MLB · Probability Engine App Coming Soon

The Edge Is
In The Numbers.

A 10,000-game engine that prices every prop, game line, and market across the MLB slate — then shows you the gap on the market. That gap is your edge.

Download on the App Store
10K
Sims / Prop
57
Factors
34+
Metrics
5s
Live
The Engine — what stands behind every number
▸ Top Play Today
Yoshinobu Yamamoto
Over 18.5 Outs · MLB
+10.8%
Model 54.5% · Market 43.7% · de-vigged
10,000-Game Sim
10K
Monte Carlo per prop
Live Refresh
5s
once games start
Factor Chain
57
per projection
Statcast Metrics
34+
bat speed · barrel · chase
01

Ingest

Statcast tracking, confirmed lineups, park factors, and a live weather read per game.

02

Simulate

57 factors and 34+ metrics run 10,000 games per matchup into full outcome distributions.

03

Price

Model probability is compared to the de-vigged market — the gap is real edge, never a quota.

The Learning System — a model that grades itself
Layer_01

Self-Correction

Every projection is settled against the result; a drift monitor catches slippage and a nightly retune corrects it.

Layer_02

Adaptive Weighting

Factor importance re-weights as the baseball environment shifts across the season.

Layer_03

Calibration

Do the probabilities match real outcome frequency? Measured continuously, never assumed.

Layer_04

Closing-Line Value

Every result is benchmarked against the closing market price — the industry standard for real signal.

Player Props — the full menu, priced off the sim
Prop_01

Hits

Modeled off contact quality, matchup, and lineup slot — not last-10 noise.

Prop_02

Total Bases

Power and contact blended; extra-base tails pulled from barrel rate and exit velo.

Prop_03

Home Runs

Barrel rate, park, and pitcher HR/9 into a true long-ball probability.

Prop_04

Hits + Runs + RBIs

A joint distribution across the batter's entire night, not three guesses stapled together.

Prop_05

RBIs & Runs

Runners-on context and lineup protection priced in — not raw rate.

Prop_06

Walks

Plate discipline against pitcher command and zone rate.

Prop_07

Stolen Bases

Jump, catcher pop time, and pitcher hold priced explicitly.

Prop_08

Multi-Hit

2+ hit probability read straight off the full contact distribution.

Prop_09

Pitcher Strikeouts

Whiff rate, opponent chase, and expected batters faced.

Prop_10

Outs Recorded

Projected length off pitch count, efficiency, and the bullpen script.

Prop_11

Earned Runs

Contact quality allowed and sequencing — not just ERA.

Prop_12

Hits & Walks Allowed

Expected contact and free passes against the day's lineup.

Game Lines — priced off the same sim
Game_01

Moneyline

Win probability from the simulated score distribution.

Game_02

Run Line

The -1.5 / +1.5 spread priced off the margin distribution.

Game_03

Game & Team Totals

Runs from starter, bullpen, lineup, and park — weather-adjusted.

Game_04

NRFI / YRFI

No-run-first-inning off the starter, top of the order, and park.

Specialty & Combos — straight off the same simulation batch
Spec_01

Moonshot Money

A dedicated home-run model surfacing the day's best long-ball value.

Spec_02

Combos

Multi-leg tickets priced from joint outcomes — real correlation, not multiplied odds.

Spec_03

Live Re-Score

Lines re-priced in real time as in-game swings open new gaps.

The Edge — the gap between price and probability

The market sets a price.
We set a probability.

Market Implied
43.7%
Diamond Edge™
54.5%
The gap is the edge+10.8%

Probability, not a lean

A real number with a distribution behind it — never a vibe or a "confidence level."

De-vigged fair value

We strip the book's margin first, so the edge you see is against true price — not the hold.

Verifiable output

Displayed probability always equals market implied plus edge. You can check the math yourself.

Negative edges shown

When the book's price beats ours, we say so. Honesty is the whole product.

Ranked, never a quota

A rank score surfaces the strongest plays — the engine never forces a pick to fill a slate.

Adversarial by design

Built to beat the market, not to sell you picks. The engine is on your side of the screen.

Features — every tool you'll actually use
▸ See the terminal

Matchup Lab

Full pitcher-vs-batter breakdowns — lineup power, bullpen usage, park & weather.

Moonshot Money

The home-run model, ranked — the day's best long-ball probabilities in one view.

Combo Builder

Build multi-leg tickets and watch model probability vs. implied update in real time.

The Scorecard

Full record, win rate, and honest commentary — including the months we ran cold.

Bet Tracker

Auto-graded results and closing-line value, so you know if you're beating the number.

Prop Research

Ranked batter and pitcher leaderboards by contact, power, strikeouts, and run prevention.

Live Tracking

Once games start, follow the pitches, at-bats, and lines update in real time.

Player Search

Jump to any hitter or arm and pull their full model card for tonight.

Pricing — one plan, everything included
Diamond Shark Pro
$29.99
per month
7-day free trial
  • The full engine — 10,000-game sim, 57 factors & 34+ Statcast metrics behind every output
  • Every market — 15 player props, game lines, NRFI/YRFI, Moonshot & Combos
  • The whole toolkit — Matchup Lab, Combo Builder, Prop Research, live tracking & the Scorecard
  • Nothing hidden — full track record, negative edges shown, math you can check
No spam. We'll email you the app link the moment it's published.
Download on the App Store
About — who's behind it

Built by people who
wanted this to exist.

We spent years looking for a baseball tool this thorough and this honest. It didn't exist — so we built it. Diamond Shark is a quantitative research platform built by a team of Wall Street probability professionals, software engineers, and serious baseball people.

This isn't a chatbot in a baseball skin. Every projection comes from a purpose-built engine — a 10,000-game simulation, 57 probability factors, and 34+ Statcast metrics on every pitcher and batter, every day. Ask a general AI for 19 million probability calculations across a full slate and it simply can't: it hands you a confident paragraph, not a distribution. The math has to be built and run — on real infrastructure, against live Statcast data, every morning. That's the engine. AI checks our work; it doesn't do our work.

The inputs are ours.
The outputs are yours —
and every one is verifiable.

And we don't hide the results. The Scorecard shows the record, the win rate, and the honest commentary — including the stretches the model ran cold. A track record you can't check isn't a track record.