2015 Combine Observations

[Telperion]

Dante Fowler and Randy Gregory can reach their (as fast as / faster than Abdullah) top speed very quickly (more quickly than Abdullah) too.

And they have 30-60lbs on him.

[dirtyharris]

DISCLAIMER: I am not trying to say Abdullah is the next great RB in the league, but...........................

Go do a Google search for "Emmitt Smith Combine Results"
I don't think anyone would would say that Emmitt was a real speed burner, but he did have a pretty good carreer none the less. He also ran behind one of the greatest O-Lines of his time, and had a really good team around him. What do you think Abdullah could do behind that line the Cowboys have today if Murray leaves and he goes to Dallas in the second?

While I am paying attention to combine results and overall performance at the college level, I believe I will put even more weight on where the second tier RB's land in the draft.

[ManuManu]

DISCLAIMER: I am not trying to say Abdullah is the next great RB in the league, but...........................

Go do a Google search for "Emmitt Smith Combine Results"
I don't think anyone would would say that Emmitt was a real speed burner, but he did have a pretty good carreer none the less. He also ran behind one of the greatest O-Lines of his time, and had a really good team around him. What do you think Abdullah could do behind that line the Cowboys have today if Murray leaves and he goes to Dallas in the second?

While I am paying attention to combine results and overall performance at the college level, I believe I will put even more weight on where the second tier RB's land in the draft.

Abdullah's issues will come down to whether he can hang onto the football. Speed really doesn't matter so much, and that's why the Speed Score formula is not a great tool. It doesn't take into account agility testing, which Abdullah is amazing at.

I think these are two good comps for Abdullah...

http://nflcombineresults.com/playerpage.php?i=6093
http://www.nfldraftscout.com/ratings/ds ... &genpos=RB

[TommyL31]

One of my favorites so far is this one published yesterday: Combine survivor effect basically stating that there's essentially a huge selection bias of people invited to the combine.

I like that article a lot - does a great job pointing out why combine results are not likely to be a great indicator. I am curious, has anyone done analysis on drills like this that includes players who did not get combine invitations but did work out at a pro day? It would maybe lessen this effect a little bit. There is still a pretty big selection bias in players working out at a pro day though.

It seems I'm on Bad Statistics Duty lately. This is a terrible, misleading article.

According to:
http://www.post-gazette.com/sports/stee ... 1409070235

"This year, there are 456 undrafted players on opening-week 53-man rosters, according to Elias, including 64 rookie free agents. Of those rookies, 36 weren't invited to the NFL Scouting Combine."

And just from another project, I have data on 304 guys from the combine for the 2014 class. I couldn't actually find a good measure of how many rookies make the final 53-man rosters (I Googled "how many rookies are in the nfl?" and found nothing useful on the first page, then gave up, my will exhausted), but given 32 teams that makes for 1696 total players. The draft lasts 7 rounds times 32 teams gives us 224 drafted guys, a number of whom I'm sure get cut before the season. Let's just go nuts and assume only half of drafted guys make final rosters, which seems underestimated to me given there are 64 undrafted rookies on rosters, but we'll run with it anyway.

That means we have 112 drafted and 36 undrafted non-combine-participants (UNCPs, obviously), meaning non-combine participant rookies make up... 36/(112+36)=.24. That suggests, even given my assumption that a full 1/3rd of 53-man-roster rookies aren't drafted, that we have combine data on 76% of incoming guys. I didn't look for how many drafted guys are not combine participants, but it would have to be a lot to throw off the results.

Hey guess what, if the author takes his arbitrary top-10% cutoff and change it to a top-76% cutoff, our data would fit pretty good.

Bottom line from my super-rough rebuttal: the reason we don't have a great way to use combine numbers to predict NFL performance isn't because of the survivor effect; it's because there are loads of omitted variables - that is, the combine doesn't measure too many factors that play into eventual success on the field.

Obviously I took some ugly shortcuts, but I still think it's illustrative of why that article ISN'T illustrative.

One of us is entirely misinterpreting the content of the article I posted. That arbitrary cutoff he's making isn't among the top 10% of combine participants or NFL players it's of AMATEUR ATHLETES. He's saying that you typically have to be really freaking good to even get invited to the combine.

