MLB is where PropsBot’s track record is longest and the ledger is deepest. The 31.7% ROI on 101,881 logged props came out of this sport. So did the 82.6% high-confidence hit rate. Today’s picks come out of the same model that produced those numbers.
Today’s MLB picks, posted before first pitch
The slate goes up in the early afternoon Eastern. Picks adjust as starting pitchers confirm, lineups drop, and weather updates roll in for outdoor parks. The last update lands around 30 minutes before the East Coast first pitch.
For longer-form market hubs (home runs, strikeouts, hits), the deeper pages handle those:
This page is the day-of curation: top-confidence picks across all MLB markets in one place.
What MLB-specific factors actually matter
Baseball variance per game is high, sample sizes per player are huge, and the lines are softer than you’d expect because pricing 30 games a day is hard work for sportsbooks. The model leans on these:
Pitcher matchup history at the batter level. Batter vs pitcher splits are noisy at the single-PA level but useful in aggregate when you weight by handedness, pitch mix, and recency. The model uses a Bayesian shrinkage on small samples — not a stat slash that says “.400 lifetime against this guy.” That’s the kind of number that gets called out in TV graphics. It’s not signal at the volume the model needs.
Park factors by handedness. Coors plays bigger for righties than lefties on home runs. Oracle plays smaller for both. Citizens Bank plays huge for left-handed power and average for right. The park adjustment isn’t one number per stadium. It’s split.
Bullpen leverage projection. A starter going six innings means the bullpen takes the seventh through ninth. The model projects which arms get those innings based on rest, recent usage, and game-script. A high-leverage righty getting the seventh in a one-run game changes what the eighth-inning projection looks like for the visiting lineup.
Weather: wind, temperature, humidity. Wind blowing out boosts HR projections at every park except domes. Hot, humid air carries the ball further than cold dry air at the same wind speed. The model bakes this in for outdoor games and zeroes it out for domes.
Pitcher fatigue and pitch-count trajectory. A starter on three days’ rest projects differently than the same starter on five. A pitcher whose last three starts averaged 102 pitches projects to a longer outing than one averaging 87.
Closing-line value across all major books. The same projection produces a no-vig fair price. The Edge value is computed against that price, not against any single book.
What “high confidence” looks like in MLB
The 82.6% hit rate on the High Confidence cohort came out of 136,953 picks across all sports. MLB drove the largest share of that volume. A high-confidence MLB pick is one where the model’s projected probability sits well above the no-vig implied probability of the line, even after accounting for variance.
The 31.7% ROI on 101,881 props came from a different filter: the High ROI Signal. That’s a smaller, edgier subset where the gap between projection and line is large enough that even at the lower hit rate, expected value is well above break-even.
Both sit on the same ledger. Both are auditable.
Where the methodology lives
Performance methodology and the public ledger
The methodology page is the audit trail for the trust strip above. Every pick on every night gets logged with timestamp, posted line, model probability, and result. The 31.7% / 82.6% / Brier-below-Vegas numbers are the sums.
If you want to read what each metric is actually measuring before trusting any pick, the methodology page is the one to read. Brier score in particular is the one that matters most to sharps — it penalizes overconfidence and underconfidence equally, which is why beating Vegas at it is a real claim.
Markets PropsBot covers in MLB
Hits, total bases, runs scored, RBIs, home runs, walks, stolen bases for batters; strikeouts, hits allowed, walks allowed, earned runs, outs recorded, win for pitchers. The batter and pitcher pools cross-pollinate — when the model has high confidence on a pitcher’s strikeout over, the same batters facing him often show up on hit unders.
The deepest signal tends to live in the strikeout, total-bases, and home-run markets. Those have the widest line dispersion across books and the most edge to capture.
Three ways to use today’s MLB slate
Single-leg the high-confidence picks. Stake even across the top three to five high-confidence plays. Over a 162-game season this is the workflow with the cleanest math. The 82.6% cohort hit rate is what makes it work.
Build a same-game parlay (carefully). SGPs are hard to value because legs correlate. The SGP value calculator does the math on which combinations are mispriced versus correlated noise. Use it before you fire.
Stack pitcher props. A pitcher with a high-confidence strikeout over and a high-confidence outs-recorded over is one underlying story (he’s going deep into the game and missing bats). Two legs, one thesis.
The hold and vig calculator and no-vig fair odds calculator help confirm the price you’re getting is actually a price worth taking.
Adjustments through the day
Lineups for evening games drop in the late afternoon. A scratched batter pulls every pick tied to him. Picks tied to teammates (a pitcher facing that lineup, for example) get re-evaluated.
Weather refreshes continuously. A wind shift at Wrigley from out to in flips the home-run picks for that game.
Pitcher confirmations matter most for double-header days and bullpen games. If a probable pitcher gets pushed back, the slate rebuilds for that game.
FAQ
How accurate are the MLB picks? The 31.7% ROI came from 101,881 MLB props on the High ROI Signal. The high-confidence cohort hit at 82.6% across all sports including MLB. The Brier score on MLB predictions is below the Vegas closing-line baseline, which means the model is better calibrated than the consensus market.
Are the picks free? Yes. The public list shows Confidence, Edge, and posted line on every pick. The full historical export and deeper filters are in the app.
How early are the picks posted? Early afternoon Eastern. Updates roll in through first pitch as lineups confirm and weather updates.
What’s the best pick for PrizePicks or Underdog from this list? Confidence 75 and above. That’s the practical floor for DFS pick’em payout structure. Stack two or three high-confidence picks in a 2-pick or 3-pick Power Play.
Do you cover home run props specifically? Yes. The home-run market is one of the deepest sources of edge in MLB because line dispersion across books is wide. The MLB home run props hub goes deeper.
How do you handle weather changes during the day? Weather refreshes continuously. A material wind or temperature shift triggers re-projection for that game. The picks list updates accordingly.
Why do some MLB picks have huge edge values (10%+)? The MLB market is softer than NFL or NBA on player props because there’s more volume to price every day. Edge values of 5-10%+ aren’t unusual on the High ROI Signal. They’re also where the variance lives — the model is right more than it’s wrong, but the wrong nights hurt.
Bottom line
MLB is the sport where PropsBot’s ledger is deepest. The picks list every day comes out of the same model that produced the 31.7% ROI on 101,881 logged props. Read the methodology page once, then trust the numbers.
