switz
Footballguy
Bettis was elite at one time, but played longer in the NFL than he played at an elite level. It doesn't mean he was not an elite RB. (see Edgerrin James)
Last edited by a moderator:
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!
An Objective Measure of Talent in RBs (1 Viewer)
- Thread starter Dinsy Ejotuz
- Start date
Bettis was elite at one time, but played longer in the NFL than he played at an elite level. It doesn't mean he was not an elite RB. (see Edgerrin James)
bonesman
Footballguy
Dinsy Ejotuz
Footballguy
Just to explain the outputs Chase was asking about above, I used peak three years when I was looking at all this since what I was interested in wasn't sustained mediocrity, but elite talent. Using the three best seasons let me ignore things like injuries, being buried behind a great back for a couple years and etc.
puckalicious
Footballguy
wdcrob -
I understand your unwillingness to share your summarized data from nfldraftscout, so I went out and compiled it myself to look at a few things. I used a Kriging surface analysis to compute the ANOVA (analysis of variance) on all the combine variables (height, weight, 40 time, bench reps, vert, broad, shuttle, cone) as well as computed variables such as BMI and Power metric to determine what was most relevant to the draft pick. I used only the players with a complete set of data, so unfortunately many highly drafted players were eliminated since they tend to not run all the drills.
The results:
BMI: 37%
Noise: 35%
40 time: 11%
Weight: 6%
Broad Jump: 6%
What this means is that for the sample of data used, roughly a third of the output (draft pick) can be described by BMI. Another third is described by unknown factors (noise), and the other third is described by several other factors such as 40 time, weight, broad jump, etc. Another important piece of the puzzle is that it is not converged yet - we need more data points. The list of drafted players with complete combine #'s is only about 50. Typically at least 100 points are needed in this type of analysis.
So this helps show that your RB categorization leans heavily on BMI. Since draft pick appears pretty dependent on BMI, your power metric is heavily influenced by BMI, and you also filter by BMI.
A similar analysis using the same players for what variables correlate the most to career fantasy points/game shows:
Noise: 76%
Draft Pick: 17%
With everything else < 5%
Can you give a summary of the "best 3 years" fantasy points/game based on your data for the "elite" and not-so-elite players? Here is my list based on career averages.
Large/Elite:
Ronnie Brown - 12.3
Jamal Lewis - 13.4
LaDainian Tomlinson - 19.9
Fred Taylor - (don't have his data)
Deuce McAllister - 12.1
Willis McGahee - 11.8
Steven Jackson - 14.1
Chris Perry - 3.9
Larry Johnson - 16.7
Kevin Jones - 10.7
Average: 12.8 pts/G (unweighted)
Large/Backup:
LenDale White - 6.7
LaMont Jordan - 6.3
Chris Henry - 4.2
Travis Henry - 10.6
Musa Smith - 2.3
Kevan Barlow - 8.5
Chris Brown - 8.3
Rudi Johnson - 11.7
Sedrick Irvin - 3.9
Marion Barber - 10.5
Michael Turner - 2.9
Average: 6.9 pts/G (unweighted)
Small/Elite:
Adrian Peterson - 17.1
Ahman Green - (don't have his data)
Brandon Jackson - 4.2
Brian Westbrook - 13.0
Clinton Portis - 16.2
Domanick Williams - 15.4
Edgerrin James - 15.7
Frank Gore - 12.3
Laurence Maroney - 9.9
Marshawn Lynch - 13.6
Ryan Moats - 3.6
Shaun Alexander - 14.8
Thomas Jones - 8.8
Average: 12.1 pts/G (unweighted)
Small/Too Old:
Carnell Williams - 9.6
DeAngelo Williams - 7.3
Joseph Addai - 13.6
Michael Bennett - 6.4
Robert Edwards - (don't have his data)
Average: 9.2 pts/G (unweighted)
I think you've done an excellent job identifying many of the "noise" factors and categorized the RB's well. This is quite possibly the best tool we can possibly have for indentifying RB's that *SHOULD* excel in the NFL, real life and fantasy. Great work!
The next step probably would be to identify the factors of the NFL team the drafted player is on that could decide success or failure. OL rating, etc.
I understand your unwillingness to share your summarized data from nfldraftscout, so I went out and compiled it myself to look at a few things. I used a Kriging surface analysis to compute the ANOVA (analysis of variance) on all the combine variables (height, weight, 40 time, bench reps, vert, broad, shuttle, cone) as well as computed variables such as BMI and Power metric to determine what was most relevant to the draft pick. I used only the players with a complete set of data, so unfortunately many highly drafted players were eliminated since they tend to not run all the drills.