I think the fact is that he's making a mathematically abstract point and not using real data. In essence though he's making a similar point that many of us (including Telperion) are hinting at. Correlation of combine data to success is pretty weak in general but there is a floor below which it's darn near impossible to be successful at the NFL level.

[AS3Dynasty]

DISCLAIMER: I am not trying to say Abdullah is the next great RB in the league, but...........................

Go do a Google search for "Emmitt Smith Combine Results"
I don't think anyone would would say that Emmitt was a real speed burner, but he did have a pretty good carreer none the less. He also ran behind one of the greatest O-Lines of his time, and had a really good team around him. What do you think Abdullah could do behind that line the Cowboys have today if Murray leaves and he goes to Dallas in the second?

While I am paying attention to combine results and overall performance at the college level, I believe I will put even more weight on where the second tier RB's land in the draft.

I think this is the best point to be made so far. If Abdullah lands in a favorable position, he likely jumps above Coleman, Ajayi, Duke, Yeldon, etc... if he goes to Dallas, he's probably the RB3 of this draft, slow 40 or not.

[jeffster]

I like that article a lot - does a great job pointing out why combine results are not likely to be a great indicator. I am curious, has anyone done analysis on drills like this that includes players who did not get combine invitations but did work out at a pro day? It would maybe lessen this effect a little bit. There is still a pretty big selection bias in players working out at a pro day though.

It seems I'm on Bad Statistics Duty lately. This is a terrible, misleading article.

According to:
http://www.post-gazette.com/sports/stee ... 1409070235

"This year, there are 456 undrafted players on opening-week 53-man rosters, according to Elias, including 64 rookie free agents. Of those rookies, 36 weren't invited to the NFL Scouting Combine."

And just from another project, I have data on 304 guys from the combine for the 2014 class. I couldn't actually find a good measure of how many rookies make the final 53-man rosters (I Googled "how many rookies are in the nfl?" and found nothing useful on the first page, then gave up, my will exhausted), but given 32 teams that makes for 1696 total players. The draft lasts 7 rounds times 32 teams gives us 224 drafted guys, a number of whom I'm sure get cut before the season. Let's just go nuts and assume only half of drafted guys make final rosters, which seems underestimated to me given there are 64 undrafted rookies on rosters, but we'll run with it anyway.

That means we have 112 drafted and 36 undrafted non-combine-participants (UNCPs, obviously), meaning non-combine participant rookies make up... 36/(112+36)=.24. That suggests, even given my assumption that a full 1/3rd of 53-man-roster rookies aren't drafted, that we have combine data on 76% of incoming guys. I didn't look for how many drafted guys are not combine participants, but it would have to be a lot to throw off the results.

Hey guess what, if the author takes his arbitrary top-10% cutoff and change it to a top-76% cutoff, our data would fit pretty good.

Bottom line from my super-rough rebuttal: the reason we don't have a great way to use combine numbers to predict NFL performance isn't because of the survivor effect; it's because there are loads of omitted variables - that is, the combine doesn't measure too many factors that play into eventual success on the field.

Obviously I took some ugly shortcuts, but I still think it's illustrative of why that article ISN'T illustrative.

One of us is entirely misinterpreting the content of the article I posted. That arbitrary cutoff he's making isn't among the top 10% of combine participants or NFL players it's of AMATEUR ATHLETES. He's saying that you typically have to be really freaking good to even get invited to the combine.

I think the fact is that he's making a mathematically abstract point and not using real data. In essence though he's making a similar point that many of us (including Telperion) are hinting at. Correlation of combine data to success is pretty weak in general but there is a floor below which it's darn near impossible to be successful at the NFL level.

Sorry, that's not right. The population of guys that we're looking at here is not "all amateur athletes." I'll quote from the article:

"What's going on is the survivor effect. Measurable aspects of performance such as speed, strength, and agility do matter greatly. But the statistical comparisons of the numbers are misleading because we can only observe a small subset of the overall population of players. In other words, instead of comparing the athletic potential of all the amateur players, we are comparing the potential of players given they were invited to the combine and given that they were chosen in the draft."

