How stats and history would pick the 2021 NCAA Tournament

Round of 64

West Region

(16) Norfolk State over (16) Appalachian State. This one is obviously not the most meaningful game in existence, but Norfolk is a tad ahead of App State on both Torvik and KenPom. It’s a coin flip, though. The favored team in First Four auto-bid games is 11-7, if you’re curious.

(11) Drake over (11) Wichita State. Torvik actually likes Wichita here, but Drake is a heavier favorite on KenPom, so I’m riding with them. I think. Maybe. The favored team in First Four at-large games is 10-8, so expect one upset.

(1) Gonzaga over (16) Norfolk State. My combined KenPom + Torvik + recent performance metric has this as literally an equal split either way; Gonzaga would be favored by 28.8 against both App and Norfolk. A game to skip.

(8) Oklahoma over (9) Missouri. Coin-flip like, but as we’ll go into further detail on, you should just pick the favorite in 8/9 games. The KenPom/Torvik (we’ll just refer to this as “metrics” the rest of the way) favorite is 38-19 across 20 Tournaments; the underdog is 8-15. I wish Oklahoma was favored by more, but neither team has great odds the next round, anyway. It’s a one-point pick.

(5) Creighton over (12) UC Santa Barbara. This one hurts. I LOVED watching UCSB this year and they fully deserved a chance to make some March moments. Sadly for the Gauchos, they’re matched up against a legitimate top 20 team with a 71% shot of winning. 5 seeds at 70% or better are 26-5 in our data set; all others are 23-26.

(13) Ohio over (4) Virginia. I really don’t know what to do here. 4 seeds at 75% or worse to win their first game are just 13-12, and somehow, we ended up with a Tournament where three of the 4 seeds are at 75% or worse. Virginia and Oklahoma State have nearly identical odds of winning. Here’s the thing: Virginia also has the second-best Sweet Sixteen odds of any 4 seed. They’re in a situation where they may either lose here or win two games. My final pick here is Ohio after applying a small COVID deduction to Virginia’s ratings; simply put, it made them the most vulnerable 4 seed.

(6) USC over (11) Drake. USC’s odds are already good for a 6 seed, and they get marginally better if Wichita wins. 6 seeds at 63% or better to win are 26-7. USC actually sits at 69% or better in every metric I use, which makes me feel even better about them.

(3) Kansas over (14) Eastern Washington. I was ready to submit an easy one here before it was revealed Jalen Wilson won’t be traveling with Kansas to Indianapolis. Kansas is already not as good as a typical 3 seed, and sans Wilson, I have them at about 76-77% to beat Eastern Washington. I’d have either team departing the field in the Round of 32, so I’ll take the risk with KU.

(7) Oregon over (10) VCU. Torvik favors VCU; KenPom, Oregon. The net result is Oregon by a hair, but neither result will be a surprise.

(2) Iowa over (15) Grand Canyon. If any 2-15 matchup this year can bring some first-round weirdness, it’s likely this one. Iowa is at 91% on KenPom and 89.6% on Torvik, and my combined metric is a hair above 90%. 2 seeds are 75-5 since 2000, but no 2 seed has ever lost when coming in as a 93.4% favorite or better (24-0). Three of the four 2 seeds are fine, though both Ohio State and Alabama cut it close. The only one that doesn’t meet this mark is Iowa. It’s still unlikely they’ll actually lose, but this is the only 2-15 I can see where an upset is even possible.

South Region

(1) Baylor over (16) Hartford. Baylor favored by 21 points or so.

(9) Wisconsin over (8) North Carolina. Moderately tough. Wisconsin’s lost 12 games this year, but all the advanced metrics adore them, likely because Wisconsin has had several dominating wins and only a couple truly wretched performances. If anything, I actually really like Wisconsin as a value pick because the UNC name alone will sway a lot of people in bracket groups.

(5) Villanova over (12) Winthrop. This is terribly tough to do, because Villanova will be missing maybe their second-best player in Collin Gillespie and they’ve lost two straight games without him. Bart Torvik’s missing players tool values Villanova as roughly three points worse without him, which is like going from playing #12 Villanova to #28 Villanova. It’s a huge deal. And yet: at #91 in KenPom, Winthrop is the lowest-rated 12 seed since Chattanooga in 2016. (They lost by 25 to Indiana.) Villanova still meets my 5 seed requirements even without Gillespie. 5 seeds at 74% or better are 20-2; Villanova is no lower than 75.6%.

