NFL Draft Scout Times All Changed for 2012 (1 Viewer)

[Dinsy Ejotuz]

Just checked the WRs, RBs and QBs from 2012, and virtually every single player is now listed considerably faster than the time that was entered after the combine last year.And as a class they're much faster than any previous class -- the majority of the WR and RB classes are now listed with times that would have been size/speed elite in any previous year. It's like someone went through and shaved a tenth of a second off almost every time -- sub 4.40 is the new normal for the WRs. Anyone have any idea what's going on over there? New way of measuring? Something else?

 

[BlackMamba24]

new way of measuring i believe --- electronic timing method

 

[Dinsy Ejotuz]

new way of measuring i believe --- electronic timing method

If you don't mind me asking... where'd you hear that? Not trying to be an ###, just want to confirm that's the difference.Without doing an actual comparison yet it seems the changes are pretty consistent changes at least. Mostly .06 to .10 faster compared to the original measure.

 

[Dinsy Ejotuz]

Looks like it's true for the TEs as well. Egnew (-.10), Hanna (-.06), and L Green (-.08) all faster as well by about the same amount.

 

[ZWK]

The changes appear to be across-the-board; they apply to offensive linemen too (Kalil -.03, Reiff -.07, Glenn -.07, etc.).The thing to do, presumably, is to find the average change (across all players at all positions) and add that to everyone's time if you want to compare to previous years (or apply formulas based on previous years).There are also some changes in relative position. David Wilson had been tied with Bernard Pierce at 4.49, but the new numbers have Wilson at 4.38 and Pierce at 4.45. I assume that the new relative numbers are more accurate than the old ones.

 

[Dinsy Ejotuz]

The changes appear to be across-the-board; they apply to offensive linemen too (Kalil -.03, Reiff -.07, Glenn -.07, etc.).The thing to do, presumably, is to find the average change (across all players at all positions) and add that to everyone's time if you want to compare to previous years (or apply formulas based on previous years).There are also some changes in relative position. David Wilson had been tied with Bernard Pierce at 4.49, but the new numbers have Wilson at 4.38 and Pierce at 4.45. I assume that the new relative numbers are more accurate than the old ones.

Thanks for looking at some other positions. I'm guessing the new #s are more accurate relative to each other as well.Might try the folks at NFLDS directly once the draft is over and they come up for air. Am curious if they have any insight that's better than just using the average change to create comparables to previous years.

 

[Xue]

I have no Idea what's going on, but I usually accept the slower times or somewhere in the middle. These times just don't match up.

 

[BlackMamba24]

new way of measuring i believe --- electronic timing method

If you don't mind me asking... where'd you hear that? Not trying to be an ###, just want to confirm that's the difference.Without doing an actual comparison yet it seems the changes are pretty consistent changes at least. Mostly .06 to .10 faster compared to the original measure.

After further review it appears they didn't actually change anything, but there was an "experiment" that used fully automated timing at last years combine. But the times weren't revealed to the teams and players. Apparently they were expecting the FAT times to actually be .20-.24 slower though. So that doesn't explain whats going on. I must have misunderstood or heard incorrectly.

http://www.cbssports.com/nfl/draft/story/17408448/new-timing-method-in-40-could-lead-to-culture-shock-for-players

 

[Xue]

NFL Network can simply just use video to determine accurate times. HD video at 60 FPS is sufficient. How can anyone put faith into a handheld stopwatch? These people aren't as smart as you think they are.

 

[Whitney26]

If i had to guess, I'd say they've simply sat on the official numbers from whatever source they use. Or, they just around to getting it updated.

FWIW, I updated them in my spreadsheet for easy perusing:

https://docs.google.com/spreadsheet/pub?key=0AotIVo4gpWF8dEtPV0x3UHNnM2hlRUxYbTBoekJ2YUE&gid=0

 

[Dinsy Ejotuz]

If i had to guess, I'd say they've simply sat on the official numbers from whatever source they use. Or, they just around to getting it updated.

I think that's part of it. They hadn't listed the 10-yard splits until the update.But the times are all faster, where in normal years the adjustments have varied up and down. It's something systematic.

