College Basketball's NCAA tournament heads into its final 3 games this weekend and Monday, known as the Final Four. Many have said that this year's tournament might have been the most boring do to the lack of upsets in the early round games, but those people are wrong. 2007 was the most boring. And this year is only the second most boring - although "boring" is the wrong word. It's not boring, it just hasn't had the number of upsets that get everyone excited when they're just watching the bracket get filled in instead of actually watching the games.
The other major thing missing from the NCAA tournament is someone on the internet doing ridiculous things with the available stats from it. So, as a public service, I've come up with a way to rank the relative upset-y-ness of each years tournament since the expansion to 64 teams in 1985. (For convenience, we'll call the 64 and 65 team bracket years the Modern NCAA tournament - obviously a tournament existed before then, but it was a different beast.) Much like popular baseball and basketball metrics, this new Tournament Upset Metric will use easily available numbers from each year's tournament, so anyone can play along at home.
The TUMHere's the Tournament Upset Metric formula for each game:
TUM = (Ws - Ls - 1) * (round + 1)/2
where Ws is the winner's seed number in the tournament bracket and Ls is the losers seed number. All negative values are rounded to 0 because we're only interested in UPSETS.
The first part of the formula simply measures the relative seeds of the two teams - but teams within one seed of each other really don't count as upsets, so we subtract one from the difference. 9 seeds beating 8s or 5 seeds beating 4s don't count as upsets, but a 10 over a 7 in the first round should get 2 points, and an 11 over a 6 is 4. The next part is a factor based on the tournament round, from 1 for the first round to 6 for the national finals. First round upsets are always fun, but the formula should give more significance to the later upsets. So the seed difference is multiplied by 1 for the first round, 1.5 for the 2nd round, 2 for the "sweet sixteen" round... up to 3.5 for final game.
BEST UPSETSThis formula yields these as the top 5 upsets in modern tournament history:
1) (tied) - #11 George Mason 86, #1 Connecticut 84 (in OT) in the 2006 "Elite Eight" round, also know as round 4 - this gets a 22.5 on the TUM meter, and also happens to be my favorite college game in recent memory.
1) (tied) - #11 LSU 59, #1 Kentucky 57 in the 1986 Elite Eight - also a 22.5
3 - #8 Villanova 66, #1 Georgetown 64 in the 1985 Finals - a 21 TUM rating - It's an excellent argument that this should be the top upset ever, but on paper it doesn't quite edge out those 11 vs 1 elite 8 games. But, this gets remembered as an even bigger upset than the seeding implies because Georgetown was so highly thought of that year, which doesn't get fully captured by its #1 seed. There's always room for improvement in a formula though.
4 - #11 LSU 70, #2 Georgia Tech 64 in the 1986 sweet 16 round - a 16 TUM upset
5 - #8 Villanova 52, #2 Memphis State 45 in the semi-finals - a 15 TUM.
BEST TOURNAMENTS EVERNow, once we have rated each upset, for it's relative upsetingness, we can add all the scores together and look at the tournaments as a whole, and rank them by year:
| Year | Total TUM | TUM per Game | Total Upset Games |
| 1986 | 129 | 2.05 | 14 |
| 1985 | 125 | 1.98 | 14 |
| 2006 | 112.5 | 1.79 | 17 |
| 2002 | 109.5 | 1.74 | 16 |
| 1990 | 108 | 1.71 | 17 |
| 1999 | 103.5 | 1.64 | 16 |
| 1987 | 91 | 1.44 | 13 |
| 1988 | 89.5 | 1.42 | 13 |
| 2001 | 86 | 1.37 | 15 |
| 2000 | 85 | 1.35 | 16 |
| 1997 | 79 | 1.25 | 13 |
| 2005 | 78 | 1.24 | 13 |
| 1991 | 69.5 | 1.1 | 12 |
| 1992 | 66.5 | 1.06 | 11 |
| 1998 | 66 | 1.05 | 13 |
| 1996 | 64 | 1.02 | 12 |
| 2004 | 63 | 1 | 10 |
| 2008 | 62.5 | 0.99 | 9 |
| 1994 | 60.5 | 0.96 | 10 |
| 2003 | 60 | 0.95 | 12 |
| 1989 | 55 | 0.87 | 11 |
| 1995 | 53.5 | 0.85 | 10 |
| 1993 | 48 | 0.76 | 8 |
| 2009* | 34.5 | 0.55 | 8 |
| 2007 | 17 | 0.27 | 4 |
*-
in progress UPDATE: final now, with no change.
