Over&Under Achieving Coaches per Bill Connelly
Interesting way of looking at which coaches over- and under-achieve using actual vs expected wins. Win expectancy “is intended to say "Given your success rates, big plays, field position components, turnovers, etc., you could have expected to win this game X% of the time." It has nothing to do with pre-game projections or opponent adjustments.”
only coaches with 3 years of experience from 2005-present are included, for a total total of 229 coaches. Bill Snyder is first with +1.12 wins per year. Other coaches of note:
- Urban Meyer +.68 10th
- David Shaw +.55. 23rd
- Les Miles +.48 31st
- Jim Tressel +.37 46th
- Mike D’Antoni +.33 50th
- Nick Saban +.28 55th
- Rich Rod +.24 64th
- Lloyd Carr +.04 108th
- Brady Hoke -.03 112th
- Jim Harbaugh -.16 139th
As Connelly points out, this metric is only one part of determining “good” and “bad” coaches. However, I’m surprised at Harbaugh ranking below Hoke and RichRod.
Dodgy methodology is dodgy. All the things listed as "given" are the things you'd expect would be the products of good coaching. I'm not sure what's left after normalizing the sample for all those things... luck?
Does the method put the same people in roughly the same ranks (or quartiles, at least) year in, year out, or does it change like a list of "Best Mutual Fund Managers" year to year? (which is mostly luck... which is why you should invest in low-cost index funds, but that's a different discussion)
Yeah, the "etc." in his description makes me wonder what makes this distinct from a simple Pythagorean expectation calculation based on scoring margin. Because it feels very much like a counting stat and not much else.
Is it completely luck? Or are some coaches better in close/rivalry games? I think it's an open question and not as simple as just luck.
In fact, the major complaint Michigan fans would have about Harbaugh right now would be literally that. It's hard to argue conclusively one way or the other, but it's not something I'm willing to just write off completely as luck.
The last part is that of course this methodology is not meant to be taken as holistic coaches ranking. It's one specific element. If you want to normalize it by something like actual season performance, then you would see Harbaugh greatly outperform Hoke.
The eye test tells you these results look pretty random.
As a Michigan fan I thought Hoke would be near the top as he seemed to win games we had no business winning. Lloyd and Jim make sense as they lost games they had no business losing and RR games were so crazy who the hell knows.
I did not read through all of the stats he linked used to determine this, but am I right to determine that be assumes big plays are luck and not produced at some level through coaching?
I take this to mean that JH has produced a lot of big plays and other positives and when his "luck" regresses to the mean he will produce more wins.
That Hoke and RichRod won as many games as they did, given their collective nincompoopery, rightly supports their overachiever statuses.
Great band name....Collective Nincompoopery.
Welp, I guess there’s only one course of action left for Manuel—fire Harbaugh and re-hire RR with Hoke as his DC.
Has RR started hiring DCs?
Niiiiiiice!
Let’s be honest - Harbaugh had one good season and it was with Hoke’s kids.
If we ignore all of his previous coaching stops, maybe. But if you also consider the job he did iat Stanford and San Diego, it is more than that. Also, Hoke only managed to go 5-7 with his kids and then Harbaugh turned them around to 10-3. Further, Jake Tudock, a Harbaugh kid was a pretty important piece of the puzzle.
Saban is considered by many to be the GOAT coach, but is only 55th on this metric. I find this metric... annoying.
To rank highly on this list you have a fair number of games the numbers say you should lose. So guys like urban and Saban rank low because their teams aren’t dominating the other teams maybe once or twice a year whereas bill Snyder probably has 5 to 6 opportunities a year.
its the same reason Nick Saban doesn’t win coach of the year every year. It’s impossible to exceed expectations when you are the best.
I don't think you understand it very well.
This is saying that these coaches have won more games (Snyder) than the performance of their team on the field that day (according to Connely's formula) would suggest.
And some, like Harbaugh, have won less than the performance on the field would suggest.
Basically the punt and the missed tackle against FSU are the two main culprits here.
Per Connelly's numbers - Michigan had only a 6% likelihood of winning the FSU Orange Bowl. That game isn't really the culprit.