 

[ZWK]

The average change looks to be about .06 or .065 seconds.

I'd made a note last year of some WRs whose 40 times seemed anomalously slow, given how well they did on the other drills. I took a look at that today - did the WRs who had better numbers on their other drills than they did on their 40 get the largest upgrades to their 40 time? In short, the answer is yes. There was a correlation of r=0.71, n=39, which is much larger than I expected. Graph here.

That's enough to convince me that the new 40 times are better calibrated than the old ones. Throw out the old ones, use the new ones, with an adjustment of +.06 or so when comparing them with previous years (or applying formulas based on previous years) since they are significantly faster than previous years.

Methodological details: For all 39 WRs who had both a 40 time and at least one jumping drill at the 2012 combine, I calculated their drill mismatch by standardizing their original 40 time, and subtracting that from their standardized jumping score (which combines standardized vertical & standardized broad jump). Then I calculated the change in their 40 time - their original time minus their new time. Those are the two variables that I plotted on the graph, and they correlate with each other at r=0.71.

 

[Dinsy Ejotuz]

Nice work, Z. I did something like that with the 2008 vertical data.

Quick question though. Why would use an adjustment of .06 when comparing them to past years if the 'new' numbers are more in line with the jumps? Wouldn't you just adopt the new numbers?

Also, did you do the same thing for RBs? Those look really off to me given what else we know about those players.

 

[Rhythmdoctor]

Nice work, Z. I did something like that with the 2008 vertical data.Quick question though. Why would use an adjustment of .06 when comparing them to past years if the 'new' numbers are more in line with the jumps? Wouldn't you just adopt the new numbers?Also, did you do the same thing for RBs? Those look really off to me given what else we know about those players.

I think he's saying to adjust the times of past players to put everyone on an equal plane. Since the new technology wasnt used for past metrics the adjustment needs to be made universally.

 

[Dinsy Ejotuz]

Nice work, Z. I did something like that with the 2008 vertical data.Quick question though. Why would use an adjustment of .06 when comparing them to past years if the 'new' numbers are more in line with the jumps? Wouldn't you just adopt the new numbers?Also, did you do the same thing for RBs? Those look really off to me given what else we know about those players.

I think he's saying to adjust the times of past players to put everyone on an equal plane. Since the new technology wasnt used for past metrics the adjustment needs to be made universally.

Ahh... makes sense. I think I missed something though. What new technology? Didn't they decide not to use the all-electronic stuff?

 

[ZWK]

Nice work, Z. I did something like that with the 2008 vertical data.Quick question though. Why would use an adjustment of .06 when comparing them to past years if the 'new' numbers are more in line with the jumps? Wouldn't you just adopt the new numbers?Also, did you do the same thing for RBs? Those look really off to me given what else we know about those players.

I think he's saying to adjust the times of past players to put everyone on an equal plane. Since the new technology wasnt used for past metrics the adjustment needs to be made universally.

I actually haven't done this rigorously; I probably should. Maybe tonight.But basically, you'd expect the average 40 time of the 2012 draft class to be similar to the average 40 time of the previous few draft classes, across all positions. Occasionally there's a draft class at one position that's unusually fast, like the 2008 RBs or the 2009 WRs, but if every position is unusually fast then that's a sign that the combine changed how they're measuring. To get slightly fancier, you can look at the other drills (vert, broad, cone, shuttle) to see how the 2012 draft class compares to other draft classes (on average). If 2012 is slightly more athletic on those drills then you'd expect them to also be slightly faster on the 40, and you can project by how much and use that to estimate more exactly how far the 40 times should be shifted.The results in my previous post should carry over to all positions - the newer times are better calibrated (meaning better at showing the relative speeds of the guys in the 2012 draft class), but if there is that 0.06 second shift it would apply to everyone.

 

[Dinsy Ejotuz]

Nice work, Z. I did something like that with the 2008 vertical data.Quick question though. Why would use an adjustment of .06 when comparing them to past years if the 'new' numbers are more in line with the jumps? Wouldn't you just adopt the new numbers?Also, did you do the same thing for RBs? Those look really off to me given what else we know about those players.

