Every pick site on the internet has the same trick: they show you the winners, count the stars, and forget the rest. No ledger. No way to check. The track record is whatever they chose to share. We built MatchWiz to work the other way. Every play we surface comes from one model, graded against the real box score at the line that was actually offered — and the score for every game locks at first pitch so nothing gets quietly revised. When we're running cold, you'll see it right there in the dashboard above. That's not a disclaimer bolted on for legal cover. It's the whole premise.
What a Play of the Day actually is
The model scores every hitter and pitcher on the slate from the matchup in front of them. Every player ends up with one projection — our best estimate of how that specific matchup should go. Most of those numbers land near what the book is offering. A starter projected for 5.2 strikeouts against a 5-K line isn't interesting; the model and the market agree. That's not a play.
A Play of the Day is the spot where model and market disagree by the widest, best-supported margin on a given board. One per market. The disagreement has to be large enough that a full-season backtest confirms it's been worth acting on — not a gut read, not a gap someone felt good about. The gap is the signal. The size of the gap is the edge.
Right now, two markets earn a play: strikeout props and team-total leans. That's it. Not because the others aren't scored — we score everything — but because those are the only two where a full-season, split-half-validated backtest shows a repeatable edge. Every other market we cover gets our number beside the market's, graded in the open, with no lean attached. You can see what the model says on everything. We just won't pretend an edge exists where the data says it doesn't.
How the model surfaces one
There are two gates between a number and a posted play.
Gate one is the math. For a strikeout prop, the model builds each starter's number from his own rates — his whiff rate, adjusted for the lefty/righty mix he's facing, blended across recent form and the full season — then runs it at the actual lineup: all nine hitters, weighted by where they bat and how they handle that kind of arm. It's not his season K/9 nudged up or down. It's a number built for tonight, this group, this park. Then we measure the gap to the book's line. The book mostly prices a K line off the name and the season total, so our number catches the spots it misses — a contact-heavy lineup sitting there priced like a whiff group, or an ace's line puffed up past what the matchup backs. When that gap is big enough to have paid in the backtest, it's the play. Team totals work the same way: our run projection for a side against the book's posted total, over a backtested edge line.
Gate two is the lineup card. A play surfaces only once the official lineup is confirmed for that game. Before lineups, the projection can still move — a key hitter scratched changes the whole math. We won't post a play that could flip by the time you see it. So the board fills as each lineup locks, usually mid-morning. Some late-announcing games push their plays closer to first pitch.
How it's graded and locked
The line you see when a play surfaces is the line it grades against — frozen the moment it posts. We don't grade against the closing line or a better number that came later. If a play surfaced at strikeouts over 6.5 at −115, that's the play in the ledger. That's the price it grades against. You could've gotten that number.
Each game's score locks at first pitch. Once the umpire calls play ball, nothing changes — not the line, not the projection, not the lean. A play that looked shaky and turned around doesn't retroactively become a confident call. A play that fell apart doesn't get quietly buried. The grade is the grade.
Every play lives in a posted_plays ledger: the market, the line, our projection, the gap, and the real-time odds we captured. The season hit rate and units in the dashboard above are graded from that frozen record — not a marketing version, not curated highlights. The same ledger that tunes the model each week. For the full methodology, see how MatchWiz works.
The full archive is public. Click any past date. Read exactly what we called and what happened.
The receipts
The dashboard at the top of this page shows the live record — every Play of the Day board and every team-total lean, graded from the posted ledger. Here's what you're reading:
- Hit rate — the raw fraction of plays that landed on the right side. The honest baseline.
- Flat 1u — one unit staked per play, regardless of edge size. What a simple flat-bet approach returned.
- ★ Recommended units — each play sized 1–3u by the gap: a 2-K strikeout divergence gets more weight than a barely-qualifying 0.8-K gap, because historically the bigger gaps have outperformed. This is how we'd actually size them, and the ★ column in the dashboard reflects that.
If we're on a cold stretch, those numbers will show it. That's the deal.
Why only these two markets
Most sites surface everything. We don't — and here's why.
Every market we cover got a full-season, out-of-sample grade in units: moneyline, strikeouts, home runs, hits, RBI, runs, total bases, game totals, and team totals. Most of them broke even against the vig, or lost. The books price these markets well. Even a well-tuned model bleeds vig on a sharp market when the edge isn't real.
