Notes on how the model actually works, and on the mistakes we made building it. All of it written from a betting record that is public, losing runs included.
Look-ahead bias hiding in the data pipeline, why an oversized edge usually means our own model is wrong, what happens when you bet into a sharper market, and why we price off probability rather than expected value.
Setting a bankroll, staking in units, why you lose more bets than you win, and how to sit through a five-week drawdown without quitting at the bottom.
Why firms restrict winning accounts, and what actually buys you time: exchanges, spreading across many bookmakers, how your stake sizes give you away, and the behaviour that gets you flagged fastest.