● ABOUT · MARCDUCK · PUBLISHER + CURATOR

About MarcDuck

MarcDuck is the publisher and curator of Bookie Bullies. Every pick on this site is made by MarcDuck with the assistance of AI software, not by AI alone. The AI surfaces edges; MarcDuck reviews, refines, and formulates the final pick before it goes on the card.

How Picks Are Made

Every MLB and NBA pick goes through a two-stage process. First, the AI model ingests up to 35 factors per game — starting pitcher xFIP/fipMinus, bullpen quality and fatigue, park factors by batter handedness, weather (temperature, humidity, wind direction × ballpark orientation), home-plate umpire tendencies, catcher framing, lineup splits vs handedness, recent form (last 7/14/30 games), travel and getaway-day penalties, defensive Outs Above Average, line movement and steam detection, day vs night offense splits, dynamic per-team home-field advantage, run-expectancy matrices, BABIP/LOB% regression flags, velocity drop, and more. The model produces win probability, expected value, and a recommended stake.

Then MarcDuck reviews the model's output. Edge cases the model can miss — late injury news, weather updates that haven't propagated, narrative-driven line moves — get human refinement before the final pick is published. The AI is the assistant; MarcDuck is the author.

Track Record

Every published pick is logged, graded, and displayed publicly at /track-record.html. As of the latest grading: 251 graded picks, 54.9% hit rate, +8.78 units, +9.9% ROI at recommended 1/8-Kelly stakes. The track record updates daily and includes splits by sport, pick type, confidence tier, and rolling 7/14/30/90-day windows.

Most public AI sports pick sites hide their track record. Bookie Bullies publishes it because long-term ROI is the only honest measure of skill — short-term wins/losses are noise.

Methodology Documentation

The full 35-factor model is documented at /methodology.html. The model uses Skellam distribution for MLB margin probabilities (correct family for integer-run differences) and Negative Binomial for run totals (real run distributions are overdispersed Poisson). NBA uses Gaussian since basketball isn't Poisson.

Probabilities blend model output with market and closing-line implied probabilities at calibrated weights. Stakes use 1/8-Kelly with variance adjustment via the Bayesian uncertainty band shown next to each pick.

Areas of Expertise

Editorial Standards

Picks are published for entertainment and informational purposes only. This is not financial advice. Past performance does not guarantee future results. Every pick can be wrong, and many are. We are not a sportsbook, broker, or registered investment adviser.

21+ entertainment only (18+ where permitted). You must be physically located in a state where sports wagering is legal. Gambling problem? Call 1-800-GAMBLER or visit ncpgambling.org. See our full disclaimer + terms.

Contact

Questions or feedback? Email hello@bookiebullies.com.

Last updated: 2026-05-13