Jump to content

2016 1st Round EDGE Prospect Athleticism Analysis


Recommended Posts

On ‎4‎/‎17‎/‎2016 at 7:38 PM, Kayoh said:

The R² is a really general number, though. That's looking at every single player and comparing their EDGE score to their AV. I'm much more concerned with my tiers. The thing is, I have the overall value, which is nice, but then how do you work a pass/fail into an R² value? My Tiers are dependent on both a player's overall EDGE score and whether they're above or below a certain threshold for the Agility score as well. Players in general that fall below the Agility threshold have nearly half the average AV than players above the Agility threshold. 

@KayohI enjoy your efforts; but i'm struggling with your regression analysis results (or maybe i'm reading other folks words). Yes, I know that a low regression doesn't necessarily mean something is bad--but as general as it is Low R Squared Values typically appear in psychology-based or human factors types of analysis. Let's face it human behavior is unpredictable. So a lower R Squared Value (less than 50%) is acceptable in behavioral based studies.

But typically physical processes do have a higher R Squared Value. They are inherently more stable and more predictable. In your case the tangibles (height, weight, forty...etc..) are all very stable and it seems you are assuming (probably correctly) a standard model fits based on stable measurements to predict potential EDGE success.

Now I don't know your entire formula--nor am I questioning it, or asking for it--but you are looking at strictly physical attributes of players, correct? So at the least you're R squared values should be in 50% or greater range, would you agree? And, no it's not a pass or fail situation

I use mini-tab regularly (i'm a Lean and Six Sigma Black Belt) and am simply curious. Regression analysis is how we test to see if our model is representative of the data.

Are you saying that you are not utilizing a linear model?

You have me curious

Link to post
Share on other sites
  • Replies 94
  • Created
  • Last Reply

Top Posters In This Topic

Top Posters In This Topic

Just now, MSalmon said:

@KayohI enjoy your efforts; but i'm struggling with your regression analysis results (or maybe i'm reading other folks words). Yes, I know that a low regression doesn't necessarily mean something is bad--but as general as it is Low R Squared Values typically appear in psychology-based or human factors types of analysis. Let's face it human behavior is unpredictable. So a lower R Squared Value (less than 50%) is acceptable in behavioral based studies.

But typically physical processes do have a higher R Squared Value. They are inherently more stable and more predictable.

Now I don't know your entire formula--nor am I questioning it, or asking for it--but you are looking at strictly physical attributes of players, correct? So at the least you're R squared values should be in 50% or greater range, would you agree? And, no it's not a pass or fail situation

I use mini-tab regularly (i'm a Lean and Six Sigma Black Belt) and am simply curious. Regression analysis is how we test to see if our model is representative of the data.

Are you saying that you are not utilizing a linear model?

You have me curious

last year when I did this I made a spreadsheet of only edge rushers who passed my major filters (<7.20 3 cone, >9'8" broad jump) and made a new formula for those players. I got my Rup to 0.76

Link to post
Share on other sites
Just now, Kayoh said:

last year when I did this I made a spreadsheet of only edge rushers who passed my major filters (<7.20 3 cone, >9'8" broad jump) and made a new formula for those players. I got my Rup to 0.76

Ahh...now .76 makes more sense. I read somewhere in the post, .14 (I think). That's pretty good R-squared value. Almost fits a pareto perfectly.

Are height and weight still a large factor in your model? Or degraded a bit?

Link to post
Share on other sites
Just now, MSalmon said:

Ahh...now .76 makes more sense. I read somewhere in the post, .14 (I think). That's pretty good R-squared value. Almost fits a pareto perfectly.

Are height and weight still a large factor in your model? Or degraded a bit?

I don't remember. I lost everything earlier this month when my laptop's hard drive failed. I didn't have anything backed up. It really sucks but I'm working on re-making the spreadsheet now and then I'm gonna end up adding more to it. I'm also being more precise with age now so rather than just having 21/22/23 I'm gonna have decimals to get a better idea of how young a guy is coming into the league.

here's something you might be interested in. I posted it on reddit but figured I'd be met with a hostile response here so I never bothered posting it here.

Quote

I separated the 4-3 DEs who pass the 3C/BJ filter from everybody else. Specifically, I'm using players who were drafted to play 4-3 DE, not necessarily concerning myself with the position they played in college. The entire goal of this is to get as accurate as possible at predicting which college players will go on to contribute as much as possible in their rookie contracts to the team that drafts them. Basically a risk/value assessment for the draft in an attempt to demystify the scouting process.

Anyway, I re-worked my formula using only that group. My new R2 comparing the result of my formula to a player's rookie contract sacks is up to 0.76. The NFL's R2 when you compare round selected to rookie contract sacks is 0.25. Theoretically, using my 3C/BJ filter in combination with my new re-worked formula is more than 3x as good as the NFL at projecting which 4-3 DE prospects will get the most sacks during their rookie contracts.

In layman's terms, using my sample data (all drafts 2003-2012) and take the top 26 players drafted (from Mario Williams who was drafted 1st overall to Da'Quan Bowers who was 51st overall, and every other 4-3 DE prospect from 2003-2012 who was taken in the top 50) and compare that group to the 26 players in this sample that pass my BJ/3C filter, my group has more career sacks. On the other hand, the fewer players you look at, the more accurate my formula becomes. Top 15 highest drafted prospects average 17.5 sacks a piece, top 15 highest graded prospects based on my formula average 26 sacks a piece. Top 10 it's 20 vs. 31. Top 5 it's 19 vs. 39.

Now, all that being said, I'm not just throwing away the rest of the players. The players who fail the broad jump benchmark but pass the 3 cone benchmark are in one spreadsheet. Players who pass broad jump but fail 3 cone are in another. Players who fail both are in a fourth. I'll eventually have all four of these spreadsheets for 3-4 OLBs as well. The idea is to be as precise as possible. My current R2 values for 4-3 DEs are as follows: 0.76 (pass both) vs. the NFL's 0.25, 0.47 (pass 3c fail BJ) vs. the NFL's 0.11, 0.57 (pass BJ fail 3c) vs. the NFL's 0.01, and 0.51 (fail both) vs. the NFL's 0.25. The fact that I'm only using predictive numbers and not really "cheating" by using round selected in my formula makes it an even bigger deal that the correlations are this strong.

Anyway, I just really wanted to give you guys an update on this. I can also use this to try and predict who's going to sort of "break out" in the near future. Devin Taylor is consistently at the top of my 4-3 DE score. 95th percentile. He's also going to be taking on more of a full time role this year from what I understand. Fully expecting him to have a huge year. Barkevious Mingo is another guy who consistently shows up as a "why does he not have more production?" kind of guy. Kony Ealy is probably going to have a great season too, the 4-3 DE formula really likes him. 75th percentile. It's obviously not perfect given that my R2 isn't a perfect 1.0 or anything like that, but at the very least the players who pass the 3C/BJ filter having an R2 of 0.76 is pretty ******* insane.

If you guys have an interest in this I can try to keep you updated here. Probably be a little while before another update since draft season has officially come to an end, but the analytics bug has bit me pretty hard so you never know. Also kinda curious if there's anything I can actually do with this info other than just kinda e-bragging. I'm sure there are NFL teams that'd love a competitive advantage and clearly they don't have this kind of info or they'd be using it already. ****, even Seattle drafted Frank Clark last season who only grades out in the 54th percentile of my 4-3 DE (pass both) formula. Meanwhile Danielle Hunter literally breaks my ******* scale, grading more than 25% better than any 4-3 DE prospect in my entire 10 year sample size. That should be interesting to watch play out.

 

Link to post
Share on other sites
2 minutes ago, Kayoh said:

I don't remember. I lost everything earlier this month when my laptop's hard drive failed. I didn't have anything backed up. It really sucks but I'm working on re-making the spreadsheet now and then I'm gonna end up adding more to it. I'm also being more precise with age now so rather than just having 21/22/23 I'm gonna have decimals to get a better idea of how young a guy is coming into the league.

here's something you might be interested in. I posted it on reddit but figured I'd be met with a hostile response here so I never bothered posting it here.

 

cool, thanks! I'll read. Nah, i'm very interested in perfecting predictability of any process, person or thing. Never boring to read. I definitely am not an expert!

Do you have mini-tab? Makes this easier than a spreadsheet!

Link to post
Share on other sites
1 minute ago, MSalmon said:

cool, thanks! I'll read. Nah, i'm very interested in perfecting predictability of any process, person or thing. Never boring to read. I definitely am not an expert!

yeah, I never found a significance f either sadly. I'm pretty bummed about losing my data. I just hope the new spreadsheet can be even better than the last. :)

Link to post
Share on other sites
Just now, Kayoh said:

yeah, I never found a significant f either sadly. I'm pretty bummed about losing my data. I just hope the new spreadsheet can be even better than the last. :)

dude, buy mini-tab...much easier to work in, export and save in a cloud or external hard drive.

Cool stuff btw

Link to post
Share on other sites
Just now, Kayoh said:

I'm just using google sheets now.

The way that you are fascinated with statistics...mini-tab will pay for itself. All Six Sigma guys use it. Plus the control charts are fascinating. You can easily plop your data in UCL and LCL models and look for variability and the reasons why, as well.

