Friday, November 11, 2005


Three cheers for distributed intelligence

So yesterday I took a crack at quantifying good coaching. I realized going in that the effort was probably futile: there's so many variables in team performance besides "Is the coach doing a good job?" that I seriously doubt we'll ever be able to numerically separate coaching performance from team performance". Coaches will continue to be discussed in terribly subjective terms: "the players have really bought in", "they've tuned out", "the players respect him", "he's not tough enough on them", etc. etc.

Anyway, my hypothesis was based on the CW about "a good road game" and "a good home game", and attempted to determine whether Road Success relative to Home Success (RSrtHS) could be correlated with good coaching.

The answer was Maybe/Inconclusive, but then Tim and sacamano had a couple of interesting comments. callmeteem:
Sometimes homecrowds can be a disadvantage. I am thinking particularly of the Montreal crowd where it seems the pressure can be unique--particularly on young players.

I was going to say that length of travel and duration of roadtrips might have some effect (i.e., teams in the West get screwed because of the brutal travel schedule)...

I decided to run the numbers for the previous two seasons as well. For the three seasons 2001-2004, each team was ranked by RSrtHS from 1st to 30th. I then averaged each team's rank over the three seasons, and used the averages to rank them all again from 1st to 30th, giving us RSrtHS for the past three seasons. That's a decent sample size: 123 home and 123 road games for each team. And look what we have here:

Eastern Conference teams are highlighted. Not exactly an random-looking distribution, is it? The six teams with the worst RSrtHS are all from the West: in short, home-ice advantage has meant the most to them (reminder: this is different from saying they're the best home teams). Give sacamano a cigar.

And I think Tim has earned one too. If was to choose the two cities where the fans are quickest to turn on the home team when things go bad (not on call-in shows or message boards, but right there in the stadium), I'd pick New York and Philly. Maybe it's a coincidence, but the two best road teams over the past 3 seasons, relative to their own home records, are the Rangers and the Flyers.

So what do have here, in the end?

1. A theory about dispassionately evaluating coaching performance debunked;

2. Another way of showing that the travel schedule of Western teams really is a bugger; and

3. Maybe, just maybe, some evidence that it really is important to support the home team, no matter how they're playing. Jennifer Kapitski, you have my qualified apology.


You're going to have difficulty selling me on any theory--about anyone's performance--that treats some regular-season games as being worth more than other regular-season games. There are two points on the table every night. There's a tendency around here to say that good teams win the games they need to win (defined variously as "games against outstanding teams" and "games against crummy ones"). The one certain thing in the regular season is that the good teams win the most games over the course of a season.

All that said, if I were looking for the footprint of good coaching, I'd start with the teams that extract maximum value from the close games--which would point you to overtime records (which show how a team does when the bench gets shorter and set plays become more important), and to the teams that outgross the expected points from their overall goal differential. But even these experiments have huge, obvious, probably irreparable problems.

I don't think the original hypothesis involved treating some games as being more important than others. It simply set out as its premise that road games have been more influenced by good (or bad) coaching than home games. This may not have been the greatest premise, but it's no more inherently illogical than stipulating that overtime games have been more influenced by coaching that those settled in regulation.

As for the bit about must-win games, your point is taken, but I do think there is a certain perspective from which all wins (or losses) are not equal. This probably needs its own post, but just as one example:

There are "bad" losses where the timing and circumstances might influence team management and/or coaches to make moves that may not be beneficial in the long term.

Obviously this distinction is senseless if you are the GM, but for the rest of us lowly fans, it makes perfect sense. (Put another way, it's OK for a fan to describe a game as must-win, but if your GM does, he's the wrong guy).


I can't see how using the OT records would be a good gauge either when the losing team still gets that lousy single. There's a different mindset when playing for a bonus single, rather than all or none.

The stats I would love to see are the "Games won going tied or down one into the 3rd period". That would be the mark of a truly well coached and well motivated team.

Not OT.

That idea is certainly no less valid than mine--but then again the optimal strategy for a team with poor talent might be to try to hold onto a tie going into overtime and sneak out with the point. In that case, a well-coached bad team might have fewer outright wins than one which unwisely played for high stakes and suffered two or three extra regulation losses for every late-and-close victory. This is a typical example of a universal and insoluble problem with this kind of analysis (which hasn't made any headway to speak of in baseball, despite the elegant statistical tools and relative masses of data available in that sport).

True enough. We were starting to see that very phenomenon in the NHL pre-lockout.

It's scary looking at the standings pre-lockout and how many games ended in ties or overtime losses.

Post a Comment

<< Home

This page is powered by Blogger. Isn't yours?