) (95) (96) (89) (90) (100) (92) 96
guppy G (76) (79) (77) (70) (99) (98) (94) (86) (90) (84) 89
Gheemoney wins. Looks like my team came in last place. More analysis to follow.
moleculo
Footballguy
Guppy drafting strategy:
Moleculo (73 wins): I use a similar strategy to FBG, but I set my baseline a little differently. To establish the baseline, I think about the league-wide number of starters, multiply by 1.5, and then average the end of year scoring for that many players. For example - there are 12 QB's that will start. 12*1.5=18. So, I average the top 18 QB overall points, which ends up with a number generally somewhere between QB 9-QB12. I like this method to establish baseline because it is less sensitive to individual fluctuations. VBD score is therefore end of season score - baseline. The VBD allows one to compare across positions.
Guppy C (96 wins) : games over baseline. First of all ,the baseline points per game was established similar to above, but the baseline score is divided by 16 to come up with an average score per game. Next, I calculated each players average score and stdev. Using these numbers and a little statistics, I calculated the probability of each player scoring above the baseline, multiplied by 16 to arrive at the expected number of games that we could expect a player to outperform the baseline. I could have simply counted the number of games that the baseline was exceeded, but I found that the expected value and the actual value were pretty similar.
Guppy D (76 wins) : Yahoo simulation. To draft a player, I looked at overall points, and picked whomever was left that scored the most. However, one caveat: draft to fill all starters before any back-ups - i.e. TE must be selected before any flex, no back-up QB or TE until all flex are filled out, etc. Emphasis on filling out the line-up over putting together a better overall team.
Guppy E: (86 wins) Target variance: I used FBG VBD and added the stdev * 1.5 as the scoring criteria. (stdev was multiplied by 1.5 to emphasize the variance - using straight stdev didn't make this list significantly different than FBG VBD). The idea here is to target guys who are more up and down.
Guppy F (99 wins) : Target consistency: same method as Guppy E, but 1.5*stdev was subtracted from FBG VBD. This was done to try to target guys who are more consistent.
Guppy G (89 wins) : straight up FBG VBD rankings.
Moleculo (73 wins): I use a similar strategy to FBG, but I set my baseline a little differently. To establish the baseline, I think about the league-wide number of starters, multiply by 1.5, and then average the end of year scoring for that many players. For example - there are 12 QB's that will start. 12*1.5=18. So, I average the top 18 QB overall points, which ends up with a number generally somewhere between QB 9-QB12. I like this method to establish baseline because it is less sensitive to individual fluctuations. VBD score is therefore end of season score - baseline. The VBD allows one to compare across positions.
Guppy C (96 wins) : games over baseline. First of all ,the baseline points per game was established similar to above, but the baseline score is divided by 16 to come up with an average score per game. Next, I calculated each players average score and stdev. Using these numbers and a little statistics, I calculated the probability of each player scoring above the baseline, multiplied by 16 to arrive at the expected number of games that we could expect a player to outperform the baseline. I could have simply counted the number of games that the baseline was exceeded, but I found that the expected value and the actual value were pretty similar.
Guppy D (76 wins) : Yahoo simulation. To draft a player, I looked at overall points, and picked whomever was left that scored the most. However, one caveat: draft to fill all starters before any back-ups - i.e. TE must be selected before any flex, no back-up QB or TE until all flex are filled out, etc. Emphasis on filling out the line-up over putting together a better overall team.
Guppy E: (86 wins) Target variance: I used FBG VBD and added the stdev * 1.5 as the scoring criteria. (stdev was multiplied by 1.5 to emphasize the variance - using straight stdev didn't make this list significantly different than FBG VBD). The idea here is to target guys who are more up and down.
Guppy F (99 wins) : Target consistency: same method as Guppy E, but 1.5*stdev was subtracted from FBG VBD. This was done to try to target guys who are more consistent.
Guppy G (89 wins) : straight up FBG VBD rankings.
I'll add more later, but here's my basic approach
1. VBD - use VBD to identify where the most value existed. A bit of "dynamic VBD" was done in my head to figure out what type of value would be left at each position at my next pick. This led me to wait on QB.
2. Average Points Per Game - Once I narrowed down the pockets of VBD value, I would look to draft the player that had the highest FP/G. This was usually because the player missed some games. Because we had perfect information, I knew I could later draft someone to fill in specifically for those games missed. This helped me identify Owen Daniels as a late TE to draft, and then pair him with Fred Davis. I think, combined, they are the same as TE4 or TE5. Not bad for late round picks.
3. Most Startable Games. Once I had my starters, I looked for back-ups that would crack my lineup most frequently. I didn't have an automated system for this. Through trial and error, I tried to find which back-ups would most often contribute to my active lineup. I think this led me to Dwayne Bowe despite a relatively low VBD.
4. Production in Weak Team Games. Towards the end of the draft, I identified the weeks when my team was underperforming, and I focused on drafting players that scored most of their points during that week. I assumed that scoring more in those weeks would be most likely to increase my all-play points.
The next for me is to (a) figure out which of these methods can be applied to a non-retrospective draft without change and (b) which of these methods can be applied through some type of simulation (e.g., you need a statistical approximation instead of counting each time a player cracks your starting lineup). Discard any methods that don't work in case (a) or (b).
1. VBD - use VBD to identify where the most value existed. A bit of "dynamic VBD" was done in my head to figure out what type of value would be left at each position at my next pick. This led me to wait on QB.