I think he's saying to adjust the times of past players to put everyone on an equal plane. Since the new technology wasnt used for past metrics the adjustment needs to be made universally.

I actually haven't done this rigorously; I probably should. Maybe tonight.But basically, you'd expect the average 40 time of the 2012 draft class to be similar to the average 40 time of the previous few draft classes, across all positions. Occasionally there's a draft class at one position that's unusually fast, like the 2008 RBs or the 2009 WRs, but if every position is unusually fast then that's a sign that the combine changed how they're measuring. To get slightly fancier, you can look at the other drills (vert, broad, cone, shuttle) to see how the 2012 draft class compares to other draft classes (on average). If 2012 is slightly more athletic on those drills then you'd expect them to also be slightly faster on the 40, and you can project by how much and use that to estimate more exactly how far the 40 times should be shifted.The results in my previous post should carry over to all positions - the newer times are better calibrated (meaning better at showing the relative speeds of the guys in the 2012 draft class), but if there is that 0.06 second shift it would apply to everyone.

So the new times fit within the class better, but to compare them to past classes you'd want to subtract (.06) from 2012 listed times?WAG that the revisions reflected two things... One is the normal revisions they do after Indy. Usually there's a round of revisions a few weeks later, but for some reason there was no update last year.Two is a change in the method of timing which resulted in the times being, on average, .065 seconds faster.So instead of adding +.05 to some of the times and subtracting (-.05) from others during the normal revision process, we're seeing the revisions that are between +.01 seconds (-.05+.06) to +.11 seconds (+.05+.06) faster.

 

[Sarnoff]

NFL Network can simply just use video to determine accurate times. HD video at 60 FPS is sufficient. How can anyone put faith into a handheld stopwatch? These people aren't as smart as you think they are.

60FPS isn't really that good for this. At best, you'll be able to differentiate 1/60th of a second. Modern timing methods are accurate to 1/100th. I read an article explaining why the combine doesn't use this method in more detail recently but can't seem to find it. I mean, it'll get you close, but not exact. Sure, more exact than a handheld stopwatch, but that's not what the agents and players want.If they have to delve into fields/interlaced frames, it'll confuse it even more, anyway. What they'd really need is a version of slit-scan photography like is used for photo finishes in horse races.

 

[cstu]

Google is putting computers into glasses and the NFL can't even measure a 40 yard dash.

 

[Xue]

NFL Network can simply just use video to determine accurate times. HD video at 60 FPS is sufficient. How can anyone put faith into a handheld stopwatch? These people aren't as smart as you think they are.

60FPS isn't really that good for this. At best, you'll be able to differentiate 1/60th of a second. Modern timing methods are accurate to 1/100th. I read an article explaining why the combine doesn't use this method in more detail recently but can't seem to find it. I mean, it'll get you close, but not exact. Sure, more exact than a handheld stopwatch, but that's not what the agents and players want.If they have to delve into fields/interlaced frames, it'll confuse it even more, anyway. What they'd really need is a version of slit-scan photography like is used for photo finishes in horse races.

60 FPS is good enough for timing the 40, at least for me. At 30 FPS, if you don't a get frame where the runner crosses the plane of the markers exactly, you get a frame of him either right before or right after.I used to analyze hitting/pitching mechanics and what I noticed is at 30 FPS, if you don't get a frame where the bat contacts the ball, you get one right before or right after. But at 60 FPS, you always get the "contact frame". With pitchers, at 60 FPS, you always get the frame where the ball last touches the hand.

 

[ZWK]

Nice work, Z. I did something like that with the 2008 vertical data.

Quick question though. Why would use an adjustment of .06 when comparing them to past years if the 'new' numbers are more in line with the jumps? Wouldn't you just adopt the new numbers?

Also, did you do the same thing for RBs? Those look really off to me given what else we know about those players.

I think he's saying to adjust the times of past players to put everyone on an equal plane. Since the new technology wasnt used for past metrics the adjustment needs to be made universally.

