No jargon. No formulas. Just the core ideas behind why a disciplined, numbers-driven approach to baseball betting works over time.
Sportsbooks are stores that sell bets. Like any store, they set their own prices. And like any store, sometimes those prices are wrong.
When you see the Dodgers listed at 1.65 odds, that's the store price. It means the sportsbook thinks the Dodgers win about 60% of the time. But what if they actually win 66% of the time? Then the store is selling something for more than it's worth — and you're the one who benefits.
Imagine a jar of 100 marbles — 66 blue and 34 red. Someone offers you a bet: "Pick a marble. If it's blue, I pay you $1.65 for every $1 you bet." You'd take that bet every single time, because the math is permanently in your favour. That's what this model looks for.
The model's only job is to figure out the real price of each game — how often each team actually wins — and then compare that to the store price. When there's a big enough gap, that's a bet.
It asks a simple question: in this specific game, how many runs will each team score, and how many will they allow?
That's it. All of baseball comes down to this: score more runs than your opponent. The model estimates both sides of that equation for every game, then converts the run gap into a win probability.
How many baserunners does he allow per inning? How many home runs? Does he strike batters out or let them put the ball in play? This tells us how many runs the opponent will likely score off him.
The starter only pitches part of the game. The model also measures the relief pitchers who take over — because a great starter followed by a bad bullpen can still lose.
How often does this team get on base? How hard do they hit? A lineup that gets on base a lot and hits for power will score more runs against the same pitcher than a weaker lineup.
Offence × opponent pitching = estimated runs for each side. The team expected to score more runs, relative to what they'll allow, gets a higher win probability.
Because no prediction is exact, the model runs each game 10,000 times with slightly different random outcomes — like playing the game on paper ten thousand times. The average of all those simulations becomes the final prediction.
The model uses the same pitcher-evaluation methods that front offices use to build their rosters. It strips out luck (a pitcher who gave up a lot of bloop singles) and focuses on skill (strikeouts, walks, home runs) — things that actually predict future performance.
Not because it picks every game right. It doesn't. Nobody does. It wins because it only bets when the math is significantly in its favour.
Imagine you could flip a coin that lands heads 54% of the time instead of 50%. On any single flip you might lose. After 10 flips, you might be behind. But after 500 flips, you are almost certainly ahead. That's the fundamental principle.
The model doesn't bet on every game. It waits for games where its calculated win probability is at least 12% higher than what the sportsbook's odds imply. That gap is called edge, and it's what makes the math work.
Sportsbooks are very good at setting lines. But they're not perfect — and they don't need to be, because most bettors aren't checking their work. This model checks their work on every single game, every single day.
The model doesn't swing at every pitch. It has strict rules that keep it out of trouble.
Won't bet unless the model sees a 12%+ advantage over the sportsbook's price
Won't bet on long shots above 2.65 odds — too much variance, even with an edge
Every bet is the same size. No chasing, no doubling down, no emotion
Flat betting is the most important rule. Every qualifying bet gets the same dollar amount, whether the model is 55% confident or 70% confident. This removes the single most common way people go broke betting on sports: increasing their bets after losses, or getting overconfident after wins.
The model also blends preseason projections with live stats throughout the year. Early in the season, when every team has only played a few games, it leans heavily on expert projections rather than trusting small, noisy samples. As the season progresses and the data builds up, it gradually shifts toward real performance.
It's not a gut feeling, a tip, or a trend. It's a structured process that runs the same way every morning.
Starting pitching is the single most predictive factor in a baseball game. The model builds its entire projection around who's on the mound, not who's hot or who's on a streak.
A pitcher who's been "unlucky" — giving up hits on softly hit balls — will look bad in traditional stats but good in the model. The model focuses on outcomes the pitcher controls: strikeouts, walks, and home runs.
Instead of one prediction, the model produces a distribution of outcomes. This means it knows when it's confident and when it's uncertain — and it won't bet on games where the noise is too high.
Every prediction, every bet, and every result is logged and tracked. There's no selective memory and no hiding from bad streaks. The numbers are the numbers.
Professional sports betting isn't about big wins. It's about a small, consistent mathematical advantage applied hundreds of times with absolute discipline. The model provides the advantage. The process provides the discipline.