2. Average Points Per Game - Once I narrowed down the pockets of VBD value, I would look to draft the player that had the highest FP/G. This was usually because the player missed some games. Because we had perfect information, I knew I could later draft someone to fill in specifically for those games missed. This helped me identify Owen Daniels as a late TE to draft, and then pair him with Fred Davis. I think, combined, they are the same as TE4 or TE5. Not bad for late round picks.
3. Most Startable Games. Once I had my starters, I looked for back-ups that would crack my lineup most frequently. I didn't have an automated system for this. Through trial and error, I tried to find which back-ups would most often contribute to my active lineup. I think this led me to Dwayne Bowe despite a relatively low VBD.
4. Production in Weak Team Games. Towards the end of the draft, I identified the weeks when my team was underperforming, and I focused on drafting players that scored most of their points during that week. I assumed that scoring more in those weeks would be most likely to increase my all-play points.
The next for me is to (a) figure out which of these methods can be applied to a non-retrospective draft without change and (b) which of these methods can be applied through some type of simulation (e.g., you need a statistical approximation instead of counting each time a player cracks your starting lineup). Discard any methods that don't work in case (a) or (b).
gonzobill5
Footballguy
Can we see how our teams did with the HTH schedule you posted? If its too much work, don't bother. I'm just curious.moleculo said:
On your calculation of the probability of each player scoring above the baseline: I can see how if you know the mean, the standard deviation, and the baseline, that you can calculate the probability by measuring the "area under the curve" for that segment. But, you have to know what the distribution is.Key question: did you assume a normal distribution? If so, why? Correlation? Best fit? Research?
I don't know the answer, but have been trying to find the answer and here's the information I've gathered:
1. Doug Drinen's research suggests that TDs follow a poisson distribution. This makes sense since TDs are effectively integers (you score 0, 1, 2, ... TDs).
2. I can't find the link (it's in the forum somewhere), but I think Drinen said that the distribution of actual performance around a player's projected performance (for a week) is log normal distribution. This is because you can't get negative points (or at least it's rare). He cited that he has data helping him determine the parameters for the log normal distribution (i.e., wider spread for higher projections).
3. I don't know the distribution for yards per game, but some smart people have suggested to me (without any looking at data) that it follows a normal distribution because yards are not fixed like TDs (you can get 52, 68, 97 yards instead of 0, 1, 2 TDs).
So, if fantasy points are equal to X*TDs+Y*Yards, then is the distribution normal, lognormal, poisson, or other? This is relevant to determining if the method used for Guppy C is an accurate way to determine the probability of scoring above the baseline.
My head hurts with all this math. Hopefully someone can help with this. I think creating a reasonable solution is a big step towards coming up with a an improved player valuation system.
moleculo
Footballguy
I've been told that unless you have reason to believe otherwise, always assume a normal distribution for a large population (CLT). I would agree that TD's (and probably catches) are not normally distributed in that one can not have negative TD's and they are discreet. Yardage is discreet but in terms of our resolution, I would say it's appropriate to assume continuous. This is only good for end of season stuff though, so what I did probably isn't rigorously accurate.What I did here is break it down to fantasy points instead of yardage/TD's. A more rigorous analysis could probably handle this more elegantly, but I'm lazy, this math is simple, and looking at the actual data (i.e. predicted # of games above baseline calculated with mean and stdev vs actual games above baseline), it followed pretty closely, so I ran with it.
Again, this is all academic unless we can find a way to predict stdev, and I have no idea if that's possible or not.
Excellent, this is helpful. I need to think about this more, but one quick thought
oisson distribution is essentially normal distribution with standard deviation equal to the square root of the mean (if you're a stats whiz, please don't kill me for my simplistic explanation). We know that TDs are poisson distributed. So I'm tempted to assume that, for each player, we have a normal distribution with mean of x fp/g and a standard deviation of sqrt(x) fp/g.gonzobill5
Footballguy
A large population isn't any more likely to be normal than a small one. The CLT is about the sampling distribution of sample means, and that comes into effect if the sample size is large enough and the original distribution has finite variance.
I've heard that fantasy points for each position follow a log normal model pretty closely. Theoretically this doesn't make sense to me, since the log normal is usually applied in practice when dealing with the product of a bunch of independent random variables, not a linear combination like we have with FPs.
I've always wondered if the log normal distribution still applies if we are limiting our population to, say, the top 50 RBs or top 25 QBs (you know, the players that will actually get drafted).
I've heard that fantasy points for each position follow a log normal model pretty closely. Theoretically this doesn't make sense to me, since the log normal is usually applied in practice when dealing with the product of a bunch of independent random variables, not a linear combination like we have with FPs.
I've always wondered if the log normal distribution still applies if we are limiting our population to, say, the top 50 RBs or top 25 QBs (you know, the players that will actually get drafted).
Just to be sure that we're talking about the same thing: if a player has an averages X fantasy points per game, what is the distribution of his actual game-by-game production (normal, lognormal, poisson, ???)? Which Shark is has a Phd in Statistics and can answer this question for us?
gonzobill5
Footballguy
Oh, I thought you were referring to the shape of the distribution of total points for the population of all fantasy players.
In that case...I think it reasonable to suspect that a player's yards are approximately Normally distributed. However, as you suggested, touchdowns most certainly aren't. And fantasy points, since they are a linear combination of TDs and yards (and perhaps receptions) which are not independent random variables, are definitely not Normally distributed. The distribution would likely be skewed and perhaps even bimodal.
I currently have the data stored in a program that can test for Normality, including measuring the skew and kurtosis. But I would have to do one player at a time, and I've input the data in such a way that I can do these calculation easily for weeks but not for players (essentially I set the weeks as the variables and the players as the cases when I compiled).
