Front Office Football Central

Front Office Football Central (http://forums.operationsports.com/fofc//index.php)
-   FOF9, FOF8, and TCY Discussion (http://forums.operationsports.com/fofc//forumdisplay.php?f=33)
-   -   What are the important combines for each position group? (http://forums.operationsports.com/fofc//showthread.php?t=89290)

Ben E Lou 09-22-2014 03:26 PM

What are the important combines for each position group?
 
This is based on a large sampling (more than 40 draft classes from both SP and MP), so I would think that a good bit of the noise is removed. I just did a simple correlation coefficient through Excel to determine the correlation between each combine and the maximum current rating that a player achieved during his career. Based on some more detailed looking at the data and running it through the MalcPow filters, I'm confident that this is quite accurate. There may be a stray case or two where one of the "also important" combines is actually the "most important" (or they're actually the same), but overall, I would think this would be helpful.

POSITION MOST IMPORTANT ALSO IMPORTANT LEAST IMPORTANT
QB Bench (10) Solecismic (28), Agility (7.80) Dash
RB Broad Jump (114) Dash (4.65), Agility (7.35), PosDrl (17) Solecismic
FB PosDrl (22) Dash (4.78), Bench (20), Broad Jump (104) Solecismic
TE Broad Jump (102) Dash (4.78), Bench (22), Agil(7.75) Solecismic
WR Dash (4.51) Agility (7.20), PosDrl (42) Broad Jump
C Dash (5.31) Bench (25), Agility (8.00) Solecismic
G Bench (27) Dash (5.27), Agility (7.90) Solecismic
OT Bench (28) Dash (5.27) , Agility (7.80), Broad Jump (84) Solecismic
P Dash (4.97) Bench (10) Agility
K Bench (9) Solecismic (23), Broad Jump (104) Agility
DE Bench (27) Agility (7.60), Dash (4.85) Solecismic
DT Agility (7.80) Bench (28) Solecismic
ILB Agility (7.60) Bench (21), Broad Jump (107) Dash
OLB Agility (7.40) Broad Jump (108), Bench (18), PosDrl (22) Solecismic
CB Bench (12) Dash (4.52), Agility (7.20), PosDrl (37) Broad Jump
S Agility (7.35) Dash (4.59), Bench (15), PosDrl (37) Broad Jump






aston217 09-22-2014 03:32 PM

Definitely not surprising to see agility in there for so many of the defensive positions. The game still seems to weigh Run Defense really heavily, including for safeties. Probably a similar thing going on with TEs and Broad Jump (Run Blocking) and RBs and Broad Jump (Endurance).

I don't think I would recommend going for run-stopping safeties necessarily, though (etc).

FBs & position drill was an interesting one.

Ben E Lou 09-22-2014 04:12 PM

Quote:

Originally Posted by aston217 (Post 2962591)
Definitely not surprising to see agility in there for so many of the defensive positions. The game still seems to weigh Run Defense really heavily, including for safeties. Probably a similar thing going on with TEs and Broad Jump (Run Blocking) and RBs and Broad Jump (Endurance).

I don't think I would recommend going for run-stopping safeties necessarily, though (etc).

FBs & position drill was an interesting one.

In most cases, the "most important" doesn't have a much stronger correlation coefficient than the "also important" ones. On safeties, for example, it's .55 for agility, and .50 for dash/bench. Just thought for newbies it would be good to highlight the highest correlation. LB, P, and WR are the only positions where the difference between the highest and second-highest correlation is more than .1.

MIJB#19 09-22-2014 04:23 PM

Does this mean these combine scores (and the corresponding bars, if still existing) are a tell of good players, telling us the corresponding bars have more weight towards the overall skill bar or is the truth somewhere in the middle?

LastWhiteSoxFanStanding 09-23-2014 12:20 AM

This chart is immensely helpful. Thank you!

