Report2026-09-18

One pool or many? The question that decides how you should fill out 20 brackets

The short version. If all your brackets are on one leaderboard, build them to cover each other: 49% vs 35% to win a 100-person pool with 20 entries. If each bracket is in its own pool, don't hedge: independent brackets win 4.33 pools out of 20, hedged ones 3.84.

The question

If you enter more than one bracket, the most important choice isn't your champion. It's whether all your brackets compete on the same leaderboard, or each in its own pool. Those two need opposite strategies.

How we tested it

• Outcomes: 100,000 simulated 2026 tournaments from our model.
• Opponents: brackets drawn from how the public really picked in 2026 (ESPN's "Who Picked Whom").
• Two ways to fill 20 brackets: independent strong brackets, or our optimizer, which adds brackets one at a time to raise the expected score of your best one.
• 24 repeats with different random seeds.

Result 1: same leaderboard, hedge

Only your best bracket counts, so 20 copies of one strong bracket are worth exactly one bracket. If bracket 1 rides Duke and Duke loses in the Sweet 16, bracket 2 should be riding someone else.

The gap shows up from 2 brackets on, and grows with every bracket you add.

Result 2: separate pools, don't

Each bracket is judged alone. Hedging means that in most pools you're riding a champion the model likes less, and none of your other brackets is in that pool to make up for it.

Caveats

• Tournaments are drawn from our own model, so absolute win rates are optimistic. Real tournaments are messier than any model thinks. The comparison is fair, because both methods trust the model equally.
• The public field is built from per-team pick rates, so it's a bit less clustered on favourites than the real public.

What this means for you

That's why the optimizer asks for your goal first. With a single bracket, neither trick does anything: the free bracket builder is all you need.