The method

The slow reference is the average of two forecasts that react slowly to news: a Poisson model (attack and defence of each team over three seasons, 180-day half-life, shrinkage towards the mean) and ClubElo's Elo converted into probabilities (home advantage 0.52, slope 0.53, draw threshold 0.60, calibrated on 2019-2023).

The market is the average of the bookmakers' odds, with their margin removed. That is what the public pages show as percentages, without ever showing odds.

For European cups, the per-league model does not apply: the reference is the global Elo. For women's football, which ClubElo does not cover, the reference is the market alone.

Over several thousand matches from six leagues and seven seasons, the obvious strategies lose: always following the favourite, as much as betting the model's "value" against the market. Only one rule comes out positive over the whole period: following the market when it moves clearly away from the slow reference — neither too little nor too much. That threshold, the measured returns and their margin of uncertainty are part of the signals reserved for members.

These results come from a backtest. Real returns will probably be lower, and long losing streaks have been observed.

• Interim coach: no bet (measured: −0.28 points per match).
• Team in great form or big seller in the transfer window: base stake.
• Gap the press does not explain: no bet.
• Staking plan: base stake plus 5% of gains, caps, automatic break after 20% losses over 30 days, ratchet, capital withdrawal.