BishopMVP 09-23-2014 01:33 PM

Hmm... On QB's and Bench, I remember weight training for QB's being super important for their early development. I wonder in FOF if a higher bench is just tied to being more developed, and thus more likely to get playing time from the AI. (Although the multiplayer data points help guard against this.)

Ben E Lou 09-23-2014 02:16 PM

Quote:

Originally Posted by BishopMVP (Post 2962773)
Hmm... On QB's and Bench, I remember weight training for QB's being super important for their early development. I wonder in FOF if a higher bench is just tied to being more developed, and thus more likely to get playing time from the AI. (Although the multiplayer data points help guard against this.)

Just checked. There's basically zero correlation between bench and % developed. The 0.02 correlation coefficient of my data set is very likely just a plain ol' 0 when the noise is removed. From doing some draft class imports, it's clear that %developed is heavily tied to height and weight. I would imagine age is also a factor. My guess is that there are no other factors tied to it.

GridGlory 09-23-2014 04:21 PM

Enjoying this & thanks! Of course some factors are more important that the current rating accounts for. Such as a very high 40 time for QBs. But that's another discussion.

BishopMVP 09-23-2014 05:07 PM

Quote:

Originally Posted by Ben E Lou (Post 2962788)
Just checked. There's basically zero correlation between bench and % developed. The 0.02 correlation coefficient of my data set is very likely just a plain ol' 0 when the noise is removed. From doing some draft class imports, it's clear that %developed is heavily tied to height and weight. I would imagine age is also a factor. My guess is that there are no other factors tied to it.

I guess Weight was the more important thing in TCY, and I just assumed bench would be tied to strength.

I do find the larger question of how to balance predictable development with just the right amount of variance interesting. And its nice to know these correlations are out there - when playing a single-player career, the small sample sizes can make some things feel like dumb luck. I feel I have a handle on drafting pretty well, if not quite MalcPow level, between the combines and Greg's DraftAnalyzer, but player development seems to be more of a crap shoot, versus some earlier FOF's where it was very linear barring injury.

biedrins 09-24-2014 05:28 AM

From my own investigations, % development is based on some inscrutable combination of height and weight with a modifier for position. There's a pattern to it, but it's not a case of, say, being close to an ideal weight, because IIRC when I had a bunch of guys with the same height and different weights it just cycled through the values for development every 40lbs or so (so a guy would have the same development at 180/220/260/300/340/380/420 and so on)

Ben E Lou 09-24-2014 07:27 AM

Quote:

Originally Posted by biedrins (Post 2962928)
... % development is based on some inscrutable combination of height and weight with a modifier for position. There's a pattern to it, but it's not a case of, say, being close to an ideal weight, because IIRC when I had a bunch of guys with the same height and different weights it just cycled through the values for development every 40lbs or so (so a guy would have the same development at 180/220/260/300/340/380/420 and so on)

This is basically what I saw. I didn't change out the ages, but when I mess around with weights for 6'2" QBs for example, it's...

200lbs: 9% (To be clear, I generated 10 QBs who were 6'2", 200 lbs. The game assigned all of them as 9% developed.)
201 lbs: 26% (Same thing...10 guys 6'2", 201 lbs...every one of them 26% developed)
202 lbs: 43%
203 lbs: 21%
204 lbs: 38%
205 lbs: 16%
206 lbs: 33%
207 lbs: 11%
208 lbs: 28%
209 lbs: 45%
210 lbs: 23%
211 lbs: 40%
212 lbs: 18%

Not sure if it would change with age, but MalcPow mentioned to me that he thinks it does.

Pyser 09-24-2014 12:09 PM

does % developed really mean anything?

sure, when its under 15% or so, that's something to take into account. but beyond that, isn't it a bit unreliable?

aston217 09-24-2014 01:39 PM

I didn't know % developed was that specifically tied to height/weight. That's neat!

