NCAA Weekly, 13.2: What's the most important month of the season? Answer: not February

controversy!

The general narrative of college basketball is that you want to be playing your best basketball at the end of the season. Obviously, it’s a sensical thing to want. If your team sucks heading into March, you’re probably not gonna expect much from them once the bracket is revealed. If your team is on fire, you simply cannot wait for the NCAA Tournament to begin.

I’ve long been curious about this, so I went about finding ways to measure it. The first thought I had, thanks to Bart Torvik’s site: what about taking the top 10 teams in every individual month, from 2008 to present, and measuring how they performed in March? The argument is certainly there to go team-by-team and measure individual cases, but I already did something similar for Field of 68 centered around UConn and Creighton. Those results, albeit part of a smaller sample, were fascinating. The teams that “peaked too early” ended up generally overperforming in March:

While the teams that came in guns blazing disappointed more often than not.

We know how it works, generally, on a team-by-team basis. What about on a macro scale? I used Torvik’s data from 2008-2022, excluding 2019-20 and 2020-21 for COVID reasons, and took the top 10 teams by month. Considering there’s four months in the regular season and 13 full seasons observed, that gives you a sample size of no fewer than 130 teams to look through and a total of 520. All in all, there were nearly 300 unique teams observed over the period.

What I found in the process likely isn’t great for commentator narratives, but can provide us with more insight into why the Tournament works like it works:

And why those above results are so darn interesting.

The case for January

If you were to go by pure win total, as seen above, it goes January > December > November > February. This makes some amount of sense, of course. January’s average top 10 seed of 3.2 was the highest by a decent margin, and of the 52 1 seeds observed, 39 (75%) played like a top-10 team in January. Those 1 seeds averaged 3.46 wins; the one seeds that didn’t play like a 1 seed in January won 2.77 games on average. (If the current 1 seeds hold, consider betting against Kansas, who was the 18th-best January side.)

If you’re looking for when the champion plays its best basketball, this is of very minor help. Both UConn teams (2011 and 2014) were exceptions here, as was 2017 North Carolina, but in general, the eventual national champion was a top-10 team every month. 10 of the 13 observed in what I’d call ‘normal’ (i.e., non-COVID-affected) seasons were a top-10 team every month. (The only team this year to meet that requirement is Houston, though Alabama, UConn, Tennessee, and UCLA have all been top-10 teams in three of the four months.)

But, in general, there’s been something to showing your strongest stuff in January. No month was stronger in terms of average seed, total wins, or expected wins in March. In general, strong January teams overperform by a hair in the NCAA Tournament, and of all four months, no month was better at determining which teams will eventually end up 1 and 2 seeds. Of the 130 January teams measured, an impressive 72 (55%) ended up on one of the top two lines. No other month topped 65.

The case for November/December

The most ignored part of the college basketball season by the general layperson ended up being the most important for figuring out March success. I thought about putting November and December in separate arguments, but they’re so intertwined that it made sense to keep them together. I mean, check it out:

• Average seed: November 3.81, December 3.82

• Total wins in March: November 295, December 298

• Performance over seed expectation (PASE): November +38.5, December +39.7

• Overperformers in March: November 74 of 130, December 75 of 130

However, the case for the first two months of the season is less about these numbers and more about what type of November/December successes end up happening down the line. It may come as a giant surprise, but November performed better than any other month in finding future Final Four teams. 29 of the eventual 52 Final Four teams were in the November top 10; no other month topped 27.

Meanwhile, December was #1 on the list in terms of finding future Elite Eight participants. 52 of the eventual 104 Elite Eight teams were in the December top 10; only January was close at 50. November was the strongest month at finding future Sweet Sixteen teams (83); December was #1 in terms of finding future top-3 seeds that made the Elite Eight or further (48 of a possible 71).

But above all else, November/December is a gold mine for finding lower-seeded teams that get hot in March.

Teams that went to the Sweet Sixteen or further as a 5+ seed AND were in a month’s top 10, 2008-2022:

• November: 12

• December: 8

• January: 6

• February: 5

Alternately, if you like the reverse of this:

Teams that lost in the Round of 64 as a 1-5 seed AND were in a month’s top 10, 2008-2022:

• November: 6

• December: 5

• January: 5

• February: 10

Or, one more:

1-4 seeds that lost before the Sweet Sixteen AND were in a month’s top 10, 2008-2022:

• November: 19

• December: 22

• January: 26

• February: 25

So maybe all of this would not be explained away as any of November, December, or January being the most predictive month. It might simply be a more negative case than that.

The case against February

The most important month on-paper in college basketball has a lot going for it. The Super Bowl and football season have finally ended, so much more attention gets dropped on the college hoops world. Ratings go up. Games feel more important because they’re for real conference stakes. Joe Lunardi puts out a new bracket every day and attempts to illustrate the exact impact a win or a loss would have on your team’s seeding.

All of that being said, across the last 15 years, February has proven itself to be the least-predictive, least-important month in the college hoops calendar. No month’s top 10 performed worse against seed expectations than February. No month’s top 10 had more March underperformers than February. Heck, no month’s top 10 had more 1 seeds fail to make the Final Four than February. (The month with the fewest 1-seed underperformers in this regard: December.)

Considering that February had the second-best average seed and therefore the second-highest average wins in March, for it to be pretty easily the worst-performing month on record is quite remarkable. Really, past the top two seed lines, February was a bad month to be a good team.

PASE for February’s top 10 teams, 2008-2022:

• 1-2 seeds (65 teams): 203 wins (+13.61 above expectation)

• 3-16 seeds (65 teams): 79 wins (-5.04 below expectation)

Teams that came into March Madness white-hot generally underperformed. Of those 65 3-16 seeds (really just 3-12), none lower than a 3 seed (2017 Oregon, 2018 Michigan, and 2019 Texas Tech) made the Final Four. It was especially bad if you ended up a 5 or 6 seed. Of those 17 teams, nine failed to win a single game, and only three cracked the Sweet Sixteen.

Those PASE stats are of additional interest because of how it worked for the other months of the season.

PASE for November’s top 10 teams, 2008-2022:

• 1-2 seeds (58 teams): 178 wins (+9.6 above expectation)

• 3-16 seeds (72 teams): 89 wins (+27 above expectation)

PASE for December’s top 10 teams, 2008-2022:

• 1-2 seeds (59 teams): 198 wins (+29.42 above expectation)

• 3-16 seeds (71 teams): 100 wins (+11 above expectation)

PASE for January’s top 10 teams, 2008-2022:

• 1-2 seeds (72 teams): 213 wins (+9.9 above expectation)

• 3-16 seeds (58 teams): 92 wins (+15.1 above expectation)

Basically, all other months’ top tens perform very well across all seed lines. The late risers of February, to put it simply, are largely not worth your time. Teams that perform really well in November and December but sag off in conference play, though? Maybe consider giving them another look.