(4) Purdue over (13) North Texas. Remember the “4 seeds at 75% or worse to win their first game are just 13-12” stat above? Purdue sits at just 73.4% to win on my sheet, and neither KenPom nor Torvik have them above 73.9%. Even 538, a system which is much more charitable to a lack of upsets than most, is only at 81% for the Boilermakers. And yet: I’m riding with Ohio over North Texas. Either pick is totally defensible and reasonable, and if I were a more daring man, I would’ve selected both. Hopefully both win, to be honest, as I have the 5 seeds in these quadrants advancing.

(6) Texas Tech over (11) Utah State. 6 seeds at 63% or worse to win are 20-27, and Texas Tech barely cracks the 60% mark. At least one 11 seed has won in 15 straight NCAA Tournaments, and I see no reason that’ll change this weekend. And yet: if UCLA wins, Texas Tech will simply be one of three 6 seeds at 63% or lower. Statistically, at least one has to win. The matchup doesn’t make a ton of sense for USU after further exploration.

(3) Arkansas over (14) Colgate. This terrifies me. A 3 seed hasn’t lost in the Round of 64 since 2016, and only eight have lost in the last 20 years. However: we have an unusually weak group of 3 seeds this year. None rank higher in KenPom than Arkansas at #18 (the equivalent of a 5 seed), and the other three rank #22 or worse. If we don’t get at least one 14 seed victory, it’ll be a moderate surprise. I only have the 3 seeds at 37% to go 4-0, the lowest collective odds since…2016. This is the first year since 2012 where three of the four 3 seeds are below 80% to win. 3 seeds below 80% are still 19-5, but that’s a far cry from the 53-3 mark if you’re above 80%.

Anyway, Colgate is ranked #84 nationally, and I think they’re going to make this very scary. You’ll see the precaution I’m taking against Arkansas in the next round.

(7) Florida over (10) Virginia Tech. Florida is around a 55-57% favorite; just by picking the favorite in the 7/10 games over the last 20 years, you would’ve gone 62-18. It’s worth the occasional loss if it means you’re beating just picking the higher seed by 17.5%.

(2) Ohio State over (15) Oral Roberts. Oral Roberts is fun, but no metric I use has them better than about a 6.5% chance to win. The OSU defense is kind of bad, though, so I could see them making this interesting.

Midwest Region

(1) Illinois over (16) Drexel. The only 1-16 game that I’m even mildly intrigued by. Drexel is a full 46 spots ahead of the next-highest 16 seed (Hartford) and represents the only 16 seed with a greater than 3% chance of winning. This still lends itself to Illinois being ~95% likely to win, but in general, most of the classic 1-16 scares are represented by teams at 5% or better to win. You should probably watch at least some of this, even though Illinois will probably win by 15.

(8) Loyola Chicago over (9) Georgia Tech. Brutal matchup for Illinois. Loyola is the second-highest rated 8 seed ever, and Georgia Tech is a quality 9 seed that ranks 32nd. Loyola is at 62.7% to win, though, and 8 seeds at 57.6% or better are 22-5.

(5) Tennessee over (12) Oregon State. Not just bias, I promise. 5 seeds at 70% or better are 26-5, and Tennessee consistently sits at 74%. If you use Torvik’s model, they’re at 78%, which would put them in an even more elite club. 5 seeds at 76% or above are a perfect 16-0.

(4) Oklahoma State over (13) Liberty. Let it be known that I despise this pick. Oklahoma State is the second-worst 4 seed of the last decade, with only 2015 Maryland sliding in ahead. Yes, they have Cade Cunningham, but they also have a negative turnover margin and an offense ranked below Liberty’s, the team they’re playing. They sit at just 74% to win, and as we’ve mentioned, 4 seeds at 75% or worse to win are just 13-12. But I can’t do it. Liberty ranks just 97th in KenPom, and 13 seeds ranked 90th or worse in KenPom are 3-26 since 2000. (All other 13 seeds are 13-38, which doesn’t sound great but is more than double the win rate of sub-90s.) I don’t have Oklahoma State going beyond the next round, anyway.

(11) Syracuse over (6) San Diego State. The toughest first-round pick resides right here. San Diego State sits precisely at 63%, which is my cutline for above/below 6 seed picks. They have unusually good Sweet Sixteen odds, sitting at 36.1%, above my 34% cutline there. They’d be favored in a second-round matchup with either potential opponent. And yet: Syracuse is #41, a very good team, and plenty dangerous for bracket purposes. I reserve the right to change this down the line.

EDIT: Indeed, I changed this from my original pick of San Diego State. Two models I look two for guidance in first-round games are Peter Keating’s Bracket Breakers and John Ezekowitz’s metrics. Keating has Syracuse at 45.7% to win; Ezekowitz, 50.1%. It’s worth the swing here.

(3) West Virginia over (14) Morehead State. West Virginia is the only 3 seed above 80%. Just hard to find the rationale for a Morehead win.