The 4 games in JH's U-M tenure where they won (lost) with a under (over) 50% chance of winning per Connelly's numbers:
1. at Utah, 2015. 58% chance, lost.
2. MSU, 2015. 68% chance, lost.
3. at Minnesota, 2015. 45% chance, won.
4. at Ohio State, 2016. 83% chance, lost.
Game 3 obviously had a coaching error (Minnesota's epic clock mis-management). Game 2 arguably did, depending how much one blames the lack of max protect. Game 4 was your classic "outplayed them but turnovers undid all of that" game. And Game 1 was probably just random noise.
It's an interesting methodology by Connelly, but in the end --- as w/ Michigan's small subset of 4 games above --- it's not all coaching. It's a mix of coaching, players/execution, and random noise.
Again this metric isn't actual wins vs Connelly's pre game win probability.
It's an analysis of actual plays from the game and whether or not the win total over or under achieved vs that performance.
Again, for UofM under Harbaugh, the coaches called the plays effectively, the plays succeeded enough to win; but, for some other reason (luck) the win total fell behind the true performance.
This is a good post, but coaching error on max protect is something only fan bases predisposed to finding the negative in another team would invoke.
The punter dropped the ball. It wasn't a rush and block. Yes they did get there sooner because the rush was on, but he dropped the ball. This is extremely rare. Then, instead of falling on it he tried to kick it. More rare. He missed it when he tried to kick it. More rare. Then the ball was bouncing around and the opposing team recovered instead of M (which you'd expect given the Sparty players are charging and M's are not, but still a chance it bounces to an OL). Then the ball bounces to a player with momentum toward the end zone with few people around to tackle him. This is an incredibly rare event that will never be repeated if the teams play for another 500 years. To put any of that on coaching is silly. In fact that play and season is a great illustration of random luck in practice: OSU was far and away the best team in the B1G that year, but MSU benefits from the punt and the horrendous game plan of OSU, along with rain/weather in Columbus, and makes the playoffs. They got trounced because they never should have been there.
Michigan handed off to TE on 4th and 1 against South Carolina. Worse yet, he wasn't supposed to get the ball, and no one - not the head coach, not the OC, not the QB, not the TE himself - called a timeout. This was an epic coaching fail in a game they should have won.
To me, the coaching blunder in the MSU game wasn’t the lack of max protect, it was the fact that the punter wasn’t coaced to fall on the damn ball.
OSU 2016 I can attribute completely to luck. O'Neill game though you have the coaching blunder on the punt itself but even more than that, it's not hard to argue the game shouldn't have been a one score game at that point. Iowa 2016 I put almost entirely on Harbaugh.
So, yes, he's had bad breaks, and since we're only talking about a handful of rivalry/close contests it's hard to separate luck from something not random, but in my opinion our 2016 Iowa loss and 2015 MSU loss cannot be fully attributed to luck.
I think you are probably correct.
The formula probably will he particularly inaccurate/unfair as a judge or coaches/teams that are regularly winning 11+ games per year because the expected win % is almost always going to be positive in the very few games those teams end up losing...
Saban don't need luck to win, he got a process to "trust".
Even Connelly seems unsure about this metric. I didn’t point out in my OP that Connelly focuses more on consistently outperforming win expectations, I.e. Pat Fitzgerald outperformed 9 out of 12 years more than just the total outperformance. Forgot to add him to the list but he ranks 5th at +.88 wins per year.
Connelly specifically said that Steve Kragthorpe is ranked close to Saban on this list...and no other.
Guy posts article with clear methodology based on stats and info this site loves and respects. Most commenters see harbaugh might be cast in a slightly negative light, don’t bother to read article and proceed to trash it because they don’t like the outcome. Fun times.
Fuck off sparty troll
Bill C would be the first to say his numbers can be squirrelly and give misleading results for a specific circumstance or another. Nobody is attacking the stat, only questioning the efficacy of a single number that includes a half dozen named factors plus "etc." in it's description.
But then, my guess is you didn't really read the description otherwise and already had your comment written the moment you saw the title.
