A public test of a baseball prediction stack.

Closerbets is a one-person research project. It exists to answer one question, in public, with receipts: can a stack of sabermetric models — one per market — consistently beat MLB closing lines over a full season?

Every pick is time-stamped before the game starts. Every result is settled the next morning. Every closing line is captured at T-10 minutes by a Lambda function and compared against the line we actually got. Nothing is editable after the fact.

Tracking started April 9, 2026. 62 settled picks so far. The full record lives on /track-record, and the model itself is documented on /methodology.

A betting model is only as good as its forward-tested track record. Backtests are easy to overfit; private spreadsheets are easy to revise. The only honest evaluation is one where the predictions are published before the game and the results are settled in the open.

Publishing also doubles as the cheapest possible test of whether the work has commercial value. If organic traffic shows up, that's a signal. If it doesn't, that's also a signal.

Every number on this site is computed from a single source — the tracked_picks table in a Postgres database, exposed read-only via three views. Three places to look:

The site is informational. Picks aren't for sale, and there aren't plans to gate them. That said, two kinds of message are welcome:

One inbox for all of the above: hello@closerbets.com.

Closerbets publishes information about an MLB prediction model. It is not a betting service, not investment advice, and not a guarantee. Past performance — including the track record on this site — does not predict future results.

Sports betting is legal only in some jurisdictions and only for adults 21 and over. If you choose to bet, do so within your means. Resources for problem gambling are available at ncpgambling.org.