Strikeout gaps and team-total leans are the two markets where that test came back positive and held up on a split-half check. Not because we like them more. Because the data earned them. If another market ever builds a sustained edge, it surfaces here automatically — the backtest decides, not a person. And if one of the two we surface ever stops earning, it disappears the same way.
We'd rather post two honest edges than fifty manufactured ones.
How to actually use it
The plays are at the top of this page, already filtered and labeled — the lean, the line, our projection, and the gap (Δ). The season hit rate and units for that board sit in each section header. Start there; you don't have to hunt.
The gap is the key number. A bigger gap means the model and market disagree more, and that's where the historical edge concentrates. The ★ recommended-units column in the dashboard scales each play by that gap — 1–3 units — so you can see what proportional sizing looks like versus a flat bet. Use it as a guide to how much conviction the model has on a call, not as a dollar amount.
For pitcher props, dig into the full strikeout board — it shows every starter's projection versus the posted line, not just the Play of the Day. The moneyline and team totals board has the full run-total model, and the NRFI board shows the first-inning model for each starter. Every board carries its own graded, roll-back-able record — so you can check whichever matters to you, on any past date.
The traps
- Judging a cold night as a trend. A good process loses individual bets all the time. One night proves nothing. The season sample is the only number that tells you anything about the process — and it's right there in the dashboard.
- Pressing markets without an edge. No play on the board today? The model is telling you there's nothing to back. There's no edge to force. Paying vig on an efficient market isn't research; it's entertainment at your expense.
- Looking for confirmation instead of information. If you already want to bet an arm and you're hoping the board backs you up — look carefully. If the board disagrees, that's the signal. The model doesn't know who you want to win, and it doesn't care.
- Confusing track record with certainty. The season record is the best evidence the edge is real. It's not a guarantee any single play lands. Baseball is a high-variance sport. Every play on this board has a real backtested edge — and any one of them can still lose.
Plays of the Day — FAQ
What is a Play of the Day?
One spot per board — the single place our MLB model disagrees most with the sportsbook's line, by a margin that actually paid in our season backtest. One per market, and only once the odds are up and the lineup's final. The gap is the whole point.
How are Plays of the Day chosen?
We score every projection against the book's posted line and keep only the disagreements large enough to have shown a real edge historically — by gap on strikeout props and by a run-total edge on team totals. The bigger the divergence from the line, the stronger the backtested edge.
Why don't you post plays on every market?
We surface a play only where a full-season backtest shows a durable, repeatable edge — currently strikeout props and team-total leans. Markets like the moneyline, game totals, home runs, hits, RBI, runs, and total bases are priced sharply and managed tightly by the books; across a full season even a well-tuned model can't find a repeatable edge in them, so we publish those as graded transparency — our number set beside the market, scored honestly, with no lean attached. If one ever earns a sustained edge, it surfaces automatically.
Does a human check the plays before they post?
Yes. The model finds the edge, then a person reviews every play before it goes on the board. Plays we pass on are never shown and never counted — the record here is what we actually posted, not everything the model flagged. We keep both sets internally so we can tell whether that review is helping or hurting.
Are Plays of the Day betting advice?
No. These are research signals showing where our model disagrees with the market — not betting advice or guaranteed outcomes. Lines move and variance is real. We publish them so you can weigh our edge against the book yourself. Gamble responsibly — 21+.
How is the track record graded?
Every play is graded against the line that was posted when it surfaced — not a later, better number. The score for each game locks at first pitch, so nothing can be quietly revised after the fact. We track a running season hit rate and ROI per board and show them next to each play, so the record reflects results you could actually have acted on.
When do plays appear on the board?
A play surfaces only once the sportsbook has posted odds for that market and the official lineup is confirmed for that game. Odds typically arrive around 10am ET; lineups go final about two hours before first pitch. Most plays are on the board by late morning. Late-announcing games add their plays closer to first pitch.
Δ shows how far our projection diverges from the book's line — the bigger the disagreement, the stronger the historical edge. Hit rates and ROI are season-to-date, graded against the posted line. Plays of the Day are a research signal — not betting advice. 21+, play responsibly (1-800-GAMBLER).