Even the article doesn't try to say what you're describing - he's just referring to all potential incoming NFL players as "amateur players" - because they're coming from college.

He's also not making the point that combine data isn't well correlated to success. He's making the point that combine data isn't well correlated to success because of survivor bias. And that's wrong. The article says "we can only observe a small subset of the overall population of players" and my point is that, no, we observe a very large subset of the overall population of players. But we still can't correlate combine data to success very well because there are too many omitted variables.

In other words, the author tried to use "fancy" (I use the term loosely) statistics to make a simple point, and while the simple point is valid his fancy statistics are wrong.

[Billy McFred]

Listed some of the "proven" backs (2+ years) with their Adjusted 40 time percentile, followed by their agility percentile.

10 out of 13 had an adjusted 40 time percentile of 50th or higher
6 out of 13 (7 if you count CJ2K) had an agility score of 50th percentile or higher

2 of 13 had a 40 time of 25th percentile or less
5 of 13 had an agility score of 25th percentile or less


------------40 time -- Agility
Foster ----- 16th - 12th
A. Morris -- 25th, 62nd
McCoy ----- 44th, 84th
Lacy ------- 62nd - 6th
Bell -------- 65th, 85th
S. Jackson - 73rd, 70th
Charles ---- 84th, 81st
Lynch ------ 87th, 10th
Forte ------ 88th, 75th
MJD -------- 91st, 22nd
Murray ----- 93rd, 29th
Peterson -- 95th, 22nd
CJ2K ------- 97th, ???

Average ---- 71st, 47th

Did i really look all this up for nothing? Basically the only "Abdullah" like player on this list is McCoy. So it's possible he can succeed, but not likely.

[jeffster]

Or here, I'll try a different way:

If survivor bias explains why combine data doesn't correlate well to NFL success, then the solution would be to get combine data on all players. Then it would correlate well to success. But no, it wouldn't, because we're missing lots of key factors that affect success, of which simple combine data are at most a small part of.

[kmbryant09]

Any chance we can talk about someone besides Abdullah?

How is Kevin White as a route runner?
Did Gordon hurt/help his stock?
Who do you prefer - Winston or Mariota?

[Telperion]

Did i really look all this up for nothing? Basically the only "Abdullah" like player on this list is McCoy. So it's possible he can succeed, but not likely.

It supports my point, yes, it's good info. I'd say though that McCoy ran a 4.5, and if Abdullah had run that we wouldn't even be having this discussion on here.

A 4.5 is (in my mind and players like McCoy verify it) fast enough for a small statured back to succeed when they excel in other areas. 4.6+ though is (in my mind and by my estimation externally validated) simply too slow for other areas to overcome.

Emmitt Smith ran a 4.55 on his pro day. But when the question becomes "is Abdullah as a total package the caliber of the NFL's all time leading rusher", I'll bet against that 10/10. Bryan Westbrook is a good comparison, and the only real comp I've seen so far.

[Cameron Giles]

Any chance we can talk about someone besides Abdullah?

How is Kevin White as a route runner?
Did Gordon hurt/help his stock?
Who do you prefer - Winston or Mariota?

White is a much improved route runner.

Gordon is still a Top 5 pick in dynasty.

There's a big gap between Winston and Mariota. Winston is a much better QB at this point and can make all the throws. Mariota is a project.

[TommyL31]

One of us is entirely misinterpreting the content of the article I posted. That arbitrary cutoff he's making isn't among the top 10% of combine participants or NFL players it's of AMATEUR ATHLETES. He's saying that you typically have to be really freaking good to even get invited to the combine.

I think the fact is that he's making a mathematically abstract point and not using real data. In essence though he's making a similar point that many of us (including Telperion) are hinting at. Correlation of combine data to success is pretty weak in general but there is a floor below which it's darn near impossible to be successful at the NFL level.

Sorry, that's not right. The population of guys that we're looking at here is not "all amateur athletes." I'll quote from the article:

"What's going on is the survivor effect. Measurable aspects of performance such as speed, strength, and agility do matter greatly. But the statistical comparisons of the numbers are misleading because we can only observe a small subset of the overall population of players. In other words, instead of comparing the athletic potential of all the amateur players, we are comparing the potential of players given they were invited to the combine and given that they were chosen in the draft."

