Live Picks Engine

// 01 · Real-Time

Connector Status

Today's & Upcoming Games

Market accuracy ratings — be honest about which calls to trust most: AMoneyline · strongest signal (2025-26 hold-out: — hit rate, Brier —)
BSpread · slightly weaker than Vegas (Margin RMSE — vs Vegas ~12-13)
CTotal · weakest market (Total RMSE —; runs systematically high in playoffs). Use lowest stakes here.

The Live Predictor

// 02 · Bayesian Inference

AUTO-CYCLE

Home Team

Away Team

Pace (poss/g)

Rest Δ

Each prediction is closed-form — a Skellam (difference of two Poissons) reduced to a Normal gives the win probability instantly, no sampling.

OUTPUT // POSEIDON

—

home team

—pts

vs.

—

away team

—pts

Win Probabilityclosed-form

— % — %

Spread

—

Total

—

Edge vs Pickem

—

Kelly @ −110

—

Margin Distributionhome pts − away pts

The Council of Mathematicians

// Foundations

Five thinkers whose work, properly applied, forms the actual mathematical backbone of any serious sports-betting system. Each has a real, computed contribution to Poseidon — no metaphors, no vocabulary borrowing.

01 // Bayes

B

Thomas Bayes

1701 — 1761

Posterior updating. Team-strength priors get updated with all new evidence; the model is recomputed weekly through the entire backtest.

Live Output

— teams modeled

refit every 7 days

02 // Pascal

P

Blaise Pascal

1623 — 1662

Expected value. We never bet on who wins — we bet when our edge over the book's de-vigged probability is positive.

Live Output

EV calculator below

de-vig + edge calc

03 // Kolmogorov

K

Andrey Kolmogorov

1903 — 1987

Signal vs noise. Pythagorean expectation separates a team's deserved record from luck.

Live Output

Luckiest: —

Unluckiest: —

04 // Thorp

T

Edward Thorp

b. 1932

Kelly criterion. Compound-growth-optimal sizing. The bankroll simulator below shows what happens at every fraction.

Live Output

2× Kelly ruin: —

why fractional matters

05 // Mandelbrot

M

Benoit Mandelbrot

1924 — 2010

Fat tails. Extreme outcomes happen more often than Gaussian intuition predicts. Real bankroll math accounts for this.

Live Output

Excess kurtosis: —

NBA margins are mildly fat-tailed

Pascal · Expected Value Engine

// 04 · Edge Detection

Plug in any matchup probability + book line. Engine computes the implied probability after de-vig, the expected value of betting, and the optimal Kelly fraction.

Use Model Pick (optional)

Your Estimated Win Probability (%)

Book Line — Side A (American odds)

Book Line — Side B (American odds)

Bankroll ($)

Expected Value Per $100 Bet

—

Implied Probability (de-vig)

—

vs your estimate

Edge

—

your probability − implied

Full Kelly Suggestion

—

% of bankroll · risky

¼ Kelly (Recommended)

—

— suggested bet

Thorp · Bankroll Monte Carlo

// 05 · Risk of Ruin

AUTO-CYCLE

1,000 simulated betting careers, each 200 bets sampled from the model's high-confidence picks at -110 odds. Cycling through Kelly fractions automatically.

Caveat: simulation assumes -110 lines match model's empirical hit rates. Real long-run edges over closing lines are 1–3%, not 30%+. The lesson is the Kelly mechanics, not a profit projection.

0.25× Kelly — wealth paths200 bets per career · 30 sample paths

Median Final

—

5th Percentile

—

95th Percentile

—

Ruin Rate

—