I actually haven't done this rigorously; I probably should. Maybe tonight.But basically, you'd expect the average 40 time of the 2012 draft class to be similar to the average 40 time of the previous few draft classes, across all positions. Occasionally there's a draft class at one position that's unusually fast, like the 2008 RBs or the 2009 WRs, but if every position is unusually fast then that's a sign that the combine changed how they're measuring. To get slightly fancier, you can look at the other drills (vert, broad, cone, shuttle) to see how the 2012 draft class compares to other draft classes (on average). If 2012 is slightly more athletic on those drills then you'd expect them to also be slightly faster on the 40, and you can project by how much and use that to estimate more exactly how far the 40 times should be shifted.

The results in my previous post should carry over to all positions - the newer times are better calibrated (meaning better at showing the relative speeds of the guys in the 2012 draft class), but if there is that 0.06 second shift it would apply to everyone.

So the new times fit within the class better, but to compare them to past classes you'd want to subtract (.06) from 2012 listed times?WAG that the revisions reflected two things...

One is the normal revisions they do after Indy. Usually there's a round of revisions a few weeks later, but for some reason there was no update last year.

Two is a change in the method of timing which resulted in the times being, on average, .065 seconds faster.

So instead of adding +.05 to some of the times and subtracting (-.05) from others during the normal revision process, we're seeing the revisions that are between +.01 seconds (-.05+.06) to +.11 seconds (+.05+.06) faster.

Your WAG sounds about right.After taking a look at the complete data (from nflcombineresults, which seems to match the new numbers on nfldraftscout), it looks like the adjustment should be about .03 or .04 rather than .06. Comparing all players who ran the 40 in 2012 with all the players who ran the 40 in 2009-2011 & 2013, the 2012 players were 3.3 hundredths of a second faster on average (according to the new times). Breaking it down by position, the gap is 4.3 hundredths if I give each position equal weight (out of the 15 positions), 4.0 hundredths if I take the median position instead of the mean, 3.9 hundredths if I only include the (11) positions where there were at least 10 players in 2012, and 2.1 hundredths if I only include the (9) positions with at least 15 players.

Compared to the other years (2009-2013), 2012 was slightly better at the vertical & shuttle, but slightly worse at the broad & cone, which suggests that they were pretty typical in terms of overall athleticism. Without adjustment, 2012 comes out as the fastest year on record (going back to 1999). So this straight adjustment, which assumes that the different years would be equally fast if not for measurement bias, seems appropriate.

 

[Dinsy Ejotuz]

WR times have been updated for 2013. Initial impression is that these are also 'fast' compared to 2011 and earlier. Does that seem right ZWK?

ETA: same deal as last year -- adjustments are almost all one way, faster. Betting we find that they're more similar to the 2012 times, and not those from earlier years.

EETA: With two years of data we might also be able to compare the faster speeds just within each position. I'm not convinced that applying a blanket readjustment is the way to go. Might be better to do one each for QBs, RBs, WRs and TEs if they seem different.

EEETA: Holy crap. Patterson now listed with 4.33 and Austin is at 4.28.

EEEETA: Didn't notice this before, but I think whatever the change they made mostly affects the 40 times, and not the intermediate splits. They seem out of whack now.

 

[Xue]

Wow, not sure why or what they're doing but these times aren't reliable anymore. The splits also don't match up to what I've measured with video.

Ryan Swope ran a 4.28? LOL.

 

[ZWK]

Xue, is your list of 40 times published online somewhere? That's one test we could run on the updates - if the new times are more strongly correlated with your times than the old times were, that's a sign that the update is an improvement (at least in terms of calibration).

I just checked Franklin's page, and it looks like the RBs have been updated as well. Before I look through all the numbers, here are three hypotheses about whose times will see the biggest improvements:

H1: Players who did better at the jumps than on their originally reported 40 time

H2: Players who did better on their 10-yard split than on their last 30 yards

H3: Players who did better on their last 30 yards than on their 10-yard split

Here are the lists of RBs who those would apply to:

H1: Better at the jumps than the 40

(higher numbers indicate better jumps than 40, negative numbers indicate better 40 than jumps)