In the last 50 years, Michigan has three losing records. Rich Rod has two. Hoke has one. That doesn’t sound like overachieving. The only year I think you can say Harbaugh underachieved was last yesr. He still went 7-5. Based on this admiitedly subjective analysis,any methodology that ranks Harbaugh lower than those two, regaurdless of whether the methodology is clear and it is based in stats seems suspect,
Have to disagree with you.
First, most posters have been willing to criticize Harbaugh when deserved. Not many "Harbaugh can do no wrong" folks left around here.
Second, the metric does not merely "cast [Harbaugh] in a slightly negative light." It ranks him among the worst coaches in CFB.
So...a bunch of posters who have been willing to criticize JH when deserved are now criticizing Connelly when deserved.
I guess I don't get how that metric has anything to do with whether a coach underachieves or not.
Just a glitch in the formula:
You see, when the Defense stopped JT short, their win probability shot up to 100% and, unfortunately, there's no algorithm to adjust for human error.
Really Connelly, "over/under achieving ".
I guess calling it luck just didn't seem sciency.
Your, uh, least impressive debuts: Colorado State’s Mike Bobo (minus-0.7) and Kansas’ David Beaty (minus-0.6).
To be fair, winning at all would be overachieving at Kansas. Beaty was right in line with expectations otherwise.
I normally agree with Bill C, but I do think this metric is a bit misleading. For example, Matt Rhule and Kevin Wilson are listed below Harbaugh. Now, they aren't amazing coaches by any means, but Rhule was the guy who took 2-10 Temple to back-to-back 10 win seaons for the first time ever. But then he went to the king of tire fires at Baylor and went 1-11 due to a metric ton of bad luck and karma, most of which he had no part in. And Kevin Wilson took Indiana and made them competitive for the first time since, what, Lee Corso was on the sideline?
By comparison, Al Golden, a man with a 59-59 career record and this image as the first one that shows up on Google is spotted, what, 40 spots higher because I guess he was just mediocre at Miami?
In totality, this metric makes some broad strokes that are helpful in quantifying good and bad coaches. But like all metrics it can be a bit dodgy and needlessly over-fitting. And I'll admit to not having the formula in front of me, but I assume this only sorta takes into relative strength of schedule, which is how Pat Fitzgerald going under .500 before the switch to East-West divisions suddenly is considered a coach who does a lot with a little.
The metric is obviously flawed.
But given the way it is described (amount below expected win rate given a range of in-game metrics), Michigan 2017, and to a lesser extent 2016, is going to measure out as hugely under performing.
A team with a great D that gets lots of stops, 3-outs, etc., and puts the offense in lots of good field position situations, an extra offensive possession or two per game, etc., coupled with an offense that fails to score much off the way above average number of good situations, describes Michigan perfectly ever since the loss at Iowa in 2016.
IOW, if you have an offense that is bad and fails to score much off of all of the great situations your top 5 D generates, then your will win games at a rate way below the expected win rate given all of those great situations for your offense...and look really bad under the formula the metric uses.
On the eyeball test Michigan has been a huge under-performer since Iowa 2016. Much of that is down to QB problems in 2017. But we all saw plenty of terrible offensive game plans and situational play calling in 2017 too. Former issue is probably fixed with Patterson. Latter issue? We will see if the coaching staff changes fix it in September.
Read the last sentence of post and thought, Stupid, Ignorance, Bias, Click Bait. Cannot wait for years four and beyond. This shit will no longer be posted here.
<iframe width="560" height="315" src="https://www.youtube.com/embed/Ct1ckS2Kb9o" frameborder="0" allow="autoplay; encrypted-media" allowfullscreen></iframe>
So, I just learned I cannot embed videos here with the new site.
This new site sucks!
Now I am trying a simple hyperlink...https://www.youtube.com/watch?v=Ct1ckS2Kb9o
this was the dumbest hecking thing I've ever seen.
Well, at least that worked.
This sounds like the luck metric on kenpom. “We expected you to win this many and you won more/less”. So I think a reversion to the mean is due for JH