A Modest Proposal: Coaching Career Sabermetric
I am not, by any stretch of the imagination, possessing any skill with numbers. That I ended up with a degree in Philosophy and not Mechanical Engineering is not an accident.
But the recent scrying of the tea-leaves to discern the fate of one chop-block lovin' coach has piqued a bit of a question in my head that I would love for someone with a bit more sabermetric skill to investigate.
The recent surge of sabermetric evaluation of football at both the pro and college level seems to suggest a very obvious but profound truth - a player is who he is. Dig through enough stats and you can get a pretty solid bead on the athletic ability of a given player and then extrapolate outward from there.
Dr. Saturday (aka SMQ, aka Matt Hinton) has done some interesting work on this in terms of evaluating quarterbacks at the college level. His hypothesis:
The working hypothesis, rather than the conventional wisdom of steady improvement, is that most quarterbacks are who they are from the beginning: initially good passers tend to decline or level off after a strong, mediocre players tend to stay mediocre, and bad debuts are followed with some improvement, but generally level off. The exceptions are the really physically gifted, probably draft-bound players, like Philip Rivers, Vince Young and Pat White, who might (or, in White's case, might not) look bad or mediocre to begin but evolve quickly into much stronger players.
The recent discussion of Tuberville's career in terms of the woeful production of his team this year reminded me of another nifty analysis Hinton did a month ago looking at the career of one Phillip Fulmer at Tennessee. The result is a clear pattern of boom-and-bust and a gradual atrophy in terms of overall quality.
The question that rises in my mind is why wouldn't a coaches career path be deterimined in the same way as we assume an athlete’s would? Clearly there should be some way to evaluate a given coaches level of ability and then, with that in hand, look for patterns in terms of career path and comparison with other coaches of similar skill and ability.
To my knowledge, there isn't any type of metric for this and I'm not sure examining win/loss records is going to be sufficient to provide what I am suggesting. I'd be very curious what OTS or you other numbers-savvy web denizens might be able to come up with.
FanPosts are just that; posts created by the readers of this site. They are in no way indicicative of the opinions of SBN and the authors of Roll Bama Roll.
0 recs |
8 comments
Comments
again, OTS’s touches on my point with his most recent post concerning the Iron Bowl…
Tommy Tuberville is simply no match for Nick Saban, and it is extremely unlikely — read next to impossible — that Auburn will be able to find a replacement for Tuberville that is on Saban’s level. Hell, truth of the matter is that even we probably could not get a comparable replacement for Saban if we had to, so you can rest assured that Auburn won’t.
ok. i think this is simple enough to agree to intuitively but my question is if this is somehow quantifiable in a way to allow more specific analysis?
by kleph on
Dec 3, 2008 10:19 AM CST
reply
actions
0 recs
Well...
…I would love to see someone do it the way some of the computers do teams currently. For instance, rank each coach based on his career record to start a season, then see how he does relative to his competition throughout the year, adjusting for wins and losses along the way. It would be a flawed system, but still interesting.
by NiceLittleSaturday on
Dec 3, 2008 10:35 AM CST
up
reply
actions
0 recs
I wouldn't know how to put the metric together
but I propose that the name of it should be the “Sabanmetric.” groans in agony
I bleed crimson and white...I puke Vol puke orange. RTR
by SugarBowl93 on
Dec 3, 2008 11:30 AM CST
reply
actions
0 recs
The only proper way to do the metric...
…would be to factor in an incredibly diverse amount of aspects of a HC’s job, i.e. win/loss, recruiting, putting together a good staff, discipline, opponents strength (the whole Alabama and Auburn can’t both be good at the same time thing), and etc. I don’t know that you could do it justice, but the way Hinton did it for Fulmer was a pretty good methond.
by Todd on
Dec 3, 2008 12:41 PM CST
reply
actions
0 recs
that's along the lines of what i was thinking...
the bare minimum of data would have to include…
1) win/loss record – including conference wins and bowl appearances but downgrading any MNC effort due to it’s poll-based designation.
2) recruiting – something like PS numbers or the ilk and complemented by nfl draft signees.
3) staff – retention of personnel and number of assistants moving to quality positions
4) pay – including buyout.
5) ncaa violations – number of investigations and/or penalties.
now as to the proper amount of emphasis for these criteria and what others might be necessary i’m all ears…
by kleph on
Dec 3, 2008 12:58 PM CST
up
reply
actions
0 recs
This is totally workable
Select which variables one would like to count toward coaching success. These variables must be quantifiable and not simply Y/N data. Once the dataset has been built I would use ordination techniques to organize the dataset. Surely there will be a lot of autocorrelation within some of the variables, but that will come out during the ordination. Using that info, one could develop a model using linear regression that would evaluate the performance of each coach.
I wouldn't piss off the boys from Alabama . . .
by I hate UT on
Dec 3, 2008 1:44 PM CST
reply
actions
0 recs
and people say my posts make their brain hurt…
by kleph on
Dec 3, 2008 1:58 PM CST
up
reply
actions
0 recs