3.0 Rex Burkhead

2.3 Christine Michael

1.9 George Winn

1.8 Robbie Rouse

1.6 Michael Ford

1.4 Ray Graham

1.2 Cierre Wood

1.1 Joseph Randle

1.0 Andre Ellington

0.7 Montee Ball

0.7 Stefphon Jefferson

0.4 Kenjon Barner

0.4 Stepfan Taylor

0.4 Zac Stacy

0.4 Montel Harris

0.4 C.J. Anderson

0.3 Giovani Bernard

0.3 Matthew Tucker

0.2 Le'Veon Bell

0.0 D.J. Harper

-0.2 Mike James

-0.3 Mike Gillislee

-0.4 Jawan Jamison

-0.5 Kerwynn Williams

-1.1 Johnathan Franklin

-1.2 Knile Davis

-1.4 Onterio McCalebb

H2 & H3: Inconsistency between 10-yard split and last 30 yards

(higher numbers indicate that they were better at the 10-yard split)

1.9 Christine Michael

1.4 Rex Burkhead

1.3 Ray Graham

0.9 Cierre Wood

0.8 Michael Ford

0.7 Onterio McCalebb

0.6 Andre Ellington

0.5 Giovani Bernard

0.5 Montel Harris

0.5 Matthew Tucker

0.4 Robbie Rouse

0.2 Kerwynn Williams

0.2 Montee Ball

0.1 Le'Veon Bell

0.1 C.J. Anderson

0.1 D.J. Harper

-0.1 Stepfan Taylor

-0.1 Stefphon Jefferson

-0.2 Knile Davis

-0.3 Johnathan Franklin

-0.6 George Winn

-0.8 Joseph Randle

-1.0 Mike James

-1.1 Zac Stacy

-1.1 Mike Gillislee

-1.7 Jawan Jamison

-1.8 Kenjon Barner

 

[Xue]

Sorry don't have them published anywhere.

 

[ConnSKINS26]

So until we hear word on their methodology, or at least reasoning behind the changes, their numbers are useless, right?

 

[ZWK]

Here are the RBs, sorted by how much their 40 time improved with the update:

new old change player4.39 4.52 0.13 Kenjon Barner4.68 4.80 0.12 Ray Graham4.43 4.54 0.11 Christine Michael4.44 4.55 0.11 Matthew Tucker4.57 4.68 0.11 Montel Harris4.46 4.56 0.10 Cierre Wood4.65 4.75 0.10 George Winn4.42 4.50 0.08 Michael Ford4.65 4.73 0.08 Rex Burkhead4.59 4.66 0.07 Montee Ball4.53 4.60 0.07 C.J. Anderson4.48 4.55 0.07 Mike Gillislee4.46 4.52 0.06 D.J. Harper4.62 4.68 0.06 Stefphon Jefferson4.62 4.68 0.06 Jawan Jamison4.28 4.34 0.06 Onterio McCalebb4.70 4.76 0.06 Stepfan Taylor4.75 4.80 0.05 Robbie Rouse4.44 4.48 0.04 Kerwynn Williams4.59 4.63 0.04 Joseph Randle4.56 4.60 0.04 LeVeon Bell4.50 4.53 0.03 Giovani Bernard4.46 4.49 0.03 Johnathan Franklin4.50 4.53 0.03 Mike James4.35 4.37 0.02 Knile Davis4.53 4.55 0.02 Zac Stacy4.66 4.68 0.02 Theo Riddick4.61 4.61 0.00 Andre EllingtonAverage change is about .06, similar to last year (where I calculated .06 for the change and about .04 for the disparity with previous years - perhaps the original times were .02 too slow?) I suspect that the new times are better calibrated than the old times, in terms of the relative speeds. Christine Michael strengthens his case to be the #2 RB behind Lacy, Kenjon Barner has my attention again, and I should probably take a closer look at Cierre Wood, Matthew Tucker, and CJ Anderson.
Testing hypothesis 1, there is a r=0.38 correlation between jump-40 mismatch (with their old 40 time) and change in 40 time. So, a modest effect in the predicted direction (not nearly as big as the 0.71 correlation I found with the 2012 WRs using the same analysis).

