march madness w/ the boys
All Articles

Does Tournament Experience Matter?

Analyzing 40 years of NCAA March Madness coaching data (1985-2024)

634 coaches • 2,557 tournament appearances • 2,518 games

65.8%
Win rate for the more experienced coach in head-to-head matchups
4.3x
Veterans (15+ apps) reach the Sweet 16 at 4.3x the rate of first-timers
2.22
Avg tournament wins for 15-19 app coaches vs 0.52 for first-timers
~50%
of the experience effect persists even after controlling for seed

Finding 1: Raw Experience Strongly Predicts Success

Coaches with 15-19 prior tournament appearances average 4x the wins of first-timers

The raw numbers are striking. A coach making their first NCAA Tournament appearance averages just 0.52 wins and only reaches the Sweet 16 about 4% of the time. That climbs steadily with experience, peaking at the 15-19 appearance bracket where coaches average 2.22 wins and reach the Sweet 16 nearly 40% of the time.

First appearance
0.52 wins
2-4 prior apps
0.86 wins
5-9 prior apps
1.27 wins
10-14 prior apps
1.60 wins
15-19 prior apps
2.22 wins
20+ prior apps
1.86 wins
Diminishing returns after 20+ appearances? The 20+ group dips slightly to 1.86 wins. Possible explanations: aging coaches past their peak, smaller sample (n=71), or regression to the mean. But 20+ coaches still massively outperform first-timers.

Finding 2: Experience Still Matters After Controlling for Seed

Among top 4 seeds, veteran coaches reach the Sweet 16 at 49% vs 31% for first-timers

The obvious objection is that experienced coaches simply have better teams (lower seeds). So we controlled for that. Among seeds 1-4 specifically, the effect persists clearly:

Experience (Seeds 1-4 only)CountAvg WinsWin 1+ GameSweet 16+
First-timers451.8484%31%
2-5 prior apps2042.1088%34%
5-14 prior apps2732.2989%38%
15+ apps (veterans)1022.6692%49%

Even when the seed is the same, veteran coaches squeeze out roughly half a win more on average. This suggests experience provides an independent edge beyond just talent — likely through game preparation, timeout management, halftime adjustments, and handling pressure moments.

The mid-seed plateau: Interestingly, for seeds 5-8 and 9-12, the experience effect is weaker and noisier. The biggest edge shows up at the extremes — top seeds where veterans maximize their advantage, and the 15+ veteran underdogs in the 9-12 range who reach the Sweet 16 at 13% vs 5% for first-timers.

Finding 3: Head-to-Head, the More Experienced Coach Usually Wins

The coach with more tournament appearances wins 65.8% of matchups
65.8%
More experienced coach wins (1,444 games)
vs
34.2%
Less experienced coach wins (752 games)

Across 2,196 tournament games where the two coaches had different experience levels, the more experienced coach won nearly two-thirds of the time. This is a large sample and a large effect — though it's partially confounded by seed (experienced coaches tend to have better-seeded teams).

Finding 4: Prior Wins Matter More Than Prior Appearances

Coaches with 30+ prior tournament wins reach the Sweet 16 at 40% vs 4% for those with zero

Simply showing up to the tournament isn't the whole story. Coaches who have actually won tournament games perform dramatically better. This makes intuitive sense — winning in March requires specific skills (adjusting to unfamiliar opponents, managing single-elimination pressure) that only come from doing it successfully.

0 prior wins
4.3%
1-5 prior wins
11.0%
6-15 prior wins
19.6%
16-30 prior wins
27.0%
30+ prior wins
40.3%

Sweet 16 appearance rate by coach's prior tournament win total

Finding 5: Cinderella Runs CAN Happen with First-Time Coaches

Some of the most memorable runs came from coaches in their first or second tournament

While experience helps on average, the exceptions are some of March Madness's greatest stories. These coaches made deep runs as low seeds with little or no prior tournament experience:

#11
VCU (2011) — Shaka Smart
First tournament appearance as head coach
Final Four
#9
Florida Atlantic (2023) — Dusty May
First tournament appearance as head coach
Final Four
#11
George Mason (2006) — Jim Larranaga
2 prior tournament appearances
Final Four
#11
Loyola Chicago (2018) — Porter Moser
First tournament appearance as head coach
Final Four
#15
St. Peter's (2022) — Shaheen Holloway
First tournament appearance as head coach
Elite Eight
#10
Gonzaga (1999) — Dan Monson
First tournament appearance as head coach
Elite Eight
The pattern: Cinderella coaches tend to be young, hungry, and have nothing to lose. They often get hired away to bigger programs immediately after their run (Shaka Smart to Texas, Dusty May to Michigan, Shaheen Holloway to Seton Hall). The lack of experience may actually help — opponents have less film and scouting data on their systems.

The Mount Rushmore of March

Top 15 coaches by total tournament appearances, 1985-2024
CoachAppearancesTournament WinsChampionshipsWin Rate
Mike Krzyzewski3510152.89/app
Roy Williams307932.63/app
Jim Boeheim295711.97/app
Rick Barnes283001.07/app
Tom Izzo265612.15/app
Bob Huggins263401.31/app
Bill Self255722.28/app
Mark Few244301.79/app
Lute Olson233911.70/app
John Calipari235712.48/app
Rick Pitino225422.45/app
Jim Calhoun204832.40/app
Lon Kruger202201.10/app
Bob Knight192411.26/app
Kelvin Sampson192601.37/app

Coach K's 101 tournament wins across 35 appearances is absurd — nearly 3 wins per trip, meaning his average tournament ended in the Sweet 16.

The Bottom Line for Our Model

Coach tournament experience should be a feature in our predictive model. Prior tournament wins and prior tournament appearances both show meaningful signal, even after controlling for seed. The effect is strongest for top seeds (where the gap between a veteran and first-timer is about half a win per tournament) and weakest for mid-seeds.

Recommended features to include: prior tournament appearances, prior tournament wins, prior deep runs (Sweet 16+), and whether the coach has ever won a championship. Prior wins appear more predictive than raw appearances.

One caveat: coach experience is likely correlated with program strength and recruiting, so its independent contribution in a multivariate model may be smaller than these bivariate numbers suggest. We'll test that in the next phase.