Instead, I want to give you three clear upset signals to consider when looking for upsets to pick in your own bracket.
Upset signal 1: Underdogs led by coaches with previous tournament successHoops pundits often discuss the importance of good coaching in the tournament, but usually when referring to coaches of powerhouse, blue-blood programs like Kentucky, North Carolina, and Michigan State.
But it turns out a coach’s record of success in the tournament can make a big difference for underdog teams as well.
When I looked at several metrics of coaching success for underdogs, one factor stood out: the number of times the underdog’s coach has previously been to the 4th round of the tournament, aka the “Elite 8".
Underdog coaches with Elite 8 experience are more likely to lead their team to a winOn the left I show the “upset rate”, defined as the percentage of games resulting in an upset.
Underdogs led by a coach with no Elite 8 visits on their resume only won around 16% of the time, but underdogs led by “Elite 8” coaches fared much better, hitting an upset rate close to 40%.
By definition, every coach with Elite 8 experience has previously led a run of at least 3 wins in the tournament.
These runs may help by giving the coach valuable experience, or could just represent evidence of the coach’s skill at managing intense in-game pressure, game-planning for the next opponent, or possibly both.
Below I show all of this year’s potential underdog teams (seeds 5–16) who have Elite 8 coaches.
For each team, I show the round and opponent for the game where they would first be considered an underdog.
Only 2 teams meet this criteria in round 1, but several potential upsets await in round 2 and beyond.
Underdogs led by coaches with previous Elite 8 experienceNotice that regardless who wins Baylor-Syracuse in round 1, Gonzaga will face an Elite 8 coach in Round 2 in either Scott Drew or Jim Boeheim.
Both are tough round 2 match-ups for #1 seed Gonzaga, meaning it could be risky to pick them to escape this region.
Syracuse was in a similar position last year as a 10-seed, when my model correctly predicted they would upset 2-seed Michigan State in round 2.
Upset signal 2: Mid-major underdogs with a strong winning record against a difficult scheduleFor the selection committee, strength of schedule is a big factor in deciding which at-large teams to invite.
As it turns out, strength of schedule is also a useful factor in predicting underdog success during the tournament.
This is particularly true for “mid-major” teams, which are those outside the “Power 5” conferences (ACC, SEC, Big 12, Big 10, PAC-12).
When I looked at mid-major teams specifically, both strength of schedule and high winning rates stood out as potential factors in predicting upsets.
Among mid-majors, underdogs with both strong schedules and winning records are most likely to pull an upsetFor all tournament games involving underdog mid-major teams, I show the underdog’s total number of pre-tournament wins on the vertical axis.
The horizontal axis shows the team’s pre-tournament strength of schedule, with higher values indicating the team played more difficult opponents.
The average mid-major strength of schedule is shown in blue.
Notice that the upsets (colored green) are clearly clustered in the upper-right.
This means upsets tend to happen when underdogs have above-average strength of schedule and at least 20 wins.
To create more specific guidance using these two factors, I grouped the mid-major underdogs by schedule strength (average or less, over average), and plotted each group’s upset rates separately for 3 categories of total wins: less than 18, 18–23, and over 23 wins.
The data here tell an interesting story.
Mid-major teams with a strong win record are indeed more likely to pull off an upset, as shown by the overall increase in upset rates from left (less than 18 wins) to right (over 23 wins) for both mid-major groups.
However, for mid-majors with a tougher schedule (green/blue line), the increase is drastically larger.
This pattern lines up with common sense: mid-majors who prove their worth by piling up wins against more difficult opponents are also those who are more likely to pull an upset in the Big Dance.
The chart above shows that upsets were most common for mid-major underdogs with over 18 wins AND a difficult schedule.
Below I show the 11 mid-major teams who meet both criteria this year.
For each team, I also show either their round 1 opponent (for seeds 11–16) or their projected future match-up (for teams seeded 5–10).
Mid-major underdogs with at least 18 wins and a difficult season scheduleTemple and St.
Mary’s look like strong upset contenders in Round 1, although Temple must first get past another mid-major (Belmont) in the First Four.
In Round 2, UCF against Duke and Cincinnati against Tennessee both stand out as mid-majors with the win record & schedule credentials to spring the upset.
Connecting this data with the coaching table above, I also notice that Michigan is guaranteed to face either a tough mid-major (Nevada) or an experienced coach (Florida) in round 2, with yet another tough mid-major (Buffalo) or the most efficient 3-seed in the country (Texas Tech) waiting in round 3.
Michigan is a great team, but strings of difficult match-ups like these are landmines that can bust brackets, and great teams take an early exit every year, including one-third of #2 seeds in round 2.
The Wolverines may be one team to consider removing early in your bracket.
Upset signal 3: Elite underdogs in a game close to homeBecause every tournament game is played at a neutral site, no teams have a true home court advantage.
However, underdogs usually end up playing further from home than the favorite, which may put them at a disadvantage, possibly due to longer travel and reduced fan attendance at the game.
But sometimes, underdogs get lucky, and end traveling a similar distance (or even less) than the favorite.
And when those teams happen to be very skilled underdogs, the upset watch is on alert.
I use the term “elite” underdogs to describe teams with a crucial upset signal that I discussed in a previous post: a Kenpom adjusted efficiency margin over 10.
Only 40% of underdogs hit this benchmark, but these teams upset the favorites around 30% of the time, compared to only 10% for other underdog teams.
As a single rule for picking upsets, you could do much worse…but I think we can also do much better.
Elite underdogs (those with Kenpom efficiency margin over 10) who play close to home upset the favorite 40% of the timeAmong the “elite” class of underdogs, playing close to home is a solid upset signal.
As I show on the left, “elite” underdogs playing less than 400 miles from campus have upset rates around 40%, while the rest of the elites have upset rates below 30%.
Elite underdogs only play this close to campus 20% of the time, but when they do, it raises the odds of an upset significantly.
Below I show the relevant games for this year’s tournament.
Cincinnati (who we already know has the upset signal for win record and strength of schedule) pops up again for their round 2 match-up with Tennessee in Columbus, Ohio.
I also notice that for Duke’s round 2 game in South Carolina, both of their potential opponents (VCU and UCF) meet the travel criteria.
Crazy enough, both teams also had the win record and strength of schedule credentials discussed above.
Elite underdogs who will travel less than 400 miles to their gameDuke is the #1 overall seed, with one of the most talented starting lineups we’ve ever seen.
Would I pick Duke to lose to UCF or VCU in a single-game format?.No, I would not.
But in bracket pools, minimizing risk is a big part of success, and that includes recognizing which of the top teams have more difficult paths to the Final Four.
So you might think twice before automatically giving Duke the title in your bracket.
Closing RemarksEven the underdogs who meet these upset signals lose more than 50% of the time, and there are many other factors that influence the outcome of each game in the NCAA tournament.
So how should you use this information to pick upsets?My advice would be to use the signals as tools, not hard rules.
If you have a hunch on an underdog team, check them against these signals, and see if they pass the test.
Try coming up with your own signals, and see how they do this year.
In the future, I also plan to release full upset predictions with confidence metrics and other data, so follow me here on Medium or Twitter @bracket_vision if you want to see more.