Link to post
Share on other sites
Just now, MSalmon said:

The way that you are fascinated with statistics...mini-tab will pay for itself. All Six Sigma guys use it. Plus the control charts are fascinating. You can easily plop your data in UCL and LCL models and look for variability and the reasons why, as well.

I don't think you quite understand just how broke I am my friend.

Link to post
Share on other sites
Just now, Kayoh said:

I don't think you quite understand just how broke I am my friend.

hah, try the free trial. I think that you get 30 days free. If  I can find my old macros excel "mini-tab" lite, i'll PM you. I'm management now, so rarely play with cool stuff these days :( I do get paid better, though :)

Link to post
Share on other sites
Just now, MSalmon said:

hah, try the free trial. I think that you get 30 days free. If  I can find my old macros excel "mini-tab" lite, i'll PM you. I'm management now, so rarely play with cool stuff these days :( I do get paid better, though :)

I don't wanna do the free trial and get hooked and then hate my life post-trial period. That **** is so far out of my price range it's practically a ferrari at this point.

Link to post
Share on other sites
5 minutes ago, Kayoh said:

I don't wanna do the free trial and get hooked and then hate my life post-trial period. That **** is so far out of my price range it's practically a ferrari at this point.

Ha ha! Fair enough. You'd do well, and get paid well, if you earned a Lean/Six Sigma black belt. Villanova has a great online program. Those guys at the minimum are six figures, many consult and own their own business. Stat geeks get paid to stabilize and improve processes by eliminating variability 

Link to post
Share on other sites
Just now, MSalmon said:

Ha ha! Fair enough. You'd do well, and get paid well, if you earned a Lean/Six Sigma black belt. Villanova has a great online program. Those guys at the minimum are six figures, many consult and own their own business. Stat geeks get paid to stabilize and improve processes by eliminating variability 

dude the full price of mini-tab is about half what I made last year

Link to post
Share on other sites
On Sunday, April 17, 2016 at 11:08 AM, AtlantaFanatic said:

That's not what she said.








So the exit is over here? I'll show myself out.

She said to give her nine inches and make it hurt - so I f#*kd her three times and slapped her....

 

 

I will be here all week - don't forget to tip your servers.

Link to post
Share on other sites
On 4/17/2016 at 6:57 PM, Kayoh said:

Every 1st rounder since 1999 who doesn't fall in tiers 1 or 2:
 

  • Nick Perry
  • Brian Orakpo
  • Vernon Gholston
  • Adrian Clayborn
  • Aaron Maybin
  • Jerry Hughes
  • Ryan Kerrigan
  • Mathias Kiwanuka
  • Shea McClellin
  • Aldon Smith
  • Erik Flowers
  • Jason Pierre-Paul
  • Anthony Spencer
  • Jarvis Moss
  • Napoleon Harris
  • Erasmus James
  • Ebenezer Ekuban
  • Brandon Graham
  • Lawrence Jackson
  • Jerome McDougle
  • Datone Jones
  • Larry English
  • Quinton Coples
  • Derrick Harvey
  • Bjoern Werner
  • Robert Ayers
  • Jarvis Jones

So out of that entire group, you've got Orakpo, Kerrigan, Aldon Smith, JPP, and maybe Napoleon Harris who've lived up to their 1st round billing. 5/27. And even then, if we're going by AV, the only players who averaged an AV higher than 7 (for a frame of reference, Patrick Kerney's average AV was 7.2) are Kerrigan & JPP. So by simply punching numbers into a spreadsheet, I've manage to find a way to avoid some of the biggest busts in the last DECADE (Perry, Gholston, Maybin, etc) while only needing to give up the possibility of drafting a small handful of players that are generally pretty good, but not great.

Tier 1 players average a yearly AV of almost 8.7. Nobody on that entire list has a yearly AV that high. Kerrigan is the highest at 7.8. JPP through his first 5 years even, before he blew up his hand, only averaged 8 AV per year

OK, Kayoh,  Bottom line, who are the edge rushers, in the 2017 draft, that fit into your first two tiers?

I am particularly interested in TJ Watt, Derek Rivers and Jordan Willis because we have a legitimate shot of drafting one of them.  

Link to post
Share on other sites
5 hours ago, etherdome said:

OK, Kayoh,  Bottom line, who are the edge rushers, in the 2017 draft, that fit into your first two tiers?

I am particularly interested in TJ Watt, Derek Rivers and Jordan Willis because we have a legitimate shot of drafting one of them.  

I posted this earlier in the thread. I'll let you know in a couple weeks, brother.

d1098a563645d50c9ca6b7e37a2de39e.png

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


×
×
  • Create New...