NCAA Major-Conference RAPM Last 6 Seasons

[tacoman206]

Below is a link to RAPM numbers for all of the major NCAA conferences for the past 6 years. I counted theA-10, AAC, ACC, Big East, B1G, Big 12, Pac-12 and SEC as the "major conferences".

https://www.dropbox.com/sh/l1yy0tzmudcg ... tdzYa?dl=0

Play-by-play was scraped off of https://www.stats.ncaa.org. Unfortunately, some games just have completely wrong substitution patterns, so about 30% of games weren't accounted for. Also, I error-checked for misspelled player names, but there still might be some out there (apparently there are 4 ways to spell Michael Gbinije).

Let me know if you have any questions or concerns.

[Crow]

And the frontier is crossed. Thanks for sharing this. I plan to look thru it later.

[DSMok1]

Below is a link to RAPM numbers for all of the major NCAA conferences for the past 6 years. I counted theA-10, AAC, ACC, Big East, B1G, Big 12, Pac-12 and SEC as the "major conferences".

https://www.dropbox.com/sh/l1yy0tzmudcg ... tdzYa?dl=0

Play-by-play was scraped off of https://www.stats.ncaa.org. Unfortunately, some games just have completely wrong substitution patterns, so about 30% of games weren't accounted for. Also, I error-checked for misspelled player names, but there still might be some out there (apparently there are 4 ways to spell Michael Gbinije).

Let me know if you have any questions or concerns.

Impressive effort, tacoman206! Certainly a good start. I know parsing NCAA PbP stats is really problematic.

Looking at the results, it appears that, as I would expect, there simply isn't enough data to separate signal from noise in an NCAA season. In the NBA, 82 games of RAPM is still not really enough; the noise is still very problematic. Multiple years are really required, and players changing teams to fix the collinearity issues. In the NCAA, the a season is 1/3 the length, and players don't change teams at all within the season. The noise and collinearity just dominates, unfortunately.

Nevertheless, this is an impressive effort, particularly the data parsing from NCAA.com!

[Kevin Pelton]

Any thought to using this to build coefficients for an NCAA-specific BPM, Daniel?

[jbrocato23]

Great work!

One question: if you're a Tacoman, why 206 instead of 253?

[DSMok1]

Any thought to using this to build coefficients for an NCAA-specific BPM, Daniel?

The results are so noisy as to not make this a worthwhile endeavor, I feel. I used long term RAPM to build BPM, because most of the APM noise was ironed out. At the college level, there will always be quite a bit of noise.

Personally, I feel that the coefficients of BPM should still work pretty well at the NCAA level--the rules are still the same, after all. Do you disagree with that?

[Crow]

I don't know how much time would be involved in building draft college BPM but I would be curious to see where the differences with the pros are or to find the posited close agreement. Is assist - rebounding versatility more or less important? 3 pt shooting vs FT rate? Blocks & steals? Don't have to conclude much or anything, just have more to think and wonder about.

What do the college to pro translation models suggest? Does the BPM comparison generally agree or not? If both are one way, maybe we understand? If they don't, maybe we don't or have to pick a model.

[Crow]

I don't know if it is accidental or intentional but most of the names are chopped in the middle, by my view.

It would be very helpful if the drafted players were highlighted.

Looking for needles in the haystack in 15-16 file, there is a Jackson, who might be Demetrius (but there are other Jacksons) that is one of the best rated drafted players at #23 with a plus 3.71 rapm estimate. Polythress is right behind at 3.58. Now I see Deyonta Davis at #17. Valentine at #28. Ingram is 299th at plus 0.83. I'll look for others.

Hield just a bit behind Ingram. Layman at #131 at just short of plus 2. Onuaku, Brogdon and Gbinije not far behind. Kris Dunn at #188 (but he was estimated worse than #1000 as a freshman), Bembry #264. T. Prince #748 of 914. Payton II at #339. Poeltl #267. Marcus Lee at #21; but nah, scouts don't agree. Ulis #71 but was much higher year before. The only Simmons included last season was... 455th. But the scouts say he is the best or will be... Scout hot pick M Chriss, #594. J Murray, #87.Boston's J Brown, if he is right one, #101. Or maybe he is the J Brown at #593. Labissiere, #201.

[Crow]

I won't ramble on about past draftees (at least not right now); but in data from 4 years ago, Oladipo was #1. And hey, looks like he is at least in top 3 or so for his draft class on last season RPM. Will look at more cases later.

[Crow]

The rules in college are not "the same" in print or interpretation regarding shot clock (and ball advancement), 3pt line, illegal defense, size of the painted area (and maybe restricted area) fouls in many ways (5 vs. 6 fouls, handchecking, what is allowed in screens and rebounding, breakaways?, intentional?, palming, walking / eurostepping, fouling 3pt shooters after the release?, flagrant fouls, flopping, contact at rim before / after shot release, continuation / and 1 patterns, charge / block, contact away from ball, college 1 and 1s, nba "fouls to give" at end of game, arguing calls, showboating / taunting, delay of game, what gets called / not in crunch / clutch, homerism, referring the margin for entertainment purposes, overall referee quality, consistency, forwardness, etc.) length of game, jump balls and perhaps other subtle things (timeouts?, rim tightness / consistency?, average ball pressure / condition?, clock operators?, replay?, the schedule (back to backs / rest days, giving up in 4th period), superstar treatment, etc.

Some of these might not be different enough to count; but many will be by at least a little or more.

Certainly skill level and skill depth are different and pace.