% developed factors into my decision-making to an extent. There's nothing that's too reliable. For me, a low % developed means the player should have a higher bureau grade, but didn't simply because he was generated with low development. If all the other signs check out, that tilts in the player's favor. Not a huge factor, but it's something I consider whenever I ask myself, "Is it plausible that this player is as good as his bars make him look?" Average combine, 4.5 grade, 45% developed -- maybe not. Average combine, 4.5 grade, 15% developed -- maybe there's something there.

Ben E Lou 12-10-2014 12:00 PM

FOF7 Constraints Added
 
OK. I've updated the first post to add the bottom-tier constraints that I've observed in FOF7. A few of these are clearly different from FOF2K7, but most are identical. A few key things to understand about these numbers.


1. These don't mean "any player better than all of these will be useful." For example, the WR key constraints are dash<=4.51, agility<=7.20, and posdrl>=42. However, there will be WRs at 4.45/7.00/50 who are scrubs.

2. However, these DO mean "any player worse than any one of these will virtually never be a 50/50 player." I posted an example of this for QBs in MalcPow's thread this morning.

Quote:

Bench press of 10 or higher (4656 game-generated FOF7 QBs examined):

5.67% made it to 50/50 or better (264 out of 4656)
3.09% made it to 60/60 or better (144 out of 4656)
1.74% made it to 70/70 or better (81 out of 4656)

But when the Bench is 9 or lower, but not zero (4672 game-generated FOF7 QBs examined):

.30% made it to 50/50 or better (14 out of 4672)
.06% made it to 60/60 or better (3 out of 4672)
0 made it to 70/70 or better (0 out of 4672)

Basically, so few guys make it to 50/50 or 60/60 that don't hit even one of these constraints that it may just be scouting error or VSOL that gets them above those numbers.

3. In nearly all cases, the constraint is also the most common combine value, and often by a very wide margin. For example, the QB agility constraint value is 7.80. In my current data, there are 1,642 QBs with an agility score of 7.80. The next-highest value is the lowest possible: 8.40. That only occurs 202 times. Below that, there are a bunch of values that show up 140-170ish times. So despite there being 148 possible values (from 6.93 to 8.40), more than one out of six non-zero results are 7.80. This is one of the more extreme cases, but pretty much all of the constraints work like this.

4. Because these constraints exist and occur with such frequency, some of our existing mathematical ways of examining combine data aren't worth very much. Stated simply, in a world where an agility score of 7.81 for QB is just about a kiss of death, 7.80 is workable, but the average is 7.87, there are "above average" QBs with agility scores in the 7.81 to 7.86 range that I would never, ever want to draft.

aston217 12-10-2014 01:21 PM

Interesting, Ben, where do you get these numbers from? If you re-did Malcpow's study, would you be able to release your numbers? A few questions/comments:
  • Do they still reflect different degrees of "constraint"? For example, the 7.80 agility constraint on OTs was a kiss of death, being a line at 425 (20). Whereas QBs and 28 sole, that line was at 525 (60).
  • It gets complicated when combines are reflected by more than one bar. For example, PH being split two or three ways (sometimes among static bars). Going off Malcpow numbers again, a 14 bench for example, technically only means there's 0% chance of PH and BnR both being >= 60. If PH is 0, maybe BnR could be higher? I don't know, it's not strictly in the data...
  • ILB bench at 21 and OLB bench at 18 certainly didn't look like hard boundaries in Malcpow's data set. Even if they were, that would mean boundaries at 575 (80) and 550 (70), respectively.
  • Did LB position drill change from 27 to 22? That's a huge deal, if so...
  • I see where you might be getting 5.31 and 8.00 on Cs from Malcpow's thread, but I don't think there's enough data there to be sure. And I don't think any of the OL positions exhibited meaningful bench press constraints.
  • There weren't any K/P numbers in his thread either, so that's intriguing :)

Nemesis 12-10-2014 02:24 PM

Quote:

Originally Posted by aston217 (Post 2983631)
Interesting, Ben, where do you get these numbers from? If you re-did Malcpow's study, would you be able to release your numbers? A few questions/comments:
  • Do they still reflect different degrees of "constraint"? For example, the 7.80 agility constraint on OTs was a kiss of death, being a line at 425 (20). Whereas QBs and 28 sole, that line was at 525 (60).
  • It gets complicated when combines are reflected by more than one bar. For example, PH being split two or three ways (sometimes among static bars). Going off Malcpow numbers again, a 14 bench for example, technically only means there's 0% chance of PH and BnR both being >= 60. If PH is 0, maybe BnR could be higher? I don't know, it's not strictly in the data...
  • ILB bench at 21 and OLB bench at 18 certainly didn't look like hard boundaries in Malcpow's data set. Even if they were, that would mean boundaries at 575 (80) and 550 (70), respectively.
  • Did LB position drill change from 27 to 22? That's a huge deal, if so...
  • I see where you might be getting 5.31 and 8.00 on Cs from Malcpow's thread, but I don't think there's enough data there to be sure. And I don't think any of the OL positions exhibited meaningful bench press constraints.
  • There weren't any K/P numbers in his thread either, so that's intriguing :)


That Malcpow's study link is this thread. Did you mean Malcpow's Combine Benchmarks?

Subby 12-10-2014 02:49 PM

Quote:

Originally Posted by Ben E Lou (Post 2983616)
OK. I've updated the first post to add the bottom-tier constraints that I've observed in FOF7. A few of these are clearly different from FOF2K7, but most are identical. A few key things to understand about these numbers.

What do you have for DE bench? 27?

gstelmack 12-10-2014 02:53 PM

Quote:

Originally Posted by Ben E Lou (Post 2983616)
4. Because these constraints exist and occur with such frequency, some of our existing mathematical ways of examining combine data aren't worth very much. Stated simply, in a world where an agility score of 7.81 for QB is just about a kiss of death, 7.80 is workable, but the average is 7.87, there are "above average" QBs with agility scores in the 7.81 to 7.86 range that I would never, ever want to draft.


It would be relatively straightforward to add a a penalty for below-threshold combine numbers to Draft Analyzer, user-settable. So a 7.81 combine QB gets a -50 (or whatever the user wants) added to their total. If that would be useful. I could probably even color the rows.

aston217 12-10-2014 03:18 PM

^ yeah, that would help a lot, I think.

I don't play around with weights, but in draftman I color below-constraint combines (according to Malcpow numbers), as far as I can find them, as green. Quick at a glance visual cue to let me know a guy has a bad score, rather than having to think, "OK...7.87...that's black, but it shouldn't be."

Ben E Lou 12-10-2014 04:59 PM

Quote:

Originally Posted by aston217 (Post 2983631)
Interesting, Ben, where do you get these numbers from? If you re-did Malcpow's study, would you be able to release your numbers? A few questions/comments:
  • Do they still reflect different degrees of "constraint"? For example, the 7.80 agility constraint on OTs was a kiss of death, being a line at 425 (20). Whereas QBs and 28 sole, that line was at 525 (60).
  • It gets complicated when combines are reflected by more than one bar. For example, PH being split two or three ways (sometimes among static bars). Going off Malcpow numbers again, a 14 bench for example, technically only means there's 0% chance of PH and BnR both being >= 60. If PH is 0, maybe BnR could be higher? I don't know, it's not strictly in the data...
  • ILB bench at 21 and OLB bench at 18 certainly didn't look like hard boundaries in Malcpow's data set. Even if they were, that would mean boundaries at 575 (80) and 550 (70), respectively.
  • Did LB position drill change from 27 to 22? That's a huge deal, if so...
  • I see where you might be getting 5.31 and 8.00 on Cs from Malcpow's thread, but I don't think there's enough data there to be sure. And I don't think any of the OL positions exhibited meaningful bench press constraints.
  • There weren't any K/P numbers in his thread either, so that's intriguing :)

I don't have time to get into all of these details. Bottom line: none of this information came from the MalcPow thread. It supercedes it.