In that case...I think it reasonable to suspect that a player's yards are approximately Normally distributed. However, as you suggested, touchdowns most certainly aren't. And fantasy points, since they are a linear combination of TDs and yards (and perhaps receptions) which are not independent random variables, are definitely not Normally distributed. The distribution would likely be skewed and perhaps even bimodal.
I currently have the data stored in a program that can test for Normality, including measuring the skew and kurtosis. But I would have to do one player at a time, and I've input the data in such a way that I can do these calculation easily for weeks but not for players (essentially I set the weeks as the variables and the players as the cases when I compiled).
gonzobill5
Footballguy
My draft strategy was to go after the players with high variance early (and strong playoff performances), and then find players with high correlation once I had a few players established. I used a giant matrix to measure the covariance, though the relationships weren't nearly strong as I expected. Essentially, I had hoped to create 8 or so dominant weeks during the regular season and I could care less how my team did the other weeks. Interestingly, I abandoned the strategy in round 6 to pick MSW, who was quite the opposite. I was worried I wouldn't hit 8 HTH wins (enough to make the playoffs by my estimate). He and Austin seemed to pair well, and had a negative correlation. This pick was the only one where I dropped down in All Play wins I think.
gonzobill5
Footballguy
Last year I tried to use a qualitative list to predict how the QBs would rank in terms of variance, and it turned out pretty awful. It was essentially a checklist
oes the QB excessively rely on a single targetDoes the QB have a suspect OLineDoes the QB play in a domeDoes the QB have a reliable safety valve (I defined this as a pass catching RB or reliable TE)and so on. I never have gone back to do an item analysis to see if any of the questions were helpful in prediction. I liked the idea a lot, but the results just didn't measure up and I gave up on it.
crush304
Footballguy
i wrote a fantasy football simulator a couple of years ago, rather then fit a projection curve (never seen any evidence that it fits a known distribution, and i'm not any kind of expert on that anyway) i'd just get a pool of possible points for each player by comparing a player to past player with similar adp (either by pos rank or by actual adp)
for example using pos rank for RB 10 (the 10th drafted rb) i'd get a pool of possible scores of all rb past 8 years who had a pos rank between 8-12
which would give me a pool of 40 possible scores. I'd do that for every player. Any time you simulate a season you just randomly pick one of the 40 seasons for each player and then use that season's weekly scores to run the simulation using best ball to pick each weekly lineup.
I'd draft teams and then run like 1000 season simulations to get win totals and also starts per player (you could see something like 60% starts for a qb you took early, and 40% starts for a qb you took later).
The orig version i wrote a couple of years ago worked in theory but took forever, it was something like 6 minutes for 1000 seasons, I was working on a full rewrite last year was much faster but between trying to round up stats, adp data, merging the 2 data sources and writing a program i never really got it usable.
The whole project did really changed the way I look at fantasy football. It was awesome in that it fully included injury history, late round guys had a shot to be top 5, early players had a change to completely bust out etc. It really got me out of the mindset that you draft couple of guys early to start, then draft other fill-ins later for bye weeks/etc, to the mindset that you draft 5 rb's and out of that pool you need to produce 2 starters each week.
for example using pos rank for RB 10 (the 10th drafted rb) i'd get a pool of possible scores of all rb past 8 years who had a pos rank between 8-12
which would give me a pool of 40 possible scores. I'd do that for every player. Any time you simulate a season you just randomly pick one of the 40 seasons for each player and then use that season's weekly scores to run the simulation using best ball to pick each weekly lineup.
I'd draft teams and then run like 1000 season simulations to get win totals and also starts per player (you could see something like 60% starts for a qb you took early, and 40% starts for a qb you took later).
The orig version i wrote a couple of years ago worked in theory but took forever, it was something like 6 minutes for 1000 seasons, I was working on a full rewrite last year was much faster but between trying to round up stats, adp data, merging the 2 data sources and writing a program i never really got it usable.
The whole project did really changed the way I look at fantasy football. It was awesome in that it fully included injury history, late round guys had a shot to be top 5, early players had a change to completely bust out etc. It really got me out of the mindset that you draft couple of guys early to start, then draft other fill-ins later for bye weeks/etc, to the mindset that you draft 5 rb's and out of that pool you need to produce 2 starters each week.
Very cool. I think running simulations is the best way to do projections and rankings. Of course, your inputs and assumptions have to be good. I know that Football Outsiders and Accuscore and one other service create their projections using a simulation of games (I think 10,000). I wish those sites would evolve more quickly to what baseball has: projections with probability of hitting those projections.Anyway, is there any way to use your simulator to simulate what we're trying to solve here? I wonder if your data could be used by Gonzobill to determine normality.Checking for normality seems like the next most important step. Do you need different data or do you need to input the data differently in order for your program to work?