As you can see, the first two years of the modern tournament were the most exciting, in terms of upsets, with the Villanova championship in '85 and LSU's final 4 run in '86. Aside from Villanova, there were also two 11 seeds and a 12 who made it to the sweet 16 round that year (which required upset wins in two rounds). The '86 tournament also included #12 Depaul and #14 Cleveland State making the sweet 16. That was one of only 2 #14 seeds who have ever made the sweet 16.
But, it's not just a case of the upsets being more likely in the '80s when maybe the tournament committee wasn't as good at ranking everyone, because we have the 2006 tournament, with George Mason's final 4 run, and #13 Bradley making the sweet 16, along with a couple of 7 seeds.
Also, this year is only the 2nd worst ever - 2007 only had 4 upsets in the whole tournament. This year we've had 8 already, and Villanova could bring that up to 10 if they beat UNC and beat Uconn in the finals (should they make it). That would still only bring the 2009 total TUM up to 42. 1993 got a surprisingly low score, despite that being the year #15 Santa Clara beat #2 Arizona in the first round. There were only 7 other upsets that year, with just one in the sweet 16 round, and no upsets in later rounds.
Best Cinderella'sAlso, since the media loves to talk about Cinderella's of the tournament, we can use this metric to rank the top Cinderella tournament runs by adding up their TUM scores for the tournament:
1985 Villanova - 61.5 TUM (won the championship as a #8, beating two #1s and 2 #2s)
1986 LSU - 53 TUM (made final 4 as an #1, beating a 1, 2, 3 and 6)
2006 George Mason - 43 TUM (made final 4 as a #11, beating a 1, 3, 6 and 7)
1988 Kansas - 25.5 TUM (won the finals as a #6, but beat a #7, #14, and #11)
1987 LSU, 2002 Kent State, 2008 Davidson - 24.5 (all three were 10 seeds who made it to the elite 8)
That '88 Kansas team would have ranked much higher, but they faced a #14 seed in the 2nd round, and a #11 in the 3rd round. Obviously that helped them along toward the final, and they don't get any TUM points for those games. Normally it would be an upset for a #6 seed to win a sweet 16 game, but not when they beat a #11.
Most Upset TeamsWe can also now tell which programs have had the most upset riddled tournament histories. Obviously, they all have to be pretty good programs or they wouldn't be in positions to be upset so often, but if we add up the total TUM in games where each team lost in the tournament, we get the following top 10 most upset teams:
| Team | Total
Tum Against | Modern
Tournament
Appearances | Upset
Losses |
| Oklahoma | 82 | 20 | 12 |
| Duke | 65 | 21 | 10 |
| Kansas | 64.5 | 22 | 9 |
| Kentucky | 62.5 | 19 | 8 |
| North Carolina | 61.5 | 20 | 8 |
| Arizona | 57.5 | 24 | 8 |
| Syracuse | 52.5 | 19 | 7 |
| Connecticut | 49.5 | 13 | 5 |
| Georgetown | 44.5 | 16 | 4 |
| UCLA | 44.5 | 19 | 6 |
Oklahoma lost the '88 finals as a 1 seed, to #6 Kansas, but most of their trouble has come in the first two rounds - they've lost 3 times as 4 seeds to 13 seeds in the first round, and once more as a 4 seed to a 12 seed in round 2. Duke, on the other hand, also lost to the '88 Kansas team, in the semi-finals, but has a history of getting upset in the 2nd and 3rd round. They've been upset 4 times in the 2nd and 4 times in the 3rd round, and of course that 1st round lost to VCU in 2007 - although Duke was only a #6 that year. Duke holds the modern tournament record for #1 seeds losing in the sweet 16, with 4. Kansas has done that twice, but no one else has more than 1 sweet 16 lose as a #1.
TUM CriticismsThe obvious flaw in this is that we're only measuring upsets based on the tournament seeding created by the guys who put together the NCAA bracket each year. We know that sometimes they need to bump a team up or down a seed to get the bracket to work out the way they want, like to keep a top team playing closer to home, or to avoid having more than 2 teams from any conference in the same 16 team region. Also, there are some #1 seeds that transcend mere #1 seeds, like that '85 Georgetown team with Patrick Ewing, the '2006 Uconn team which was loaded with NBA lottery picks, and the '91 UNLV team that lost to Duke in the final 4. Many people consider that UNLV-Duke game a big upset, since UNLV was having an undefeated season and rolling over everyone, but Duke was a 2 seed so that game doesn't even score a TUM. But, most stats simply work with the measurements that are easily available, and don't take situation or other intangibles into account, so this works pretty well on the whole.
What does this have to do with baseball or HBPs?This has nothing to do with baseball or people getting hit by pitches, but if you insist... the average TUM score for the modern NCAA tournament is 78.5. In years when the tournament has a below average TUM, batters get hit 8% more often than in years when the TUM is above average. Which means this could be a good year for HBPs.