Ben E Lou 12-10-2014 05:00 PM

Quote:

Originally Posted by Subby (Post 2983653)
What do you have for DE bench? 27?

Yes. Fixed.

corbes 12-10-2014 06:12 PM

Quote:

Originally Posted by gstelmack (Post 2983656)
It would be relatively straightforward to add a a penalty for below-threshold combine numbers to Draft Analyzer, user-settable. So a 7.81 combine QB gets a -50 (or whatever the user wants) added to their total. If that would be useful. I could probably even color the rows.


Either way would be helpful. I've been using my own rough estimate of the constraints for a few FOOL drafts now and essentially begin Analyzer by marking players with any one below-threshold combine number with an "X" rating for "do not draft." A way of saving time with that process--either with a user-settable rating penalty or an identifying color--would be helpful...

gstelmack 12-10-2014 06:49 PM

Draft Analyzer pushed with a new weight to apply a penalty for below-threshold combines and coloring of those combines, may have just saved me from a huge mistake in the 4th round of the FOOL draft...

corbes 12-10-2014 07:38 PM

And colored chemistry groups... You are the wind beneath my wings

Ben E Lou 12-11-2014 05:04 AM

Quote:

Originally Posted by Ben E Lou (Post 2962589)
There may be a stray case or two where one of the "also important" combines is actually the "most important" (or they're actually the same)

Beginning the process of looking at the correlation coefficients on a larger dataset now. Having looked at QBs, it appears that there won't be any major differences, but that the quoted statement is correct. On a larger, cleaner dataset, it appears that Sole/Bench/Agility are of equal value for QBs. If there is one that is "most important", it's Sole. But they're so close in a sampling of over 12,000 QBs that trying to make the distinction would be a waste of time. I'll check the others and report if anything changes, but I don't see it happening.

Ben E Lou 12-13-2014 03:28 AM

Added Agil to TE. The overall correlation is quite a bit lower than Bench/Broad/Dash, but I just noticed that guys worse than 7.75 virtually *never* pan out. Agil isn't crucial for them to make it into the 70s, but they'll basically never hit the 50s without a 7.75. So I suppose the play there is to view 7.75 as a constraint, but don't be too concerned if it's not much above that.

IrishGuy65 12-13-2014 07:04 AM

I should read these forums more often. Now I see why, in recent drafts, people have jumped me in getting guys who normally would be around a long time in the past. I used to be able to wait until the 4th round to get a quality starter for my team on the defensive line, now they are all gone because you have people looking at the same things that I've been looking at for years. A lot of this stuff worked great for me pre FOF7...

gstelmack 12-13-2014 08:32 AM

TE Agil threshold added to Draft Analyzer.

gstelmack 12-13-2014 09:06 AM

PlayerTracker now supports reporting the thresholds in the UI and the Draft Class report. I'll be interested to see the FOOL 2014 draft class once Ex1 is past - after 3 seasons, the first draft class definitely has a couple of below-threshold players at the top, but I want to see what they look like after that fourth season of reporting. That seems to be a key point, I'm seeing a fair number of players who may "peak" in their 2nd or 3rd seasons, but I'm beginning to wonder how much FOF is lying to us about their real CUR value early on. Of course it will be difficult to tell that versus players who are "1 hit wonders", where maybe guys below the threshold can be that great rookie/2nd year player, but don't last much longer. So maybe instead of lying about CUR, FOF is correct, but instead models flame outs that aren't VSOD-related.

gstelmack 12-13-2014 09:17 AM

To give a specific example, I'm curious about this guy. Clearly his Pass Rush Strength matches his below-threshold combine, but everything else is pretty high. He went from 38/70 to a peak of 69/69, but showed 66/66 at the end of his 3rd season. Will he decline quickly, or is he simply the occasional exception to Ben's rule (part of the 0.3%)?