moleculo
Footballguy
I've gone ahead and finished out the H2H season - gheemoney won the overall title. Final standings:
Division #1 wins losses PF PAfubar 6 7 1624.45 1479.8guppy D 6 7 1543 1653.8moleculo 4 9 1594.05 1621.7 Division #2 wins losses PF PANo Way Jose 10 3 1586.35 1389.35guppy E 6 7 1593 1688.1gonzobill5 5 8 1501.25 1449.8 Division #3 wins losses PF PAguppy F 7 6 1567.7 1758.4Truman 5 8 1482.6 1437.25Football Critic 4 9 1482.95 1733.15 Division #4 wins losses PF PAgheemoney 9 4 1680.65 1710.8guppy G 9 4 1619.1 1490.95guppy C 7 6 1620.75 1482.75division Winners: No Way Jose
gheemoney
guppy F
fubar
Wild Card:
guppy G
guppy C
playoffs: round 1 (week 14):
No Way Jose bye
gheemoney bye
guppy F 131
guppy C 141
fubar 117
guppy G 94
Round 2 (week 15):
No Way Jose 145
guppy C 109
gheemoney 137
fubar 111
Championship (week 16):
No Way Jose 102
gheemoney 115
Division #1 wins losses PF PAfubar 6 7 1624.45 1479.8guppy D 6 7 1543 1653.8moleculo 4 9 1594.05 1621.7 Division #2 wins losses PF PANo Way Jose 10 3 1586.35 1389.35guppy E 6 7 1593 1688.1gonzobill5 5 8 1501.25 1449.8 Division #3 wins losses PF PAguppy F 7 6 1567.7 1758.4Truman 5 8 1482.6 1437.25Football Critic 4 9 1482.95 1733.15 Division #4 wins losses PF PAgheemoney 9 4 1680.65 1710.8guppy G 9 4 1619.1 1490.95guppy C 7 6 1620.75 1482.75division Winners: No Way Jose
gheemoney
guppy F
fubar
Wild Card:
guppy G
guppy C
playoffs: round 1 (week 14):
No Way Jose bye
gheemoney bye
guppy F 131
guppy C 141
fubar 117
guppy G 94
Round 2 (week 15):
No Way Jose 145
guppy C 109
gheemoney 137
fubar 111
Championship (week 16):
No Way Jose 102
gheemoney 115
moleculo
Footballguy
I know the below is hard to read - if anyone has any tips on making this more readable, I'm all ears. I'm working on uploading this to my google docs account, but the innernet must be all jammed up or something. if anyone wants me to e-mail it to them, send me a PM w/ your e-mail adress and I'll be happy to forward it along.
Anyways, below is the entire draft. If you can get this into excel ,the columns are as follows:draft pick by round, overall pick, player name, position, # of games this player was a starter, All Play Wins Over Replacement (APWOR) , number of H2H wins (vs replacement), total points scored by the player as a starter, percentage of team all-play wins attributable to the player, and % of team points attributable to the player.
I haven't had a chance to sit down and look at things very closely yet; I expect to do that tomorrow. The first thing that's apparent though - APWOR isn't really a very good descriptor of a players value - it's very good at describing the impact a player had on his team, but it's tough to compare across teams. For example, if someone has two very good, relatively equivalent QB's, their APWOR wouldn't be very high because the team could continue along and not miss a beat...in that manner APWOR is really a comparison of player X vs player Y on the bench.
Additionally, measuring value over replacement in terms of team wins is kind of goofy in that if a team is overall very poor, he will be limited in how big his value over replacement is...that's why I included the metric of % of team wins.
There's definitely some interesting things here, it will be interesting to tear this down further.
Anyways, below is the entire draft. If you can get this into excel ,the columns are as follows:draft pick by round, overall pick, player name, position, # of games this player was a starter, All Play Wins Over Replacement (APWOR) , number of H2H wins (vs replacement), total points scored by the player as a starter, percentage of team all-play wins attributable to the player, and % of team points attributable to the player.
I haven't had a chance to sit down and look at things very closely yet; I expect to do that tomorrow. The first thing that's apparent though - APWOR isn't really a very good descriptor of a players value - it's very good at describing the impact a player had on his team, but it's tough to compare across teams. For example, if someone has two very good, relatively equivalent QB's, their APWOR wouldn't be very high because the team could continue along and not miss a beat...in that manner APWOR is really a comparison of player X vs player Y on the bench.
Additionally, measuring value over replacement in terms of team wins is kind of goofy in that if a team is overall very poor, he will be limited in how big his value over replacement is...that's why I included the metric of % of team wins.
There's definitely some interesting things here, it will be interesting to tear this down further.
Code:
draft pick team player pos_rank # games started all play wins H2H wins points % of all play wins % of team points 1.01 1 gheemoney Chris Johnson RB1 13 41 4 258.3 38.0% 12% 1.02 2 guppy D Maurice Jones-Drew RB2 15 27 2 195.6 36.0% 10% 1.03 3 No Way Jose Aaron Rodgers QB1 12 48 4 242.2 50.0% 13% 1.04 4 guppy G Adrian Peterson - min RB3 13 31 3 177.3 35.2% 9% 1.05 5 fubar Ray Rice RB4 15 27 1 208.9 30.7% 11% 1.06 6 Football Critic Drew Brees QB2 11 26 1 184.6 35.6% 10% 1.07 7 Truman Andre Johnson WR1 14 30 4 206.4 37.5% 11% 1.08 8 guppy E Wes Welker WR2 13 25 2 170.1 29.4% 9% 1.09 9 guppy C Dallas Clark TE1 14 29 2 225.5 30.5% 11% 1.10 10 gonzobill5 Frank Gore RB5 10 22 1 169.2 24.4% 9% 1.11 11 moleculo Reggie Wayne WR3 13 16 0 174.4 20.0% 9% 1.12 12 guppy F Larry Fitzgerald WR4 13 31 2 174.1 31.6% 9% 2.01 13 guppy F Brandon Marshall WR5 15 34 1 171 34.7% 9% 2.02 14 moleculo Randy Moss WR6 11 23 0 184.7 28.8% 9% 2.03 15 gonzobill5 Miles Austin WR7 10 22 2 197.9 24.4% 10% 2.04 16 guppy C Steve Smith NYG WR8 14 28 2 159.5 29.5% 8% 2.05 17 guppy E Peyton Manning QB3 12 24 3 177.5 28.2% 9% 2.06 18 Truman Roddy White WR9 13 27 2 156.9 33.8% 8% 2.07 19 Football Critic DeSean Jackson WR10 12 27 2 155.1 37.0% 8% 2.08 20 fubar Antonio Gates TE2 14 27 1 175.3 30.7% 9% 2.09 21 guppy G Vernon Davis TE3 13 24 4 153.2 27.3% 8% 2.1 22 No Way Jose Vincent Jackson WR11 11 27 2 157.1 28.1% 8% 2.11 23 guppy D Matt Schaub QB4 11 22 2 141 29.3% 7% 2.12 24 gheemoney Joseph Addai RB6 13 21 2 139.4 19.4% 7% 3.01 25 gheemoney Hines Ward WR12 11 23 1 133.8 21.3% 6% 3.02 26 guppy D Sidney Rice WR13 11 21 2 125.8 28.0% 7% 3.03 27 No Way Jose Ricky Williams RB7 12 25 2 147 26.0% 8% 3.04 28 guppy G Steven Jackson RB8 13 27 3 110.7 30.7% 6% 3.05 29 fubar Chad Ochocinco WR14 13 18 1 147.5 20.5% 7% 3.06 30 Football Critic Jamaal Charles RB9 9 14 1 93.5 19.2% 5% 3.07 31 Truman Jonathan Stewart RB10 10 13 1 100.8 16.3% 5% 3.08 32 guppy E Santonio Holmes WR15 12 16 2 106.8 18.8% 6% 3.09 33 guppy C Marques Colston WR16 11 19 2 128.6 20.0% 6% 3.1 34 gonzobill5 Ryan Grant RB11 13 22 1 134.6 24.4% 7% 3.11 35 moleculo Brent Celek TE4 14 18 0 155.2 22.5% 8% 3.12 36 guppy F Tony Gonzalez TE5 12 21 3 106.9 21.4% 5% 4.01 37 guppy F Philip Rivers QB5 11 35 3 170 35.7% 9% 4.02 38 moleculo Tom Brady QB6 12 26 0 252.9 32.5% 13% 4.03 39 gonzobill5 Tony Romo QB7 9 22 2 171.85 24.4% 9% 4.04 40 guppy C Thomas Jones RB12 13 26 2 134.3 27.4% 7% 4.05 41 guppy E DeAngelo Williams RB13 11 10 1 107.9 11.8% 6% 4.06 42 Truman Rashard Mendenhall RB14 8 12 2 100.4 15.0% 5% 4.07 43 Football Critic Anquan Boldin WR17 13 20 3 120.3 27.4% 7% 4.08 44 fubar Brett Favre QB8 11 20 0 159.9 22.7% 8% 4.09 45 guppy G Derrick Mason WR18 12 18 2 116.2 20.5% 6% 4.1 46 No Way Jose Jason Witten TE6 14 23 2 132.9 24.0% 7% 4.11 47 guppy D Kellen Winslow TE7 10 16 0 101.8 21.3% 5% 4.12 48 gheemoney Greg Jennings WR19 13 17 2 102.4 15.7% 5% 5.01 49 gheemoney Steve Smith car WR20 10 19 1 103.8 17.6% 5% 5.02 50 guppy D Donald Driver WR21 11 14 1 99 18.7% 5% 5.03 51 No Way Jose Tim Hightower RB15 13 20 1 117.1 20.8% 6% 5.04 52 guppy G Matt Forte RB16 10 12 2 70.6 13.6% 4% 5.05 53 fubar Robert Meachem WR22 10 9 0 91.1 10.2% 5% 5.06 54 Football Critic Ronnie Brown RB17 8 10 1 82.3 13.7% 4% 5.07 55 Truman Cedric Benson RB18 10 12 2 104.8 15.0% 6% 5.08 56 guppy E Heath Miller TE8 11 18 0 95.3 21.2% 5% 5.09 57 guppy C Eli Manning QB9 8 20 1 127.1 21.1% 6% 5.1 58 gonzobill5 Mike Sims-Walker WR23 9 14 0 122.7 15.6% 6% 5.11 59 moleculo Pierre Thomas RB19 12 10 0 101.5 12.5% 5% 5.12 60 guppy F Fred Jackson RB20 8 18 2 63.75 18.4% 3% 6.01 61 guppy F kevin smith RB21 9 14 2 67.6 14.3% 3% 6.02 62 moleculo reggie bush RB22 8 3 0 72.1 3.8% 4% 6.03 63 gonzobill5 Jerome Harrison RB23 8 18 1 99 20.0% 5% 6.04 64 guppy C Calvin Johnson WR24 12 11 0 106.3 11.6% 5% 6.05 65 guppy E T.J. Houshmandzadeh WR25 9 18 1 83.8 21.2% 4% 6.06 66 Truman Ben Roethlisberger QB10 11 19 2 172.3 23.8% 9% 6.07 67 Football Critic Visanthe Shiancoe TE9 11 15 2 90.6 20.5% 5% 6.08 68 fubar LaDainian Tomlinson RB24 12 14 0 82.5 15.9% 4% 6.09 69 guppy G Kurt Warner QB11 10 23 2 140.85 26.1% 7% 6.1 70 No Way