There are actually several ways to combine hypotheses 2 & 3. One possibility is that those who were better on the 10 than the 30 would see the biggest improvements and those who were better on the 30 than the 10 would see the smallest improvements. A second possibility is the exact opposite - biggest improvements for those with good 30 & bad 10. A third possibility is that those who had the biggest 10-30 mismatch would see the biggest improvements, regardless of whether they were better at the 10 or the 30 (and those whose 10 & 30 times were most consistent with each other would see the smallest improvement). Possibility 3 is one that I've considered using in my RB ratings. A fourth possibility is that those whose 10 & 30 times were most consistent with each other would see the most average improvement (closest to 0.04), while those who had a large mismatch between their 10 & 30 (in either direction) would be prone to both unusually large improvements and unusual small improvements (the idea here is that a 10-30 mismatch is a sign of a large measurement error, which could be in either direction).

Testing the first two of these possibilities, I just look at the correlation between 10-30 mismatch (the number that I gave in my previous post) and change in 40 time. The data is slightly in favor of possibility 1, r=0.22, which means that it goes against possibility 2 (which is the exact opposite). For possibility 3, I take the absolute value of the number that I gave (10-30 mismatch) - did those with a large mismatch in either direction see more improvement in 40 time? The answer is yes, somewhat more strongly: r=0.42 for the correlation between the absolute value of the 10-30 mismatch and the improvement in 40 time. To test possibility 4, I also re-scale the change in 40 time (subtracting .06 so that it's centered around zero) and take the absolute value; this correlation is similar in size, r=0.38 for the correlation between the absolute value of the 10-30 mismatch and the absolute value of the (rescaled) change in 40 time.

These correlations provide some evidence that the new numbers are better calibrated than the old, since they're what I'd expect to find if the old 40 numbers had predictable errors.

 

[Dinsy Ejotuz]

So in order to compare last year and this year to 2011 and earlier, you'd recommend using the listed times and adding .04 seconds?

 

[ZWK]

So in order to compare last year and this year to 2011 and earlier, you'd recommend using the listed times and adding .04 seconds?

Yes for 2012. For now, that is probably also the best thing to do for this year, but I'd like to look at the full set of this year's new 40 times before I settle on something for sure. nflcombineresults is still showing the old times, so I don't have those numbers in readily accessible form.

 

[Dinsy Ejotuz]

Great work on all this Z!

When you have the 2013 #s, could you take a quick look at the adjustments by position? I'm wondering if the RBs and WRs (faster) should be treated the same as the QBs, TEs and linemen (slower)?

IOW should .04 (or whatever) be a blanket fix or would an individual adjustment for each position (or for each range of 40 times) make more sense.

 

[ZWK]

Great work on all this Z!When you have the 2013 #s, could you take a quick look at the adjustments by position? I'm wondering if the RBs and WRs (faster) should be treated the same as the QBs, TEs and linemen (slower)? IOW should .04 (or whatever) be a blanket fix or would an individual adjustment for each position (or for each range of 40 times) make more sense.

I'd expect the fix to be essentially constant across positions. The one thing that it might vary by is the players' speed, but since it's presumably coming from measurement error/technique at the start and the finish it shouldn't matter how long it takes the players to run in between. And in 2012, there was no correlation between how fast a position was and how much its 40 times differed from other years. Just apparently random variation centered at .04 or so.

 

[Dinsy Ejotuz]

Awesome ... thanks for taking a look. Much appreciated!

 

[Xue]

I've decided to publish "my" times. Will include Pro Day times as well. If I can't find video on a player, they won't be listed.

https://docs.google.com/spreadsheet/ccc?key=0Aph2QMe0NEl5dE1ud05CZUVkYk9WdUd6LXhweHRkVkE#gid=0

Will update as I go.

 

[ZWK]

I've decided to publish "my" times. Will include Pro Day times as well. If I can't find video on a player, they won't be listed.

https://docs.google.com/spreadsheet/ccc?key=0Aph2QMe0NEl5dE1ud05CZUVkYk9WdUd6LXhweHRkVkE#gid=0

Will update as I go.

