Wizard Posted March 22, 2010 Posted March 22, 2010 A STATISTICAL ANALYSIS I've made this as simple as possible realizing statistics can be boring too many people; in addition, I don't know much about a 4-3 or 3-4 defense, so I assume that everyone has a varying degree of statistical knowledge as well. Get past the first few paragraphs, and I think you will find an easy to understand (simplified) analysis of who the Bills will draft. If you hate reading a lot, then read the summary, evaluation, and conclusion. It saves time and headaches! For Statistic Guys & Non-Stat Guys...A Brief Background of the Study Background: I'm in a graduate level statistical analysis class, and I chose the NFL Draft as my topic. I took 30 draft sites (considered Pro or Media) by the Hail Redskins site that keeps track of all the mock drafts. In order to perform the analysis, I used both descriptive statistics (mean, median, mode, and range) and inferential statistics (Standard Deviation, z-scores, t-scores, and ANOVA analysis) in order to predict the likelihood of the Bills drafting a particular player. Sites Chosen: I isolated these 15 sites so that they would be automatically included. Part of my theory is that media/professional sites are more likely to have less bias. I won't go into those details, but I didn't want a site by a Bills Fan or a Raiders Fan. In theory, media people and "professional" mock draft people probably know a little bit more! The other 15 sites were selected through a stratified sample. I've not included the additional 15 sites for space purposes. Scoutsnotebook.com, Football Expert, Signature Sports, FF Toolbox, Draft Season, Universal Draft, Draft Season, Draft Zoo, Draft Tek, NFL Draft Dog, Pro Football Talk, Football's Future, The Huddle Report, Walter Football, ESPN Mock Draft Mel Kiper/Todd McShay (March 2010)Results of 19 Players Projected as 1st Round Picks by these 30 Sites (Simplified Results) If you are a statistics guru, send me an e-mail at loganandkara@gmail.com and I will give you all the results. The 19 Players Chosen Bradford, Suh, McCoy, Okung, Bulaga, Trent Williams, Haden, Campbell, Berry, McClain, Bryant, Pierre-Paul, Spiller, Morgan, A. Davis, Dan Williams, Graham, Iupati, Clausen Mean Results (Sum of all the site's placement of a player in the draft divided by the 30 sites) Results: #1. Bradford, Suh, McCoy, Okung, and Berry will not be available to us. There is about a 5% chance that one of them will fall to the Bills at #9.#2. These players based on mean averages of the 30 sites are the most likely to be available at the #9 spot. In other words, these are the best players who have the highest chances of being available for the Bills to draft. Please keep in mind that if you like Iupati, Anthony Davis, or Bruce Campbell, this is okay. These results are according to the 30 sites, not personal preferences. Haden, McClain, Paul, Morgan, Bryant, Spiller, Bulaga, Clausen (In that order) What this Means in simple terms Based on statistics, we can stop dreaming about Bradford, Suh, McCoy, Okung, and Berry. If the 30 sites are accurate in rating a player's ability, the best players according to probability available to the Bills will be Haden, McClain, Paul, Morgan, Bryant, Spiller, and Bulaga, Clausen (See Below). Otherwise, the other players, Trent Williams, Dan Williams, Anthony Davis, Graham, Iupati, Campbell will be available after the 9th pick. Haden (95% chance he'll be available at #9) McClain (about 88% chance he'll be available at #9) Paul (About 85% chance he'll be available at #9) Morgan (About 84% chance he'll be available at #9) Spiller (About a 80% chance he'll be available at #9) Bryant (About a 70% chance he'll be available at #9) Bulaga (About a 52% chance he'll be available at #9) Clausen (About a 40% chance he'll be available at #9) Other Results: Mode & Range (The most frequently occuring #. I added the second most frequent # just for informational purposes. The second # isn't a Mode. It is just information. The range goes from the highest draft slot to the lowest rated draft slot by the 30 sites) Bradford: #1, #4 (Mode) 1-6 (Range) Suh: #2, #3 (Mode) 1-3 (Range) McCoy #3, #2 (Mode) 2-6 (Range) Okung #4, #5 (Mode) 2-5 (Range) Bulaga #5 & #9 (Mode) 4-14 (Range) Trent Williams #13, #6 (Mode) 6-23 (Range) Haden #7, #10 (Mode) 5-17 (Range) Campbell #8, #23 (Mode) 8-27 (Range) Berry #5, #6 (Mode) 3-9 (Range) McClain #15, #11 (Mode) 7-25 (Range) Bryant #11, #12 (Mode) 5-27 (Range) Paul #10, #12 (Mode) 8-32 (Range) Spiller #14, #17 (Mode) 10-23 (Range) Morgan #16, #14 (Mode) 8-16 (Range) A. Davis #13, #23 (Mode) 9-23 (Range) D. Williams #20, #28 (Mode) 11-32 (Range) Graham #19, #26 (Mode) 16-26 (Range) Iupati #18, #21 (Mode) 11-27 (Range) Clausen #9, #14 (Mode) 4-29) What This Means Clausen and Bulaga are the most frequently favored players according to the Mode. The next most favored guy for the Bills to draft is Anthony Davis. Summary 1. Bradford, Suh, McCoy, Berry, and Okung are a pipedream 2. If we believe in the Best Player Available Philosophy, then the Bills will be looking at at least four of these seven (7) players come draft time. (Haden, McClain, Paul, Morgan, Spiller, Bulaga, and Clausen). These would be the "best" players available. Haden, McClain, Paul, and Morgan are the most likely statistically to still be available at #9 because Bulaga and Clausen both have enough predictions that statisically they have a better chance of going before #9. 3. If the Bills are going for positional need and talent, then the Bills should be drafting Bulaga or Clausen. After that it depends on how someone personally thinks Morgan, Paul, or McClain would fit into a 3-4 defense. Evaluation: Obviously, there could be a run on OT's that would be hard to statistically measure. But, if statistically analysis provides in insight it is this information: 1. OT's- Okung will be gone by #9. Bulaga has a 52% chance of being available at #9. Campbell has a 58% chance of being available (The Raiders alter this statistic significantly). If the Bills like A. Davis or Trent Williams it is at least 80% likely that either will be available. 2. NT-Dan Williams won't go higher than #11 out of 30 sites. He is most likely going to be available at #20, or #28 by the Mode. By the Mean average, he'll be available between #17-#18. In other words, if NT is the priority, then if the Bills trade back to #17 there is still a 80%+ chance Dan Williams will be there in the 1st Round. 3. ILB/OLB/DE Paul & McClain should be available at #9. Morgan would be available too. All three are rated pretty close by Mean average availability. If the Bills are interested, then they take the best player to fit their 3-4 system. Personally, I would take Morgan. Graham can be had by #16-#19, if this is the Bills choice to move back. 4. QB- Bradford won't be available. Clausen is about 40% likely to be available. In other words, if the Bills are sold on him, then he is the best player with a reasonable probability of being available. 5. WR/RB/G Spiller and Bryant are undeniably talented. Both will be available but don't trade down past #14; otherwise, both of them will statistically be gone. Iupati will be available at #9, but he would be a reach. The Bills can trade back to #17 and still have a 90% chance of drafting him. Conclusion: If the Bills don't like what is in front of them at #9, then there are a lot of talented players including Dan Williams, Iupati, Brandon Graham, and one of the following: Trent Williams, Bruce Campbell, or Antony Davis available at #17.
Coach Tuesday Posted March 22, 2010 Posted March 22, 2010 Have you applied this analysis to previous years' drafts (and mocks) to test its accuracy? I assume yes but I wasn't sure.
Wizard Posted March 22, 2010 Author Posted March 22, 2010 Have you applied this analysis to previous years' drafts (and mocks) to test its accuracy? I assume yes but I wasn't sure. No. I haven't yet to previous draft year (mocks)..but I do plan on doing a comparison the week before the this draft (April 15) to this report. And, then I plan on doing an analysis after the 1st round is over. I will be posting these around the 20th of April, and the final one a week after the draft to compare the results. If my hypothesis has any statistical validity, my April 15 Round #1 should be more accurate than almost any mock draft, if (and it is a big IF), the mocks of professional drafters and media are more accurate than fan mock drafts.
Coach Tuesday Posted March 22, 2010 Posted March 22, 2010 No. I haven't yet to previous draft year (mocks)..but I do plan on doing a comparison the week before the this draft (April 15) to this report. And, then I plan on doing an analysis after the 1st round is over. I will be posting these around the 20th of April, and the final one a week after the draft to compare the results. If my hypothesis has any statistical validity, my April 15 Round #1 should be more accurate than almost any mock draft, if (and it is a big IF), the mocks of professional drafters and media are more accurate than fan mock drafts. You probably would also want to weight the various mocks you're using according to their previous accuracy, no?
Green Lightning Posted March 22, 2010 Posted March 22, 2010 That was a hell of an effort! Thanks for taking the time and being so transparent in your presentation. It's nice to read an intelligent post like this. I've copied it and will see how things stack up on draft day! I'm sure the nattering nabobs will weigh in and challenge it but I for one appreciate your work.
Wizard Posted March 22, 2010 Author Posted March 22, 2010 You probably would also want to weight the various mocks you're using according to their previous accuracy, no? Yes. I will only be doing the comparison between the previous 30 Mocks selected. But, yes, the validity of my study's accuracy will have to compare March data with April data and then a comparison to the actual Draft results.
Trader Posted March 22, 2010 Posted March 22, 2010 A STATISTICAL ANALYSISI've made this as simple as possible realizing statistics can be boring too many people; in addition, I don't know much about a 4-3 or 3-4 defense, so I assume that everyone has a varying degree of statistical knowledge as well. Get past the first few paragraphs, and I think you will find an easy to understand (simplified) analysis of who the Bills will draft. If you hate reading a lot, then read the summary, evaluation, and conclusion. It saves time and headaches! For Statistic Guys & Non-Stat Guys...A Brief Background of the Study Background: I'm in a graduate level statistical analysis class, and I chose the NFL Draft as my topic. I took 30 draft sites (considered Pro or Media) by the Hail Redskins site that keeps track of all the mock drafts. In order to perform the analysis, I used both descriptive statistics (mean, median, mode, and range) and inferential statistics (Standard Deviation, z-scores, t-scores, and ANOVA analysis) in order to predict the likelihood of the Bills drafting a particular player. Sites Chosen: I isolated these 15 sites so that they would be automatically included. Part of my theory is that media/professional sites are more likely to have less bias. I won't go into those details, but I didn't want a site by a Bills Fan or a Raiders Fan. In theory, media people and "professional" mock draft people probably know a little bit more! The other 15 sites were selected through a stratified sample. I've not included the additional 15 sites for space purposes. Scoutsnotebook.com, Football Expert, Signature Sports, FF Toolbox, Draft Season, Universal Draft, Draft Season, Draft Zoo, Draft Tek, NFL Draft Dog, Pro Football Talk, Football's Future, The Huddle Report, Walter Football, ESPN Mock Draft Mel Kiper/Todd McShay (March 2010)Results of 19 Players Projected as 1st Round Picks by these 30 Sites (Simplified Results) If you are a statistics guru, send me an e-mail at loganandkara@gmail.com and I will give you all the results. The 19 Players Chosen Bradford, Suh, McCoy, Okung, Bulaga, Trent Williams, Haden, Campbell, Berry, McClain, Bryant, Pierre-Paul, Spiller, Morgan, A. Davis, Dan Williams, Graham, Iupati, Clausen Mean Results (Sum of all the site's placement of a player in the draft divided by the 30 sites) Results: #1. Bradford, Suh, McCoy, Okung, and Berry will not be available to us. There is about a 5% chance that one of them will fall to the Bills at #9.#2. These players based on mean averages of the 30 sites are the most likely to be available at the #9 spot. In other words, these are the best players who have the highest chances of being available for the Bills to draft. Please keep in mind that if you like Iupati, Anthony Davis, or Bruce Campbell, this is okay. These results are according to the 30 sites, not personal preferences. Haden, McClain, Paul, Morgan, Bryant, Spiller, Bulaga, Clausen (In that order) What this Means in simple terms Based on statistics, we can stop dreaming about Bradford, Suh, McCoy, Okung, and Berry. If the 30 sites are accurate in rating a player's ability, the best players according to probability available to the Bills will be Haden, McClain, Paul, Morgan, Bryant, Spiller, and Bulaga, Clausen (See Below). Otherwise, the other players, Trent Williams, Dan Williams, Anthony Davis, Graham, Iupati, Campbell will be available after the 9th pick. Haden (95% chance he'll be available at #9) McClain (about 88% chance he'll be available at #9) Paul (About 85% chance he'll be available at #9) Morgan (About 84% chance he'll be available at #9) Spiller (About a 80% chance he'll be available at #9) Bryant (About a 70% chance he'll be available at #9) Bulaga (About a 52% chance he'll be available at #9) Clausen (About a 40% chance he'll be available at #9) Other Results: Mode & Range (The most frequently occuring #. I added the second most frequent # just for informational purposes. The second # isn't a Mode. It is just information. The range goes from the highest draft slot to the lowest rated draft slot by the 30 sites) Bradford: #1, #4 (Mode) 1-6 (Range) Suh: #2, #3 (Mode) 1-3 (Range) McCoy #3, #2 (Mode) 2-6 (Range) Okung #4, #5 (Mode) 2-5 (Range) Bulaga #5 & #9 (Mode) 4-14 (Range) Trent Williams #13, #6 (Mode) 6-23 (Range) Haden #7, #10 (Mode) 5-17 (Range) Campbell #8, #23 (Mode) 8-27 (Range) Berry #5, #6 (Mode) 3-9 (Range) McClain #15, #11 (Mode) 7-25 (Range) Bryant #11, #12 (Mode) 5-27 (Range) Paul #10, #12 (Mode) 8-32 (Range) Spiller #14, #17 (Mode) 10-23 (Range) Morgan #16, #14 (Mode) 8-16 (Range) A. Davis #13, #23 (Mode) 9-23 (Range) D. Williams #20, #28 (Mode) 11-32 (Range) Graham #19, #26 (Mode) 16-26 (Range) Iupati #18, #21 (Mode) 11-27 (Range) Clausen #9, #14 (Mode) 4-29) What This Means Clausen and Bulaga are the most frequently favored players according to the Mode. The next most favored guy for the Bills to draft is Anthony Davis. Summary 1. Bradford, Suh, McCoy, Berry, and Okung are a pipedream 2. If we believe in the Best Player Available Philosophy, then the Bills will be looking at at least four of these seven (7) players come draft time. (Haden, McClain, Paul, Morgan, Spiller, Bulaga, and Clausen). These would be the "best" players available. Haden, McClain, Paul, and Morgan are the most likely statistically to still be available at #9 because Bulaga and Clausen both have enough predictions that statisically they have a better chance of going before #9. 3. If the Bills are going for positional need and talent, then the Bills should be drafting Bulaga or Clausen. After that it depends on how someone personally thinks Morgan, Paul, or McClain would fit into a 3-4 defense. Evaluation: Obviously, there could be a run on OT's that would be hard to statistically measure. But, if statistically analysis provides in insight it is this information: 1. OT's- Okung will be gone by #9. Bulaga has a 52% chance of being available at #9. Campbell has a 58% chance of being available (The Raiders alter this statistic significantly). If the Bills like A. Davis or Trent Williams it is at least 80% likely that either will be available. 2. NT-Dan Williams won't go higher than #11 out of 30 sites. He is most likely going to be available at #20, or #28 by the Mode. By the Mean average, he'll be available between #17-#18. In other words, if NT is the priority, then if the Bills trade back to #17 there is still a 80%+ chance Dan Williams will be there in the 1st Round. 3. ILB/OLB/DE Paul & McClain should be available at #9. Morgan would be available too. All three are rated pretty close by Mean average availability. If the Bills are interested, then they take the best player to fit their 3-4 system. Personally, I would take Morgan. Graham can be had by #16-#19, if this is the Bills choice to move back. 4. QB- Bradford won't be available. Clausen is about 40% likely to be available. In other words, if the Bills are sold on him, then he is the best player with a reasonable probability of being available. 5. WR/RB/G Spiller and Bryant are undeniably talented. Both will be available but don't trade down past #14; otherwise, both of them will statistically be gone. Iupati will be available at #9, but he would be a reach. The Bills can trade back to #17 and still have a 90% chance of drafting him. Conclusion: If the Bills don't like what is in front of them at #9, then there are a lot of talented players including Dan Williams, Iupati, Brandon Graham, and one of the following: Trent Williams, Bruce Campbell, or Antony Davis available at #17. Nice work now I know why my car insurance keeps going up!
1B4IDie Posted March 22, 2010 Posted March 22, 2010 Nice work. I use this site as another MockDraft database. The Hail to the Redskin site makes my skin crawl (XXVI) http://www.fftoolbox.com/nfl_draft/2010/nf...ft-database.cfm If you do any time of previous year's acuuracy study can you private message me. Being in a Dynasty League a read many mock drafts but I very rarely see a site account for the predictions. Its real easy to do a mock but which "experts" actually have a good track record. You couldn't just do a simple right and wrong as far as player is concerned you'd have to look at position the player drafted and also the next best player ranked at the position of need. (i.e. If Clausen and Bradford are off the board a team may trade down or not take a QB because the #3 ranked QB is a much lower ranked player as compared to players at other positions) It'd be really complicated to come up with a rating system but I'm sure you have some more projects to do.
Wizard Posted March 22, 2010 Author Posted March 22, 2010 Thanks for the information. Yes...once I get the foundation for my first study, I will have a better idea of what variables to tweak. I think my goal is to take the 10-15 best sites/magazines, and then do an analysis/rating system of which have the best track record. However, this is a step or two away from where I am currently at. Does anyone know what are considered the most reliable/accurate sites and/or most read sites? I picked the 15 from my study without a lot of knowledge other than the sites were purportedly coming from pro's, media personnel, and/or above the average fan such as myself. Thanks again. Wizard Nice work. I use this site as another MockDraft database. The Hail to the Redskin site makes my skin crawl (XXVI) http://www.fftoolbox.com/nfl_draft/2010/nf...ft-database.cfm If you do any time of previous year's acuuracy study can you private message me. Being in a Dynasty League a read many mock drafts but I very rarely see a site account for the predictions. Its real easy to do a mock but which "experts" actually have a good track record. You couldn't just do a simple right and wrong as far as player is concerned you'd have to look at position the player drafted and also the next best player ranked at the position of need. (i.e. If Clausen and Bradford are off the board a team may trade down or not take a QB because the #3 ranked QB is a much lower ranked player as compared to players at other positions) It'd be really complicated to come up with a rating system but I'm sure you have some more projects to do.
Bob in SC Posted March 22, 2010 Posted March 22, 2010 I am not impressed. Perhaps, if you didn't describe standard deviation as an inferential statistic, your predictions might have more credibility. Better forget the Bills for a while and concentrate on your studies, my friend!
Wizard Posted March 23, 2010 Author Posted March 23, 2010 I am not impressed. Perhaps, if you didn't describe standard deviation as an inferential statistic, your predictions might have more credibility. Better forget the Bills for a while and concentrate on your studies, my friend! Bob, you are right. Standard Deviation is a descriptive statistic. However, descriptive statistics also often serve a role in evaluation within inferential statistics. Forgive my semantics. If semantics is your only point of fault, then you win. If semantics is your hang-up for not providing an intelligent response, then I'm sorry that was your hang-up. Mountain or mole hill, Bob?
KOKBILLS Posted March 23, 2010 Posted March 23, 2010 Conclusion:If the Bills don't like what is in front of them at #9, then there are a lot of talented players including Dan Williams, Iupati, Brandon Graham, and one of the following: Trent Williams, Bruce Campbell, or Antony Davis available at #17. [/size][/font] Exactly...If there was ever a year to Trade down it's this one cause The Bills will need the extra Picks to Trade back up and get Mt. (Buffalo Bill) Cody later in the 1st...
PS 56 Posted March 23, 2010 Posted March 23, 2010 This was a great read... and a lot of work. Thanks for posting
Bob in SC Posted March 23, 2010 Posted March 23, 2010 No hang-up here (and thanks for the insight about the role of descrptive statistics in inferential analyses). As a fellow Bills fan, I appreciate your efforts to predict the results of the draft (just like the others that appear here) even though, if we are really honest about it, there are far too many known and unknown variables to insert into your statistical analysis, given your sample size. It is an interesting approach to what is, essentially, a crap-shoot. My objection to your post had far less to do with your mis-categorization of standard deviation (or z - scores for that matter) than to its condescending nature. It is what, in 30+ years of teaching this content, I have railed against, i.e., trying to impress others with our knowedge(?) of statistical analyses, and why it is superior to anyone else's "best guess." It remains an imprecise art, at best. Does this constitute an intelligent response by your definition?
KOKBILLS Posted March 23, 2010 Posted March 23, 2010 No hang-up here (and thanks for the insight about the role of descrptive statistics in inferential analyses). As a fellow Bills fan, I appreciate your efforts to predict the results of the draft (just like the others that appear here) even though, if we are really honest about it, there are far too many known and unknown variables to insert into your statistical analysis, given your sample size. It is an interesting approach to what is, essentially, a crap-shoot. My objection to your post had far less to do with your mis-categorization of standard deviation (or z - scores for that matter) than to its condescending nature. It is what, in 30+ years of teaching this content, I have railed against, i.e., trying to impress others with our knowedge(?) of statistical analyses, and why it is superior to anyone else's "best guess." It remains an imprecise art, at best. Does this constitute an intelligent response by your definition? Define response...
Kelso_Helmet Posted March 23, 2010 Posted March 23, 2010 you didn't factor in Tebow's "will to win" seriously tho'...that was a cool post. thanks for dumbing it down so even i could understand it...
Kelso_Helmet Posted March 23, 2010 Posted March 23, 2010 Bob, you are right. Standard Deviation is a descriptive statistic. However, descriptive statistics also often serve a role in evaluation within inferential statistics. Forgive my semantics. If semantics is your only point of fault, then you win. If semantics is your hang-up for not providing an intelligent response, then I'm sorry that was your hang-up. Mountain or mole hill, Bob? it's hard to do descriptive statistics in the NFL
San Jose Bills Fan Posted March 23, 2010 Posted March 23, 2010 Exactly...If there was ever a year to Trade down it's this one cause The Bills will need the extra Picks to Trade back up and get Mt. (Buffalo Bill) Cody later in the 1st... I would guess that at least half of the people on this board would like to see a trade down. But as has been pointed out hundreds of times in the last week, one needs a trading partner. And as this statistical analyses suggests, it's not whether you trade down...it's how far you can trade down and still have the player (or several of the players) you covet, available.
Wizard Posted March 23, 2010 Author Posted March 23, 2010 No hang-up here (and thanks for the insight about the role of descrptive statistics in inferential analyses). As a fellow Bills fan, I appreciate your efforts to predict the results of the draft (just like the others that appear here) even though, if we are really honest about it, there are far too many known and unknown variables to insert into your statistical analysis, given your sample size. It is an interesting approach to what is, essentially, a crap-shoot. My objection to your post had far less to do with your mis-categorization of standard deviation (or z - scores for that matter) than to its condescending nature. It is what, in 30+ years of teaching this content, I have railed against, i.e., trying to impress others with our knowedge(?) of statistical analyses, and why it is superior to anyone else's "best guess." It remains an imprecise art, at best. Does this constitute an intelligent response by your definition? If my response was condescending, I sincerely apologize. And, no I'm not trying to be sarcastic. Of course, the draft is an art, and there are too many variables too possibly make any perfect predictions on the draft. I'd be foolish to think so. I'm doing this for a class assignment, and I thought I would post it for people's thoughts, comments, or disagreement. I only care about learning in my assignment and hoping the Bills draft a quality player. However, my only disagreement with you is that statistics (if used appropriately and not to "twist" information) does provide generally a more educated guess to just stating a personal opinion. Otherwise, statistics wouldn't be used in just about every measurement, including the NFL Combine. It is far from perfect, but statistics can provide an advantage to speculation. And, if my presentation seemed to come across as a display of superior knowledge, please know that was never intended either. It was merely a sharing of some work I had done for a class with the intention of creating a lively discussion. Best, Wizard
Recommended Posts