Jose Percy Harvin WR26 9 24 3 92.9 25.0% 5% 6.11 71 guppy D Marion Barber RB25 10 8 0 64.3 10.7% 3% 6.12 72 gheemoney Owen Daniels TE10 8 18 2 103 16.7% 5% 7.01 73 gheemoney Steve Slaton RB26 8 14 1 89.7 13.0% 4% 7.02 74 guppy D Zach Miller oak TE11 10 13 1 66.9 17.3% 4% 7.03 75 No Way Jose Jermichael Finley TE12 7 15 1 79.1 15.6% 4% 7.04 76 guppy G Mario Manningham WR27 8 13 2 79.2 14.8% 4% 7.05 77 fubar Austin Collie WR28 10 19 2 90.9 21.6% 5% 7.06 78 Football Critic Devery Henderson WR29 9 8 1 63.4 11.0% 3% 7.07 79 Truman Beanie Wells RB27 6 10 1 67.8 12.5% 4% 7.08 80 guppy E Darren Sproles RB28 8 6 1 87.3 7.1% 5% 7.09 81 guppy C Michael Turner RB29 7 12 1 86.9 12.6% 4% 7.1 82 gonzobill5 Jay Cutler QB12 7 15 0 102.65 16.7% 5% 7.11 83 moleculo Knowshon Moreno RB30 11 4 0 54.9 5.0% 3% 7.12 84 guppy F Santana Moss WR30 8 15 1 68.8 15.3% 4% 8.01 85 guppy F cadillac williams RB31 9 13 1 56.1 13.3% 3% 8.02 86 moleculo Terrell Owens WR31 9 7 0 66.5 8.8% 3% 8.03 87 gonzobill5 Todd Heap TE13 12 16 1 106.1 17.8% 5% 8.04 88 guppy C Donovan McNabb QB13 8 14 2 93.5 14.7% 5% 8.05 89 guppy E Hakeem Nicks WR32 9 10 1 66.4 11.8% 3% 8.06 90 Truman Kevin Boss TE14 12 16 3 72.1 20.0% 4% 8.07 91 Football Critic Justin Forsett RB32 9 12 2 71.1 16.4% 4% 8.08 92 fubar Carson Palmer QB14 5 10 1 70.3 11.4% 4% 8.09 93 guppy G Nate Burleson WR33 8 11 2 81.1 12.5% 4% 8.1 94 No Way Jose Jeremy Maclin WR34 9 11 1 71.9 11.5% 4% 8.11 95 guppy D Joe Flacco QB15 5 8 1 61.6 10.7% 3% 8.12 96 gheemoney Fred Davis TE15 9 16 2 103.7 14.8% 5% 9.01 97 gheemoney Kyle Orton QB16 10 33 4 174.55 30.6% 8% 9.02 98 guppy D Brandon Jacobs RB33 8 7 1 58.1 9.3% 3% 9.03 99 No Way Jose Matt Hasselbeck QB17 4 12 2 62.15 12.5% 3% 9.04 100 guppy G LeSean McCoy RB34 7 3 0 23.8 3.4% 1% 9.05 101 fubar Steve Breaston WR35 8 10 0 75.4 11.4% 4% 9.06 102 Football Critic Ahmad Bradshaw RB35 9 8 0 77.4 11.0% 4% 9.07 103 Truman Pierre Garcon WR36 8 12 3 59.8 15.0% 3% 9.08 104 guppy E Laurence Maroney RB36 9 11 0 47 12.9% 2% 9.09 105 guppy C Julius Jones RB37 7 12 1 61.1 12.6% 3% 9.1 106 gonzobill5 jerricho cotchery WR37 10 18 1 80.9 20.0% 4% 9.11 107 moleculo Braylon Edwards WR38 6 2 0 46.6 2.5% 2% 9.12 108 guppy F Mike Wallace WR39 7 12 0 50.5 12.2% 3% 10.01 109 guppy F John Carlson TE16 9 13 1 46.3 13.3% 2% 10.02 110 moleculo Chester Taylor RB38 6 7 0 26.5 8.8% 1% 10.03 111 gonzobill5 Jason Snelling RB39 9 7 0 53.8 7.8% 3% 10.04 112 guppy C Devin Hester WR40 7 12 2 65.9 12.6% 3% 10.05 113 guppy E Jeremy Shockey TE17 9 3 0 45.1 3.5% 2% 10.06 114 Truman Roy E. Williams WR41 9 9 2 55.3 11.3% 3% 10.07 115 Football Critic Jason Campbell QB18 5 6 1 52.95 8.2% 3% 10.08 116 fubar Willis McGahee RB40 5 6 0 60.4 6.8% 3% 10.09 117 guppy G Chris Chambers WR42 8 11 1 53.5 12.5% 3% 10.1 118 No Way Jose Correll Buckhalter RB41 7 8 1 38.9 8.3% 2% 10.11 119 guppy D Bernard Berrian WR43 10 4 0 37.2 5.3% 2% 10.12 120 gheemoney Alex Smith QB19 6 14 2 90.3 13.0% 4% 11.01 121 gheemoney Dwayne Bowe WR44 7 11 1 60.4 10.2% 3% 11.02 122 guppy D Lee Evans WR45 7 10 1 49.5 13.3% 3% 11.03 123 No Way Jose Kevin Walter WR46 7 7 0 52 7.3% 3% 11.04 124 guppy G Greg Olsen TE18 4 4 0 34.6 4.5% 2% 11.05 125 fubar Donnie Avery RB42 6 7 0 52.8 8.0% 3% 11.06 126 Football Critic Earl Bennett WR47 10 8 1 53.8 11.0% 3% 11.07 127 Truman Dustin Keller TE19 6 6 0 45 7.5% 2% 11.08 128 guppy E David Garrard QB20 4 5 0 32.15 5.9% 2% 11.09 129 guppy C Kevin Faulk RB43 7 8 0 40.4 8.4% 2% 11.1 130 gonzobill5 Michael Crabtree WR48 8 10 1 47.9 11.1% 2% 11.11 131 moleculo Vince Young QB21 4 9 0 57.7 11.3% 3% 11.12 132 guppy F Matt Ryan QB22 5 10 1 50.5 10.2% 3% 12.01 133 guppy F Marshawn Lynch RB44 6 7 0 31.6 7.1% 2% 12.02 134 moleculo Benjamin Watson TE20 6 3 0 47.7 3.8% 2% 12.03 135 gonzobill5 Brandon Pettigrew TE21 6 5 0 43.4 5.6% 2% 12.04 136 guppy C Chris Cooley TE22 4 6 0 40.2 6.3% 2% 12.05 137 guppy E Kenny Britt WR49 5 6 0 35.5 7.1% 2% 12.06 138 Truman Chad Henne QB23 5 4 0 30.8 5.0% 2% 12.07 139 Football Critic Marcedes Lewis TE23 5 3 0 27.4 4.1% 1% 12.08 140 fubar Tony Scheffler TE24 3 2 0 22.2 2.3% 1% 12.09 141 guppy G Matt Cassel QB24 6 13 2 53.2 14.8% 3% 12.1 142 No Way Jose Mohamed Massaquoi WR50 7 9 1 47.3 9.4% 2% 12.11 143 guppy D Michael Bush RB45 3 4 0 23.5 5.3% 1% 12.12 144 gheemoney Ryan Moats RB46 4 6 1 38.1 5.6% 2%No Way Jose
Footballguy
Good stuff here guys....I will however leave the math to you guys.
The way that I drafted was pretty simple
1. I wanted to have the highest scoring QB tandem and I did that with Rogers and Hasselbeck. Aaron Rogers scored so many points week to week and most of the leagues I played in this past year, the winner usually had a top 5 qb.
2. After that I went with the top WR,RB and then I wanted a top TE because they really dropped off after the first few went off the board.
3. Once I got the basic team I drafted guys that had big spikes on random weeks.
My team was ugly, but I guess it kinda worked
The way that I drafted was pretty simple
1. I wanted to have the highest scoring QB tandem and I did that with Rogers and Hasselbeck. Aaron Rogers scored so many points week to week and most of the leagues I played in this past year, the winner usually had a top 5 qb.
2. After that I went with the top WR,RB and then I wanted a top TE because they really dropped off after the first few went off the board.
3. Once I got the basic team I drafted guys that had big spikes on random weeks.
My team was ugly, but I guess it kinda worked
moleculo
Footballguy
1. Had gheemony drafted MJD instead of CJohnson, he would have been the 6th seed in the playoffs and lost in the first round, in addition to losing 11 all-play wins and two H2H wins. I went through and replaced every draft pick gheemony made with the next guy drafted on down the line – ie. Replace Addai (drafted by gheemony @ RB6) with ricky Williams (RB7) and so on…no other single player impacted gheemony from making it to the championship game. Oddly, without Ryan Moats (Mr Irrelevant, last pick of the draft), gheemoney would have lost the championship. The moral of the story is that CJohnson was absolutely instrumental to winning a championship. I suppose that should be obvious, but w/ cjohnson, gheemoney wins. W/o cjohnson, gheemoney does not. Can all of fantasy football simply break down to finding that one special player?
2. There’s a big difference between Reggie Wayne and Larry Fitzgerald. These guys were drafted back to back at 1.11 and 1.12. Wayne accounted for 16 all-play wins for moleculo, Fitzgerald accounted for 31 all-play wins for Guppy F. If I swap the players and teams, moleculo would have won 5 additional all-play and one H2H games w/ Fitzgerald, whereas guppy F would have lost 6 all-play and one H2H w/ Wayne. Clearly, Fitzgerald was more valuable than Wayne, despite having nearly identical end of season numbers. Fitzgerald had a slightly lower stdev, meaning he was more consistent. I talked about the games above baseline number before – looking at the actual games over baseline (not the one calculated assuming normal distribution), Wayne had 10 games over the baseline of 11.5; Fitzgerald had 13 games over the baseline. Looking at some other single game numbers, Wayne had 4 games over 30, Fitzgerald had one. Wayne had 4 games with < 10, Fitzgerald had 2. So, in terms of winning as many FF games as possible, you want guys who consistently score high, not guys who have a few really high scoring games and a few clunkers. This is probably obvious, but it should demonstrate that end of season numbers may not be all that helpful at the micro level.
3. Dallas Clark accounted for 29 all-play wins, consistent with his draft piers. IMO, he justified being worth a 1st round selection. I think the days of only a few TE’s being worthy of a high pick are over - TE’s should be considered as being valuable much early than we are generally comfortable with. As far as the season goes, Guppy C (whom drafted Clark in the 1st) finished 4th in all-play wins, 4th in H2H wins, 3rd in overall points, and lost in the semi-finals, so drafting TE early paid off.
4. gheemoney waited and drafted a starting qb late. Clearly, with the benefits of hindsight, QBBC is a great option.
5. During the draft, Jamaal Charles and Jonathan Stewart jumped Ryan Grant and Thomas Jones – i.e. they were selected before Grant and Jones, despite having lower end of season totals. Subbing in Grant or Jones in Charles place, football Critic would have won three additional all-play games. Subbing in for Jonathan Stewart yielded similar results. Season long, drafting Grant/Jones was only a marginally better proposition than Grant/Stewart, despite being having significantly better end of season numbers. Of course, Football Critic did this as a plan to platoon Brown (good 1st half of the season) w/ Charles (good 2nd half). Had Football Critic drafted Ryan Grant and Cedric Benson instead of Charles/Brown, he would have had 8 additional all-play games. I think that platooning RB’s like Football Critic was trying to do was a good idea, but he probably should have waited another round or so to do it.