Ben E Lou 12-13-2014 11:20 AM

Heh. That's actually the one constraint that I debated going into long detail over, but decided not to so that I wouldn't cloud the issue. DT bench press is the "softest" of all the constraints posted above.

Here are the facts about DT bench press:
  • The percentages of 50+ players at 25, 26, and 27 are nearly identical: 1.49%, 1.41$, 1.52%.
  • But once you hit 28 and up, there's a dramatic increase: 5.69%, 7.51%, and it keeps increasing as the bench increases.
  • The samples are roughly the same size: 2,558 DTs combined with 25/26/27 vs. 2407 with 28. (Like I said, there is almost always a BIG player creation bias toward the bottom of the constraints.)
  • So few decent players at 23 and below that it's probably just VSOL or coach-related.
So in short, yes, there's a somewhat-significant chance that a low-bench DT will make it into the 50s or 60s. (No idea what happens after that; we'll see with the guy in FOOL.) But it's roughly four times less likely than if he hits the 28 constraint, and we're talking less than once every three seasons, and zero instances of the guy making it to 70.

Ben E Lou 12-13-2014 11:31 AM

And Combine Checker does a good job reflecting this:

FOF7 Combine Checker. (Link reflects that guy's combines.)

Note that there are only two guys under the constraints in each database showing as similar, and they aren't great matches. (Anything over .5 is just so-so at best.)

gstelmack 12-13-2014 12:22 PM

I do think part of the key with this particular player is that everything else is so high, with 2 red and 2 blue combines. That gives you 3 near-max bars, 2 average, and the one really bad bar. In most other cases, the other combines are rarely going to be that high to compensate for the very low pass rush strength.

aston217 12-13-2014 02:06 PM

Quote:

Originally Posted by gstelmack (Post 2984272)
To give a specific example, I'm curious about this guy. Clearly his Pass Rush Strength matches his below-threshold combine, but everything else is pretty high.



Unless I'm not understanding this correctly, Ben's looking at "what kinds of combines do 50+ (or whatever number) OVR players generated by FOF7 typically have".

If the combine correlations haven't been changed in FOF7, then a low bench score in a DT or in a C will tell you exactly what you expect and see: low strength bars.

Their inclusion on this list, then, might just be reflecting "Players with 50+ OVR rarely have poor strength bars." Or, "Players with poor OVR rarely have good strength bars."

So broadly, this is useful, especially in earlier rounds where you might be looking at a player and trying to decide if he's that stud OVR guy you hope.

But a lack of strength bars doesn't mean a lack in other areas, even if their OVR rarely gets up there. Your guy might be on the extreme side, but there are useful players with low strength bars (do these bars even do much?), just as there are useful TEs with low run blocking, useful DTs with low run stopping, etc. Look at the rest of the player to decide if he might be one of them or not.

RedKingGold 12-20-2014 08:20 AM

Quote:

Originally Posted by Ben E Lou (Post 2984242)
Added Agil to TE. The overall correlation is quite a bit lower than Bench/Broad/Dash, but I just noticed that guys worse than 7.75 virtually *never* pan out. Agil isn't crucial for them to make it into the 70s, but they'll basically never hit the 50s without a 7.75. So I suppose the play there is to view 7.75 as a constraint, but don't be too concerned if it's not much above that.


Almost on cue, it'll be interesting to follow the career of this guy in FOOL:

http://www.younglifenorthdekalb.com/...n_game_id=6766

Ben E Lou 12-20-2014 11:00 AM

Interesting case there, and one that makes me appreciate the player creation module in this game even more. My big draft class database now has over 14,000 TEs, yet there's really not a player quite like him. 10'6" is insane for TE broad jump and has the highest correlation to success for that position of any combine. (10'8" is still the max for TEs. Only 17 of those 14,000+ TEs achieved 10'6" or better.) In all those TEs, there's never been one worse than 7.75 but with a Brj in that stratosphere. Combine checker doesn't have constraint barriers yet when checking for matches, so most of the matches it's finding are 7.75 or better, and none are particularly good matches: FOF7 Combine Checker.

