How SkegBets makes picks

SkegBets is a sports betting analysis publication. It is not a sportsbook, and it does not take bets or hold money. What it does is estimate, for each game and market, how likely an outcome really is, then publish a pick when that estimate is far enough from the sportsbook price to be worth backing.

The model

Every pick starts with a machine-learning model trained on years of play-by-play data across the sports we cover. For a given matchup it produces a probability for each market: who wins, how many runs or points score, whether a player clears a prop line. That probability is then compared to the price the sportsbook is offering.

A price implies a probability of its own. When the model's probability sits meaningfully above the price's implied probability, the bet has positive expected value, and only then does it become a published pick. Most games produce nothing worth betting, which is the point. The job is to wait for the spots where the number and the price disagree for a reason.

The point-in-time rule

A model is only honest if it never sees the future. Every input the model uses is computed strictly from data that existed before the game started. No final score, no box score, no line that moved after first pitch ever reaches a training example. This is enforced in code, not by convention: the training pipeline refuses to run if any feature can be shown to depend on information dated on or after game time.

The reason to be this strict is that hindsight is the easiest way to fool yourself. A backtest that quietly peeks at the result will look brilliant and lose money live. Holding the line here is what makes a backtest worth trusting.

Graded in the open

Every pick that posts is graded and added to a public ledger. The ledger carries the final result, the running win rate, unit profit and loss, and average odds. Losing picks stay on the record. There is no private tout channel and no quietly deleted history.

You can read the full record on the track-record page, and the live picks on the picks page.

Why closing-line value matters

Win rate over a few weeks is mostly noise. The measure that separates a real edge from a hot streak is closing-line value: did the pick beat the market by the time the line closed. If picks consistently land on the right side of where the market settles, the edge is real even before the games are played. SkegBets reports closing-line value next to win rate and ROI so the record can be judged the way a sharp bettor would judge it.

What "SkegBets editorial" means

The written analysis on each pick is produced with the help of a language model, working only from the verified data the pick is built on, and every number and name in the copy is checked against that data before it publishes. The analysis explains a decision the model already made. It does not invent the pick, and it is not allowed to cite a statistic the data does not contain.

The honest caveats

No model guarantees a result. Sports are variance-heavy, and any edge shows up only over a long run of bets. SkegBets does not promise profit, and no one should bet more than they can afford to lose. If betting stops being fun, step away and use the resources on our terms page.

Common questions