Even the article doesn't try to say what you're describing - he's just referring to all potential incoming NFL players as "amateur players" - because they're coming from college.

He's also not making the point that combine data isn't well correlated to success. He's making the point that combine data isn't well correlated to success because of survivor bias. And that's wrong. The article says "we can only observe a small subset of the overall population of players" and my point is that, no, we observe a very large subset of the overall population of players. But we still can't correlate combine data to success very well because there are too many omitted variables.

In other words, the author tried to use "fancy" (I use the term loosely) statistics to make a simple point, and while the simple point is valid his fancy statistics are wrong.

How is that not implying that he's not talking about all amateur players not just ones who will play in the NFL?

I'm not even going to argue that his point is demonstrably true because the magnitude of such an effect is completely unknown.

To answer your question about why we don't test all players at the combine to get a more strong correlation, it's because there are certain thresholds that can be seen with the naked eye that make it a waste of time to invite most NCAA players to the combine. Do you really need to see fringe college players who run a 5.5 40 to know that speed has a strong correlation to success in the NFL? Nobody does, but is it possible that we tend to overblow the different between a 4.5 and 4.55 because, as you said, there are plenty of other factors that affect success.

[Billy McFred]

By the time we are doing our rookie drafts, Winston is going to be hyped out of this world. He's already being compared to Peyton Manning, and he will end up in a great situation throwing to 2 potential young studs in Mike Evans and ASJ.

[TommyL31]

Any chance we can talk about someone besides Abdullah?

How is Kevin White as a route runner?
Did Gordon hurt/help his stock?
Who do you prefer - Winston or Mariota?

White is a much improved route runner.

Gordon is still a Top 5 pick in dynasty.

There's a big gap between Winston and Mariota. Winston is a much better QB at this point and can make all the throws. Mariota is a project.

I think Gordon helped his stock among RBs because so many of them under performed and Gurley couldn't participate. The fact that he 'held serve' makes him look like an even better option. He is still a top 4 pick in dynasty but he may have lost a little because people loved Kevin White's performance so much. Before the combine I think it was Gurley/Gordon/Cooper in some order for most people and I think it's now Gurley/Gordon/Cooper/White in some order for most.

One guy who I'm not sure what to make of his draft stock now is Tevin Coleman. Normally a guy missing the combine is a bad thing because people forget about him but pre-combine I think the rookie tiers were something like:

Gurley/Gordon
Ajayi/Coleman/Duke Johnson/Yeldon/Abdullah

And I don't think Coleman was very strong in that tier. But outside of Ajayi, I don't think anyone was impressed with the other tier 2 RBs performances so Coleman might have moved up just by default.

[ManuManu]

Did i really look all this up for nothing? Basically the only "Abdullah" like player on this list is McCoy. So it's possible he can succeed, but not likely.

It supports my point, yes, it's good info. I'd say though that McCoy ran a 4.5, and if Abdullah had run that we wouldn't even be having this discussion on here.

A 4.5 is (in my mind and players like McCoy verify it) fast enough for a small statured back to succeed when they excel in other areas. 4.6+ though is (in my mind and by my estimation externally validated) simply too slow for other areas to overcome.

Emmitt Smith ran a 4.55 on his pro day. But when the question becomes "is Abdullah as a total package the caliber of the NFL's all time leading rusher", I'll bet against that 10/10. Bryan Westbrook is a good comparison, and the only real comp I've seen so far.

McCoy ran a 4.5 during a pro day, which players notoriously run faster at. He gets caught from behind A LOT, so I'd say there's little chance he's a true 4.5 guy. More like a 4.6. The point is, you don't need long speed to be a good running back. For him, it's his rare agility that makes the difference. Abdullah has that kind of agility.

I'm not saying Abdullah is a lock to be a good running back. After all, he's on the small side, doesn't break a ton of tackles, fumbles, etc. But he has plenty of burst to get through holes and he has the agility to make defenders look silly, much like Brian Westbrook.