For more in the "this is a really weird one" category, there are 194 TEs in the db with raw scores of 6.3 or better. Not a single one of them has a single combine score that is worse than *any* of the four constraints above. This guy is really unique.

Ben E Lou 12-21-2014 07:38 AM

Just noticed a typo that made its way into Draft Analyzer: CB agility was at 7.35. Should be 7.20.

gstelmack 12-21-2014 09:13 AM

Fixed.

Ruckuz 12-21-2014 10:32 AM

Bench (10) Solecismic (28), Agility (7.80)
what do the numbers mean? Min values to look for?

Ben E Lou 12-21-2014 02:20 PM

Quote:

Originally Posted by Ruckuz (Post 2986311)
Bench (10) Solecismic (28), Agility (7.80)
what do the numbers mean? Min values to look for?

See post #14 in this thread for an explanation.

Front Office Football Central - View Single Post - What are the important combines for each position group?

Sharkn20 12-23-2014 01:56 PM

Any diferences between the diferents combines correlations? Or this nice study is based on a 100 combine correlation??

gstelmack 12-24-2014 01:11 PM

I believe the current interpretation of combine correlation is not that it affects the range of combine values for a given player talent level, but rather that it spreads them out more, flattening the bell curve if you will. So the average stays the same, but the standard deviation increases.

OldSchool 12-27-2014 08:30 AM

Fantastic info, thanks Ben.

garion333 12-31-2014 09:15 AM

Ben, without making the OP confusing is there any way you can post the correlations you found for the individual attributes?

I'd be curious to know what the dropoff is between, say, Dash in a WR to Agility.

Ben E Lou 12-31-2014 07:43 PM

It's more laziness than confusion in that I don't want to have to get into explaining correlation coefficients and tables don't post well here, so the effort to post it isn't something that I really want to deal with.

Here are the numbers for WRs.
Code:

Cor_Sole        Cor_Bench        Cor_Brdj        Cor_PosDrl        Cor_Agil        Cor_Dash
0.326331738        0.353612529        0.209346083        0.411168029        0.493956585        0.629523465


Ben E Lou 01-05-2015 02:19 PM

Quote:

Originally Posted by gstelmack (Post 2984272)
To give a specific example, I'm curious about this guy. Clearly his Pass Rush Strength matches his below-threshold combine, but everything else is pretty high. He went from 38/70 to a peak of 69/69, but showed 66/66 at the end of his 3rd season. Will he decline quickly, or is he simply the occasional exception to Ben's rule (part of the 0.3%)?


I've got him at 69/69 now in year 6. And as my database of draftees grows, it becomes more and more clear what an anomaly this dude is. I've got 1,333 DTs (300+ draft classes) with 25 bench press. Only 2 of them made it to 60/60. 0 made it to 65/65. None. Here are some numbers for DTs right below and a little above the 28 magic number.
Code:

POS        COUNT        BENCH        AVG        50+        55+        60+        65+        70+        75+
DT        832        21        9.8        0.6%        0.1%        0.0%        0.0%        0.0%        0.0%
DT        726        22        12.2        1.1%        0.1%        0.1%        0.0%        0.0%        0.0%
DT        944        23        12.6        0.8%        0.2%        0.2%        0.0%        0.0%        0.0%
DT        1343        24        14.1        0.9%        0.4%        0.0%        0.0%        0.0%        0.0%
DT        1333        25        16.2        1.3%        0.6%        0.2%        0.0%        0.0%        0.0%
DT        1286        26        17.6        1.5%        0.8%        0.4%        0.0%        0.0%        0.0%
DT        1003        27        18.5        1.7%        0.5%        0.0%        0.0%        0.0%        0.0%
DT        3539        28        23.2        5.7%        3.1%        1.4%        0.7%        0.5%        0.3%
DT        955        29        27.2        8.7%        6.0%        3.0%        1.7%        1.3%        0.7%
DT        921        30        28.1        9.6%        5.5%        3.3%        2.1%        1.4%        0.5%
DT        941        31        30.0        13.0%        7.8%        4.5%        3.0%        1.9%        1.1%


