The Selective Upset Strategy

1The Spectrum of Selectivity

We tested four variations, from aggressive to surgical, across 21 tournaments (2003-2024, excluding 2020).

2The Hit Rate Sweet Spot

The more selective you are about which coaching upsets to pick, the higher your accuracy on those picks.

At the gap \u2265 5 threshold, you're calling upsets that actually happen more often than not. That's a coin flip that's slightly loaded in your favor — and in a pool of chalk-picking competitors, those 2-3 correct upsets per tournament are what separate you.

3Why R1-Only Beats All-Round Coaching

The point breakdown reveals exactly why constraining coaching to Round 1 works.

All the damage (and all the upside) is concentrated in Round 1. Rounds 2-6 are identical because both strategies revert to chalk after R1. The R1-only approach gives up 250 points across 21 seasons in the round of 64 — about 12 points per season. But the gap \u2265 5 variant gives up almost nothing because it's so selective.

4Season-by-Season Results

The R1 Coaching (gap \u2265 5) strategy across every tournament since 2003.

The strategy wins 5 seasons, loses 8, and ties 8. When it wins, it often wins by 20-30 points. When it loses, it typically loses by just 10. The asymmetry is favorable: your upside is larger than your downside because you're only making a handful of flips.

5The Bracket Pool Argument

The gap \u2265 5 R1 coaching strategy threads this needle perfectly:

Why it works for pools

Minimal downside: You're changing only 2-3 picks per tournament, all worth just 10 points each. Your worst-case loss vs chalk is ~30 points — roughly one Sweet 16 game. Meanwhile, chalk pickers routinely lose 80-160 points on a single bad Elite 8 or Final Four pick.

Real differentiation: When you hit 2 of 3 coaching upsets in R1, you gain 10-30 points over the chalk-heavy field. In a tight pool, that's often the margin of victory.

52% hit rate: You're not gambling. You're picking upsets that historical data says happen more often than not. This isn't a coin flip — it's a slightly loaded die.

Low correlation with the field: Nobody else is using coaching experience as their upset heuristic. They're picking based on mascots, conference loyalty, or which 12-seed "looks dangerous." Your edge is systematic and uncorrelated.

6The Economics Lens

In finance, this maps directly to the concept of tracking error budgeting. You have a limited amount of "deviation from benchmark" you can afford before the cost of being wrong overwhelms the benefit of being right. The gap \u2265 5, R1-only approach has a tracking error of roughly 12 points per season — tight enough to stay competitive, loose enough to generate alpha when conditions are right.

The deeper lesson: edge and implementation are separate problems. We proved coaching experience creates a real predictive edge (52% on targeted upsets). But the last report showed that implementing it naively — throughout the bracket — destroys value. This report shows that implementing it surgically — R1 only, high-conviction only — preserves the edge while managing the risk.

It's the difference between a research paper and a trading strategy. The research is the same. The execution is everything.