Sunday, October 2, 2016

How much do unbalanced conference schedules matter? Not much.

One of the many downsides of the trend towards mega-conferences is the death of the round-robin conference schedule in basketball. The resulting unbalanced schedules raise the distinct possibility that regular season conference championships will be decided by quirks of scheduling fate rather than talent and coaching tantrums directed at referees as God intended.

The Investigation

But how much difference do unbalanced schedules actually make? To find out, I looked at every season since 2008-09 and used T-Rank to calculate each team's:

1) Expected wins against the conference schedule it actually played;
2) Expected win percentage against a true round robin, and used this to calculate expected wins based on actual number of conference games;
3) Expected chance of winning a share of the title against actual schedule;
4) Expected chance of winning a share of the title against round-robin schedule.

The differences between 1 & 2, and 3 & 4 show the effect of an unbalanced schedule, though in slightly different ways. So I looked at both.

First, the conclusion

Overall, the data is pretty clear that unbalanced schedules are rarely extreme enough to be decisive. By far the dominant force in college basketball is variance, and that's why we love it. All the data is available here.

An Extreme Example: Wisconsin 2016

Let's look at the results for Wisconsin last year as an example.

Expected wins against actual schedule: 10
Expected wins against balanced schedule: 10.8
Chance of winning title against actual schedule: 1.8%
Chance of winning title against balanced schedule: 4%

Obviously, the Badgers got a bad draw last year. Indeed, the -.8 wins is the second most difficult schedule in the entire database. But it's hard to say that even this extreme schedule had much of an effect on anything. The Badgers finished a full 3 games behind Indiana. Even accounting for Indiana's favorable schedule (+.34 expected wins) the Badgers still finished 1.86 adjusted games back:

Team Wins EW Diff Adj. Wins Actual EW Blnced EW Actual Ch% Blnced Ch% Ch% Diff  Adj GB
Indiana 15 0.34 14.66 13.4 13 29 22.4 6.6 0
Wisconsin 12 -0.8 12.8 10 10.8 1.8 4 -2.2 1.86
Michigan St. 13 0.22 12.78 14.5 14.3 61.9 56.6 5.3 1.88
Iowa 12 -0.59 12.59 11.3 11.9 6.3 10.8 -4.5 2.07
Maryland 12 -0.37 12.37 11.5 11.9 6.8 11.8 -5 2.29
Purdue 12 -0.2 12.2 13 13.2 23.3 25.7 -2.4 2.46
Ohio St. 11 0.23 10.77 8.3 8.1 0.1 0.1 0 3.89

Perhaps more importantly, the Badgers were an extreme long-shot to win the title under a balanced schedule (just 4%) and while the unbalanced schedule cut those long odds by more than half, it's still hard to complain too much about that.

The few examples where it made a difference

Now for the fun stuff: Looking for examples where an unbalanced schedule actually made a difference. First, I looked for teams that didn't win the title and finished less than .5 "adjusted games back." Here they are:

Team Year Conf Wins EW Diff Adj. Wins Actual EW Blnced EW  Adj GB
Coastal Carolina 2014 BSth 11 -0.22 11.22 10 10.3 0.43
VCU 2012 CAA 15 0 15 14.3 14.3 0.47
Dayton 2009 A10 11 -0.46 11.46 9.7 10.2 0.48

So just three teams over 7 seasons finished less than .5 adjusted games back, and even these three barely cleared the .5 threshold.

VCU is an interesting one because their schedule actually did not disadvantage them compared to a balanced one. The problem was that that the Colonial champs that year, Drexel, had a very favorable schedule, worth +.53 wins. Most importantly, Drexel only had to play VCU once, and that game was at Drexel. This is when an unbalanced schedule can be really unfortunate: when there is a clear top two that play only once, the team that gets the home game has a big advantage.

The other way I looked at this was to calculate likelihood of winning a championship. I think this is less good than looking at the "adjusted wins" because it's entirely hypothetical. For example, Wisconsin vastly outperformed expectations in conference play last year -- winning 12 games when T-Rank would have expected a team of their quality to win only 10 on overage. As it happened, Indiana also won two more games than expected. But if they'd won 13, as expected, it would have been fair to say that their easier schedule and Wisconsin's harder schedule combined to rob UW of a championship. But if you just look at the simulated difference in Championship expectations (4% vs. 1.8%) for Wisconsin, that doesn't show up.

That said, here are the five teams whose championship odds were negatively affected by 10% or more:

Team Conf Year Wins Actual Ch% Blnced Ch% Ch% Diff GB
Mount St. Mary's NEC 2010 12 37.1 53.9 -16.8 3
VCU CAA 2010 11 28 43.5 -15.5 4
VCU CAA 2012 15 48.6 60 -11.4 1
Syracuse BE 2013 11 12.9 24.3 -11.4 3
Butler A10 2013 11 15.1 26 -10.9 2

VCU's 2012 team shows up again, as does its 2010 team. But the 2010 team illustrates the downside of this purely hypothetical analysis: that team actually finished in a tie for 5th pace, a full four games behind Old Dominion. Similarly, each of the other teams (other than 2012 VCU) finished at least 2 games out of first. So although all these teams definitely got screwed by their schedules, they didn't perform well enough to really feel sorry for them.


So the overarching conclusion is that most of the time unbalanced schedules are not that big a deal. But a couple caveats:

1) The conclusion that unbalanced schedules don't really affect conference championships doesn't mean that there aren't other effects. Clearly, a tough schedule can easily cost a team one win, and it's not that unusual that one win is the difference between making the NCAA tournament and sitting at home. Pertinently, the 2010 VCU team noted above has a good claim to losing one win based on a bad schedule. That team ended up losing in OT to Old Dominion the Colonial Tournament Championship game, and didn't make the tourney -- despite being No. 51 in Kenpom. It was the next year that VCU snuck into the tournament, as a First Four participant, despite a significantly worse profile, but rode Almighty Variance to the Final Four.

2) Although most schedules are balanced enough most of the time, this is only true in the end. It's still very important to look at who's played whom at a given point in the season. For example, Indiana got a lot of flack about its Big Ten schedule last year, even though in the end it was just marginally favorable. But the real issue was that their schedule was extremely unbalanced temporally, with a very soft 7-game stretch to start. That was a legitimate thing to point out at the time, even though IU continued to surprise even when things got tougher, and cruised to the title in the end.

3) I haven't looked into this systematically, but the trend does seem pretty clear toward more extreme results recently. This is not a surprise, as the rise of the super-conference is relatively recent, and is still in progress. So although we haven't definitely seen it yet, we likely will see an unbalanced schedule decide a major-conference championship soon enough.

No comments:

Post a Comment