(10) Rutgers over (7) Clemson. Rutgers is the better team in every metric I use, but for some reason, the public heavily favors Clemson. A good chance to earn some Round of 64 value, as, again, the favorite in 7/10 games is 62-18.

(2) Houston over (15) Cleveland State. Houston is the highest-favored 2 seed in the field.

East Region

(16) Mount St. Mary’s over (16) Texas Southern. Not incredibly meaningful either way, but MSM is favored by a tiny amount most everywhere.

(11) UCLA over (11) Michigan State. If you have BYU in this round (as I currently do), you want Michigan State to win. BYU’s odds of seeing the Round of 32 fall a full 5% if they draw UCLA instead of Michigan State, and indeed, they’d fall below my 63% cutline for auto-advancing 6 seeds.

(1) Michigan over (16) Mount St. Mary’s. Michigan is a 98% likely winner against either team.

(9) St. Bonaventure over (8) LSU? This is the biggest coin-flip of the entire Round of 64. St. Bonaventure sits at 50.1% to win in my combined ratings. KenPom has them as the outright favorite, but Torvik likes LSU by a hair and the Tigers have been much better as of late. I ended up going with St. Bonaventure for one specific reason: no first-round pick would hold greater value if it worked out. Only 35% of ESPN users are picking the Bonnies, which means you’d gain .15 points on the average group with an SBU win.

(12) Georgetown over (5) Colorado. Colorado is the only 5 seed in the field that falls below 70% to win in the first round, and 5 seeds below 70% are 23-26. Worse yet, in the metrics combination, Colorado sits at 62%. 5 seeds with less than a two-thirds chance to win (66.7%) are just 16-23. This is the trendiest 12 seed upset pick, but for once, it’s a good pick. None of the other 5 seeds are very close to Colorado in terms of upset likelihood, and that’s before I adjusted Colorado’s numbers downward due to their unusually large home/away-from-home splits.

(4) Florida State over (13) UNC Greensboro. The lone 4 seed I actually feel safe picking. 4 seeds at 80% or higher to win in Round One are 31-1 (2011 Louisville the one loser), and Florida State is anywhere from 81-88% depending on which site you prefer.

(11) UCLA over (6) BYU. As mentioned above, BYU really wanted a Spartan victory. They would’ve been a 66.7% favorite against MSU in my metrics combination, and Torvik gives MSU just a 31% chance at an upset. However, a UCLA win massively complicates our plans. UCLA is 1.5 points better than Michigan State and drags BYU’s odds all the way down to 61.9%, where they’re below the 63% cutline for 6 seeds and would be in serious danger. Worse yet, a UCLA win of any kind almost certainly shrinks those odds further, as it would improve their adjusted efficiency margin. Go Bruins!

(14) Abilene Christian over (3) Texas. This pick makes my stomach turn, because I generally have a rule against picking 14 seeds. In my actual bracket, I don’t know if this is a trigger I’ll pull. However, I chose Texas as the 3 seed to go down instead of Arkansas for two key reasons: 1. KenPom has #26 Texas-#86 Abilene as the closest 3/14 game since West Virginia/Stephen F. Austin, the most recent 14 over 3 upset; 2. While I don’t advise doing this for the entire bracket, if you measure performance over the last 10 games and in 2021, Texas/Abilene is the closest of the 3/14 matchups.

Anecdotally, Texas could be a team to exploit for bracket group value, as they just won the Big 12 Tournament despite only playing two games (Kansas bowed out due to a COVID issue in the potential semifinal), and both wins were by six points or less in a sport where we’ve written for a long time about how close games have a substantial amount of luck involved. As I type, 93% of ESPN users have picked Texas in this game, and KenPom gives the Longhorns just a 75% chance of winning. If this doesn’t work out, it is what it is, but you could gain massive value by selecting a team that 71% of users are taking to the Sweet Sixteen and 33% to the Elite Eight.

One more thing: no game in the first round has a greater difference in turnover margin per 100 possessions than this one. Abilene averages a +7.3 turnover delta per 100; Texas sits below the Mendoza line at -2.3. That roughly converts to 6.7 extra possessions for Abilene in this game.

(7) Connecticut over (10) Maryland. UConn is ranked 16th, which makes them more like a 4/5 seed than your traditional 7. Maryland is a 10 seed ranked 31st, making them more like an 8/9. Essentially, Alabama got screwed, but you could benefit. Take UConn, the favorite everywhere.

(2) Alabama over (15) Iona. Iona doesn’t meet the metrics for a potential upset, but Alabama’s poor defensive rebounding could keep an Iona team ranked 32nd in OREB% in it for a half or so.

NEXT PAGE: Round of 32

One thought on “How stats and history would pick the 2021 NCAA Tournament

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s