After taking a look at the complete data (from nflcombineresults, which seems to match the new numbers on nfldraftscout), it looks like the adjustment should be about .03 or .04 rather than .06. Comparing all players who ran the 40 in 2012 with all the players who ran the 40 in 2009-2011 & 2013, the 2012 players were 3.3 hundredths of a second faster on average (according to the new times). Breaking it down by position, the gap is 4.3 hundredths if I give each position equal weight (out of the 15 positions), 4.0 hundredths if I take the median position instead of the mean, 3.9 hundredths if I only include the (11) positions where there were at least 10 players in 2012, and 2.1 hundredths if I only include the (9) positions with at least 15 players.

I was counting 2013 in my baseline here, which was a mistake. Using only 2009-2011 as a baseline, the 2012 players were only 2.3 hundredths of a second faster, on average (using the new nflds numbers). Using 2006-2011 as the baseline, it's only 1.8 hundredths. Which suggests that the adjustment should only be about .02 rather than .04. The gap is small enough to maybe just be normal year-to-year variation (although it was still the fastest time on record, after several years of stagnation, in a year when performance on other drills did not stand out).The change that nflds made to their 2012 40 times made them about 7 hundredths of a second faster, on average. Here is a graph showing the average 40 time across all positions, from 1999-2013, which shows both the pre-adjustment (blue) and post-adjustment (red) numbers. You can see that the pre-adjustment numbers are much more out-of-line with the previous years than the post-adjustment numbers. (Average 40 times have been relatively consistent since 2006, although they were significantly higher from 1999-2004.)

The 2013 pre-adjustment numbers were way faster than the 2012 pre-adjustment numbers, so if the latest change brings another .07 second improvement then the 2013 post-adjustment numbers will be really out of whack.

I'll take a look once I get that data set. My current plan is to wait until nflcombineresults updates their numbers to match the new nfldraftscout numbers, and then download them and run some analyses. If anyone has an alternative which involves less waiting, and not too much more work, I'm open to suggestions.

 

[Dinsy Ejotuz]

I've decided to publish "my" times. Will include Pro Day times as well. If I can't find video on a player, they won't be listed.https://docs.google.com/spreadsheet/ccc?key=0Aph2QMe0NEl5dE1ud05CZUVkYk9WdUd6LXhweHRkVkE#gid=0Will update as I go.

Excellent Xue.Your times are pretty close to dead on for most players compared the originally posted times -- before they were adjusted.

 

[ZWK]

I've decided to publish "my" times. Will include Pro Day times as well. If I can't find video on a player, they won't be listed.https://docs.google.com/spreadsheet/ccc?key=0Aph2QMe0NEl5dE1ud05CZUVkYk9WdUd6LXhweHRkVkE#gid=0Will update as I go.

For the 12 players that you have so far, the correlation between Xue's time and the original nfldraftscout time is r=.937. The correlation between Xue's time and the updated nfldraftscout time is r=.938. If those numbers continue to be identical when the sample size is larger, that suggests that the old numbers and new numbers are equally calibrated; they each have about the same amount of noise. I expected the correlation with the new nflds numbers to be larger than the correlation with the old nflds numbers, based on the assumption that the new numbers are more accurate (in terms of the relative times of different players). (The correlation between the original and updated nfldraftscout times for those 12 players is r=.968, and if you average the old and new numbers together they correlate with Xue's numbers at r=.945)The average 40 time (for those 12 players) is:4.559 Xue4.578 original nflds4.524 updated nflds

 

[Dinsy Ejotuz]

Check out the WRs tab too ZWK -- he's started filling some of those in as well.

 

[ZWK]

Check out the WRs tab too ZWK -- he's started filling some of those in as well.

Rights, tabs. With n=21 (12 RB, 9 WR):Correlations

.952 Xue & Old nflds time

.960 Xue & New nflds time

.962 Xue & Average of old and new nflds time

.975 Old & New nflds times

Suggesting that the new nflds are a bit better calibrated than the old ones (since .960 > .952), but you might be able to do slightly better (reduce the noise in the measurements) with some kind of weighted average of the old and new (or even an unweighted average, since .962 > .960).

Averages

4.517 Old nflds times

4.458 New nflds times

4.502 Xue

We don't have any data on how Xue's method compares to previous years of nflds times, so this comparison tells us very little about what blanket adjustment to use (if any).