AVG=avg peakcur
50+/60+/etc. = percentage of players achieving that cur rating

In other words, I'm still sticking with 28 as a pretty hard constraint there for drafting. Roughly once every four seasons there's a DT with a bench under 27 who makes it to 50. It's really not worth bothering with them unless he's some freak like the guy that you bring up.

garion333 01-05-2015 02:59 PM

Quote:

Originally Posted by Ben E Lou (Post 2988464)
It's more laziness than confusion in that I don't want to have to get into explaining correlation coefficients and tables don't post well here, so the effort to post it isn't something that I really want to deal with.

Here are the numbers for WRs.
Code:

Cor_Sole        Cor_Bench        Cor_Brdj        Cor_PosDrl        Cor_Agil        Cor_Dash
0.326331738        0.353612529        0.209346083        0.411168029        0.493956585        0.629523465



This is awesome. I'm not a big math guy (since college ended), but I do understand the basics of correlation having helped my gf with her MBA math classes recently. So, this makes sense to me.

This is my small petition to get the rest of the data.

Ben E Lou 01-05-2015 04:02 PM

I actually ended up needing it for a little project I'm working on, so the effort to post it is negligible now. I never finished going back through and comparing to the original post, so if someone wants to do that, be my guest.

Code:

posgrp        pos        sole        bench        brdj        posspec        agil        dash        dev        adjgrd
1        QB        0.435        0.428        0.329        0.246        0.415        0.191        0.295        0.726
2        RB        0.411        0.388        0.495        0.428        0.427        0.460        0.111        0.756
3        FB        0.280        0.467        0.502        0.484        0.313        0.400        0.208        0.772
4        TE        0.338        0.524        0.582        0.448        0.456        0.549        0.158        0.781
5        WR        0.326        0.354        0.209        0.411        0.494        0.630        0.147        0.796
6        C        0.022        0.581        0.447        0.000        0.527        0.642        0.157        0.783
7        G        0.011        0.588        0.372        0.000        0.402        0.575        0.170        0.819
8        T        0.022        0.544        0.394        0.000        0.478        0.514        0.185        0.822
9        P        0.178        0.443        0.027        0.000        0.039        0.775        0.108        0.848
10        K        0.419        0.514        0.522        0.000        0.018        0.243        0.178        0.762
11        DE        0.217        0.591        0.361        0.000        0.548        0.457        0.172        0.826
12        DT        0.182        0.581        0.336        0.000        0.594        0.351        0.166        0.818
13        ILB        0.255        0.400        0.440        0.382        0.624        0.333        0.198        0.808
14        OLB        0.250        0.382        0.473        0.393        0.604        0.373        0.198        0.826
15        CB        0.247        0.570        0.111        0.498        0.419        0.559        0.203        0.833
16        S        0.277        0.518        0.101        0.469        0.575        0.476        0.188        0.817


gstelmack 01-06-2015 08:47 AM

Given those correlations, do I need to add the Adjusted Grade as a factor in Draft Analyzer?

Ben E Lou 01-06-2015 08:50 AM

I'm not sure. The correlation is always so high that I am not sure how useful it is especially when you factor in that it is often closely correlated with bars.


All times are GMT -5. The time now is 03:01 AM.

Powered by vBulletin Version 3.6.0
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.