Score markets — over/under goals, both teams to score, totals — modelled from the same fitted distributions as the match tips, plus the accumulator analysis.
Ranked by Kelly-staked edge against the de-vigged Betfair price, model confidence as the tiebreak. Fixtures vetoed by the AI risk officer are excluded; risk and momentum flags from the daily AI checks are shown where the fixture was assessed today. Exactly this slate is recorded to the £10 ledger below.
A flat £10 on every priced score pick featured above, logged each morning before results are known and settled at the real Betfair price we showed. Green line = bankers (short-priced, model-agreeing picks), amber = edge plays. Unpriced picks are never in this ledger — a paper bet with an invented price would measure nothing.
Every settled £10 bet from the card, grouped by the day it settled — the same dates and the same pounds the P&L chart above is drawn from. The price is the one recorded the morning the bet was featured. Bets whose event has finished but which haven't been graded yet are listed separately below as awaiting a result.
Rolling Brier skill against each market's own base rate — the only honest yardstick for markets whose base rate sits near 50% ("53% of matches go over 2.5", so 53% accuracy is a coin flip). Above the dashed line the probabilities genuinely know something; below it they are worse than a constant. Lines start drawing once a module has ~40 settled markets.
Both lines stake exactly £10 a day and differ only in how it is spent: rolled into one accumulator of that morning's sport bankers, or split flat across the same legs as singles. Same money, same picks, same days — so the gap between the lines is the effect of stacking them, and nothing else. An acca only counts once every one of its legs has settled. A banker acca multiplies the bookmaker's margin along with the returns — bankers carry no priced edge, so rolling them up compounds the vig. This is the acca that FEELS safe and is structurally the weaker bet (see the edge acca).
The two right-hand tiles are the £10 Challenge on its own basis — £10 on every tip, so a four-banker day stakes £40. They are not comparable in pounds to the two on the left, which stake £10 for the whole day. Per-sport bankers-vs-edges charts are on each sport page.
Both lines stake exactly £10 a day and differ only in how it is spent: rolled into one accumulator of that morning's sport value plays, or split flat across the same legs as singles. Same money, same picks, same days — so the gap between the lines is the effect of stacking them, and nothing else. An acca only counts once every one of its legs has settled. For independent legs each carrying real edge, the acca's EV compounds (Π(1+e) − 1 beats the sum) — so IF the edges are real, this is the only acca with a theoretical case. The price of that: a long staircase down punctuated by rare spikes. Read it against the expected-wins number, never against the last few days.
Both lines stake exactly £10 a day and differ only in how it is spent: rolled into one accumulator of that day's recorded top-3 priced score picks (O/U goals + BTTS), or split flat across the same legs as singles. Same money, same picks, same days — so the gap between the lines is the effect of stacking them, and nothing else. An acca only counts once every one of its legs has settled. Legs are the same priced score picks the Scores ledger records — an acca only counts once every leg's fixture has a final score.
Score markets are derived from the same fitted models as the match tips — Dixon-Coles scoreline grids for football, a league-environment totals model for the score sports — and every probability is graded against what actually happened. Informational research and paper-trading only — no real money is staked.