 

[Whitney26]

NFLDS have updated the combine 40s and the 10s & 20s. They are different than what was originally reported. For instance, Goodwin is credited with a 4.21.

 

[Xue]

NFL.com/40 has video of past combine participants. I'm going to go over those and make a separate spreadsheet. There's also some footage floating on youtube as well that I will be using.

One little secret everyone needs to know is that Chris Johnson didn't run a 4.24, but he does have the fastest 10 and 20 splits.

 

[ZWK]

With four TEs added (n=25 total), the correlations are:

0.958 Xue & Old nflds time

0.969 Xue & New nflds time

So it's looking like the nfdls update was an improvement in terms of calibration.

I'm tracking these numbers in a google spreadsheet, which I plan on updating if Xue adds more times to his 2013 spreadsheet.

 

[Xue]

I thought it was crazy for combine RBs to run 4.7+ (Burkhead, Winn). That's D2/D3 type of athleticism.

Two RBs that are better than their 40 indicates is DJ Harper and George Winn. Both with very good 10-yards splits.

 

[Xue]

Here's my spreadsheet of past combine times: https://docs.google.com/spreadsheet/ccc?key=0Aph2QMe0NEl5dFpXbXJqUTJvN2h0V3o3V0F3RUk2UFE#gid=0

 

[Xue]

Experimenting with Speed Score for 10-yard split, and 40-10, 40-20 splits:

https://docs.google.com/spreadsheet/ccc?key=0Aph2QMe0NEl5dDRJbU10N1dwc1ZnZUE5aXd0ODFfX0E#gid=1

 

[ZWK]

n=53 now for the 2013 draft class.

Correlations
.955 Xue & Old nflds time
.968 Xue & New nflds time
.968 Xue & Average of old and new nflds time
.971 Old & New nflds times

Suggesting that the new nflds are better calibrated than the old ones (since .968 > .955). If you wanted to take a weighted average of the old and new nflds times in order to maximize the correlation with Xue's times, the optimal weights would be about 72% new times and 28% old times.

Averages
4.558 Old nflds times
4.496 New nflds times
4.545 Xue

 

[rodgev]

Wonder how fast Chris Johnson and Deon Sanders really ran?

 

[Xue]

Wonder how fast Chris Johnson and Deon Sanders really ran?

If you click the spreadsheet 3 posts up, you will see Chris Johnson's time.

 

[Dinsy Ejotuz]

I'm getting ready to do my WR projections for 2013, so finally got around to updating the changes for 2012 and 2013. And I took a look at how the last two years compared to previous years. I included only records where I also had the 10- and 20- yard splits, and excluded those where the only times I had were from pro days. So N = 276 for 1999-2011, and N= 51 for 2012-2013.

I'm not quite sure what they've changed, but it sure looks like they're definitely doing something differently in the last couple years. Average times for first ten, second ten and flying twenty.

99-11 12-13 Change 1st 10 1.551 1.570 (0.019)2nd 10 1.050 0.985 0.065Last 20 1.880 1.888 (0.008)40 yard 4.481 4.443 0.038

So the 10-yard split is actually slower by about two 100ths of a second. Considering the entire range for 2012-2013 is only 16/100ths (1.49 to 1.65 seconds) that's sizeable.

The final twenty didn't change much, less than a hundreth.

But the 2nd 10 yards got faster by 6.5 100ths. That's a ton and accounts for the fact that overall times are about 4/100ths faster in the last two years.

I'm guessing they're using a new timing method -- obviously runners didn't actually get faster in the second ten and slower through the rest of the race.

Also, the 10- and 20-yard splits now correlate much more strongly to the forty time. So whatver they've done seems to be more internally consistent. The splits now account for 62% of the variance in forty times, where from 1999-2011 they accounted for only 46%.

Until we see data for 2014+ I'm going to add four hundredths to the forty time and nick two off the 10-yard split.

If somone wants to run T-tests(?) for the two populations go for it. I'd be curious to know what they show, but not so curious that I want to dust off the stats books to figure out which test I need here.