We have MIL at 36.5% to win in San Diego. The market is paying +228.

What we found

The biggest gap between our simulation and a posted number today is Milwaukee's moneyline at San Diego. The market is offering +228 on MIL. Our plate-appearance Monte Carlo puts Milwaukee's win probability at 36.5%. That combination implies an expected value of +19.83% per unit staked.

To be explicit about what that does and does not say: we think Milwaukee loses this game roughly two times out of three. This is not a call that Milwaukee wins. It is a claim that +228 is too long a price for a team we model as a 36.5% underdog, and that the gap between those two numbers is large enough to be worth publishing.

That is the entire nature of underdog value, and it is why this kind of position is miserable to hold in public. The single most likely outcome — by a wide margin — is that we are on the losing side tonight and the market looks correct. One game carries essentially no information about whether the edge is real. A 36.5% claim can only be evaluated across hundreds of similar claims, not one.

Three other disagreements cleared our threshold on the same slate:

- AZ ML (COL @ AZ) — 19.82% EV - SF ML (HOU @ SF) — 17.64% EV - NYM ML (NYM @ ATL) — 14.58% EV

That is four positions, and the sample size caveat is the point: four is not a track record. It is four data points that will be logged, graded, and folded into the running record whether they win or lose.

What it means

An edge this large is a flag as much as a finding. When a model disagrees with a liquid market by nearly twenty points of EV, the two candidate explanations are that the market is slow and that the model is wrong — and the second is the higher prior. Large edges concentrated on longshots are the classic signature of a model that has mispriced tail outcomes, not one that has found free money.

So we are publishing the claim in the form that lets you falsify it. The number is 36.5%. The price is +228. If our probabilities are honest, positions taken at this kind of gap should compound over a large sample of them. If our probabilities are inflated on underdogs specifically, this exact bucket — big-EV longshots — is where that failure shows up first, and it will show up as a losing bucket in our own record long before it shows up in any single result.

We would rather be measured on that bucket than on tonight.

Nothing here is a prediction of what any player or team will do. It is a probability and a price, and the probability is well under half.

What would change our mind: if our big-EV underdog positions grade out below breakeven across a meaningful sample, the problem is our tail pricing, not the market's — and we recalibrate rather than defend the model. 21+ where legal.