6. Gonzo bill selected Jerome Harrison over LaDanian Tomlinson. Harrison was good for 6 more all-play wins vs Tomlinson, and 11 overall points.
7. Roders, Brees, and Manning were all very close in terms of end of season points, but Rodgers accounted for many more all-play and H2H wins. I believe this is a function of Hasselbeck matching up very well with Rodgers, so good job on that.
8. Moleculo didn’t draft a RB until the end of the 5th (although Guppy F did the same). This appears to be where my draft went wrong, and I don’t think I could have fixed it. Had I taken Thomas Jones instead of Celek in the 3rd and then Visanthe Shiancoe instead of Bush in the 6th, I would have had 8 additional all-play wins, three additional H2H wins, and I would have made the playoffs.
2. There’s a big difference between Reggie Wayne and Larry Fitzgerald. These guys were drafted back to back at 1.11 and 1.12. Wayne accounted for 16 all-play wins for moleculo, Fitzgerald accounted for 31 all-play wins for Guppy F. If I swap the players and teams, moleculo would have won 5 additional all-play and one H2H games w/ Fitzgerald, whereas guppy F would have lost 6 all-play and one H2H w/ Wayne. Clearly, Fitzgerald was more valuable than Wayne, despite having nearly identical end of season numbers. Fitzgerald had a slightly lower stdev, meaning he was more consistent. I talked about the games above baseline number before – looking at the actual games over baseline (not the one calculated assuming normal distribution), Wayne had 10 games over the baseline of 11.5; Fitzgerald had 13 games over the baseline. Looking at some other single game numbers, Wayne had 4 games over 30, Fitzgerald had one. Wayne had 4 games with < 10, Fitzgerald had 2. So, in terms of winning as many FF games as possible, you want guys who consistently score high, not guys who have a few really high scoring games and a few clunkers. This is probably obvious, but it should demonstrate that end of season numbers may not be all that helpful at the micro level.
3. Dallas Clark accounted for 29 all-play wins, consistent with his draft piers. IMO, he justified being worth a 1st round selection. I think the days of only a few TE’s being worthy of a high pick are over - TE’s should be considered as being valuable much early than we are generally comfortable with. As far as the season goes, Guppy C (whom drafted Clark in the 1st) finished 4th in all-play wins, 4th in H2H wins, 3rd in overall points, and lost in the semi-finals, so drafting TE early paid off.
4. gheemoney waited and drafted a starting qb late. Clearly, with the benefits of hindsight, QBBC is a great option.
5. During the draft, Jamaal Charles and Jonathan Stewart jumped Ryan Grant and Thomas Jones – i.e. they were selected before Grant and Jones, despite having lower end of season totals. Subbing in Grant or Jones in Charles place, football Critic would have won three additional all-play games. Subbing in for Jonathan Stewart yielded similar results. Season long, drafting Grant/Jones was only a marginally better proposition than Grant/Stewart, despite being having significantly better end of season numbers. Of course, Football Critic did this as a plan to platoon Brown (good 1st half of the season) w/ Charles (good 2nd half). Had Football Critic drafted Ryan Grant and Cedric Benson instead of Charles/Brown, he would have had 8 additional all-play games. I think that platooning RB’s like Football Critic was trying to do was a good idea, but he probably should have waited another round or so to do it.
6. Gonzo bill selected Jerome Harrison over LaDanian Tomlinson. Harrison was good for 6 more all-play wins vs Tomlinson, and 11 overall points.
7. Roders, Brees, and Manning were all very close in terms of end of season points, but Rodgers accounted for many more all-play and H2H wins. I believe this is a function of Hasselbeck matching up very well with Rodgers, so good job on that.
8. Moleculo didn’t draft a RB until the end of the 5th (although Guppy F did the same). This appears to be where my draft went wrong, and I don’t think I could have fixed it. Had I taken Thomas Jones instead of Celek in the 3rd and then Visanthe Shiancoe instead of Bush in the 6th, I would have had 8 additional all-play wins, three additional H2H wins, and I would have made the playoffs.
Yup. You can't lose your league in the first round, but you can certainly win it.Aaron Rodgers accounted for 48 All Play wins, while Chris Johnson accounted for 41. The other 10 first rounders accounted for 26.4 wins, on average. The difference between Chris Johnson and an average 1st rounder- an MJD or an Adrian Peterson- was huge. You can absolutely win your league in the first round.
FUBAR
Footballguy
moleculo
Footballguy
if that's the case, there is no amount of players on anyone's roster that would be too much to trade for a guy like CJohnson. If you had him last year, you were easily the favorite to win the league. If you didn't have him, it would have been worth trading pretty much everyone you had to get him.So, I guess the take away is that if you can identify that player, make every effort to get him, regardless of the cost.*of course, this by no means implies that CJohnson will have as dominant a 2010, so buyer beware.
FUBAR
Footballguy
I had CJ in my dynasty league and didn't win the Super Bowl.
The CJ owner didn't win in the two leagues I am in. Let's not overstate the importance of the most dominant player in a year. You still have to have a strong team. It can't be full of replacent level players.
I can think of so many similar great years by players that didn't lead to championships: Brady, Manning, LT, Faulk, Alexander, Holmes, .... This certainly puts the value of this measurement in to qurstin.
I can think of so many similar great years by players that didn't lead to championships: Brady, Manning, LT, Faulk, Alexander, Holmes, .... This certainly puts the value of this measurement in to qurstin.