The results:
BMI: 37%
Noise: 35%
40 time: 11%
Weight: 6%
Broad Jump: 6%
What this means is that for the sample of data used, roughly a third of the output (draft pick) can be described by BMI. Another third is described by unknown factors (noise), and the other third is described by several other factors such as 40 time, weight, broad jump, etc. Another important piece of the puzzle is that it is not converged yet - we need more data points. The list of drafted players with complete combine #'s is only about 50. Typically at least 100 points are needed in this type of analysis.
So this helps show that your RB categorization leans heavily on BMI. Since draft pick appears pretty dependent on BMI, your power metric is heavily influenced by BMI, and you also filter by BMI.
A similar analysis using the same players for what variables correlate the most to career fantasy points/game shows:
Noise: 76%
Draft Pick: 17%
With everything else < 5%
Can you give a summary of the "best 3 years" fantasy points/game based on your data for the "elite" and not-so-elite players? Here is my list based on career averages.
Large/Elite:
Ronnie Brown - 12.3
Jamal Lewis - 13.4
LaDainian Tomlinson - 19.9
Fred Taylor - (don't have his data)
Deuce McAllister - 12.1
Willis McGahee - 11.8
Steven Jackson - 14.1
Chris Perry - 3.9
Larry Johnson - 16.7
Kevin Jones - 10.7
Average: 12.8 pts/G (unweighted)
Large/Backup:
LenDale White - 6.7
LaMont Jordan - 6.3
Chris Henry - 4.2
Travis Henry - 10.6
Musa Smith - 2.3
Kevan Barlow - 8.5
Chris Brown - 8.3
Rudi Johnson - 11.7
Sedrick Irvin - 3.9
Marion Barber - 10.5
Michael Turner - 2.9
Average: 6.9 pts/G (unweighted)
Small/Elite:
Adrian Peterson - 17.1
Ahman Green - (don't have his data)
Brandon Jackson - 4.2
Brian Westbrook - 13.0
Clinton Portis - 16.2
Domanick Williams - 15.4
Edgerrin James - 15.7
Frank Gore - 12.3
Laurence Maroney - 9.9
Marshawn Lynch - 13.6
Ryan Moats - 3.6
Shaun Alexander - 14.8
Thomas Jones - 8.8
Average: 12.1 pts/G (unweighted)
Small/Too Old:
Carnell Williams - 9.6
DeAngelo Williams - 7.3
Joseph Addai - 13.6
Michael Bennett - 6.4
Robert Edwards - (don't have his data)
Average: 9.2 pts/G (unweighted)
I think you've done an excellent job identifying many of the "noise" factors and categorized the RB's well. This is quite possibly the best tool we can possibly have for indentifying RB's that *SHOULD* excel in the NFL, real life and fantasy. Great work!
The next step probably would be to identify the factors of the NFL team the drafted player is on that could decide success or failure. OL rating, etc.
Dinsy Ejotuz
Footballguy
Wow... gonna have to take a longer look at this tonight. And I'll PM you.
Absolutely awesome topic but ANY system that depends on humans will NEVER be broken down by numbers. Political systems, the stock market and fantasy football success fall into this realm. Take a baseline of 5'10"/210 as a minimun and at least 5lbs for every 1" beyond 5'10". Some guys can do it if they're under 5'10" but they can't drop below 210lbs. Watching the guy play is the other 65% of the equation. Simple and easier for us morons!
Dinsy Ejotuz
Footballguy
I'm going to wait until some drafts are over to revise the first post in this thread and make some very specific player predictions, but the folks saying I was over fitting the data were more right than not. There's a simpler solution that doesn't involve some of the variables I'd said mattered.
And what interesting is that the key variable is the same as the WR research - the player's final playing weight. Why I didn't look at that for the RBs before now I don't know.
Anyhow, if you know what weight a RB will actually play at you can predict ability very well. Since you can't know that at the time of the draft, there will always be an element of uncertainty above and beyond the possibility of chronic injuries and horrific situations.
And what interesting is that the key variable is the same as the WR research - the player's final playing weight. Why I didn't look at that for the RBs before now I don't know.
Anyhow, if you know what weight a RB will actually play at you can predict ability very well. Since you can't know that at the time of the draft, there will always be an element of uncertainty above and beyond the possibility of chronic injuries and horrific situations.
Last edited by a moderator:
