MLB Picks

Nearly every "MLB picks" site works the same way: it hands you a lock, tells you it's a five-star play, and quietly forgets the ones that lost. There's no ledger, no accountability, and no reason to believe the record is anything but the wins they chose to remember. We built MatchWiz to do the opposite. Our picks come from one model, they're graded against the real result whether they hit or miss, and the entire track record is public and rolls back day by day. If that sounds less exciting than a guaranteed winner, good — that's the point.

What a "pick" means here

A pick, to us, isn't a hunch or a lean someone felt good about. It's a spot where our model's number and the market's number disagree by enough that the disagreement has been worth acting on historically. We call those the Plays of the Day: the handful of matchups each slate where the projection diverges most sharply from the posted line. Everything starts from the model, and the model doesn't have favorites — it scores a superstar and a September call-up with the exact same math, so what you're getting is a read on the matchup, not on a name or a narrative.

How the model gets there

Under the hood it's a matchup model. For a hitter, it weighs how he's actually performed against the kind of pitcher he's facing — handedness matters a lot — against what that starter and the bullpen behind him tend to allow, then adjusts for how much he's likely to play, the ballpark, the weather, and how he's swinging lately. Pitcher boards run the same logic in reverse: how the starter projects against the specific lineup he's drawing. Every player on the slate ends up with one comparable number.

A projection on its own is just trivia, though. It only becomes a pick when it meets a price. So wherever a book posts a line, we set our number beside it and measure the gap — and the size of that gap, not the name attached to it, is what decides whether something rises to a Play of the Day. The thresholds aren't guesses; they come from backtesting where the disagreement has actually paid.

Where the edge actually is — and where it isn't

This is the part most sites will never tell you: most MLB markets are efficient, and we won't pretend otherwise. We've run a full-season backtest on every market we cover, graded in units, not hit rate. The moneyline, game totals, home runs, hits, RBI, runs, total bases — the books price all of them well enough that even a sharp model can't find a repeatable edge. So we publish those as graded transparency: our number next to the market's, scored honestly, but with no lean attached. They're there for you to use, not for us to oversell.

The edges we have found and can defend with a track record are narrow and specific: strikeout-prop gaps — where a big projection-vs-line divergence has held up across a full season and a split-half check — and team-total leans, the softer run-total market where our strongest disagreements have paid. Those are the only markets we surface as plays. Two honest edges beat fifty manufactured ones, and if a market ever earns a durable edge it'll show up here on its own.

The receipts

Anybody can claim a hit rate. The thing that should make you trust — or distrust — a source is whether you can check it. So we grade every projection against the actual box score, at the line the market really offered, and we lock the score for a game at first pitch so no number can be nudged after the fact to look smarter. The full archive is public: pick any past date on any board and read exactly what we said and exactly how it landed.

The season hit rate and ROI for the Plays of the Day sit right next to the plays and on our how-it-works page, updated as games settle. It won't always be green — no honest record is — but it's the same ledger that tunes the model, not a marketing number we picked out. When we're on a cold stretch, you'll see it. That's the deal.

Why we won't sell you a lock

A single MLB game is close to a coin flip, and a single prop is barely better. Anyone promising you a guaranteed winner is either fooling you or fooling themselves. What actually compounds over a season isn't any one call — it's consistently taking spots where the price is a little off, and knowing the difference between a real edge and variance dressed up as one. That's a boring, unglamorous truth, and it's the entire business. We'd rather be the site that tells it to you straight and shows its work than the one with the loudest guarantee and the quietest track record.

How to actually use it

Start with the Plays of the Day — that's the shortlist, the spots where the model most disagrees with the market, with the record attached. From there, the full boards let you go deeper: the strikeout board for pitcher props, the Fantasy Points board and the DFS lineup optimizer if you play daily fantasy, and moneyline & run totals for the team-level model. Every board carries its own graded, roll-back-able record, so nothing asks you to take our word for it.

The traps to avoid

• Judging a pick by one result. A good process loses plenty of individual bets; a bad one wins plenty. One night tells you almost nothing — the sample does. That's why we show the sample.
• Chasing the loudest source. Confidence is free. A public, graded record is not. If a pick doesn't come with a way to check the track record, treat the confidence as decoration.
• Betting the efficient markets anyway. If the number says a market is priced right, forcing action there is just paying the vig for entertainment. The edge is in the few spots that aren't efficient — not everywhere.
• Confusing a projection with a promise. Our number is our best estimate, published so you can weigh it — not a guarantee, and never betting advice.

See today's reads

The whole thing is free, and all of it is graded in the open. Check the day's Plays of the Day, dig into any board, or read exactly how the model works and how it's doing. Then decide for yourself whether the numbers are worth trusting — which is the only thing we'd ever ask you to do.