The stories around a match are convincing. Most do not survive a check of the sample. Here are the usual traps, and then one match goes through the whole process.
The usual traps
- Dead rubbers: a team that has secured or lost its target may field a different side. The catch is that you know the motive, not what the manager will do. If the market has not moved, do not overrate it.
- Derbies: the emotions are real, but the statistics show no stable pattern. You usually hear "anything goes in derbies". That is a story, not a number.
- New manager: the "bounce" of the first games mostly disappears when you compare with similar teams in poor form that did not change manager. Often it is a return to the average.
- Head-to-head: few games, with different players and managers. It is not a measurement.
One whole match
Made-up data. Team A hosts Team B. Base from team strength: A 46.0%, draw 27.0%, B 27.0%.
- Absences: A’s striker is out (−2.0 on A’s win) and so is B’s centre-back pair (+2.0 on A’s win).
- Schedule: B played two days ago (+0.5 on A’s win).
- Form and motivation: no correction. The sample is small and nothing changes it.
| Step | A | Draw | B |
|---|---|---|---|
| Base | 46.0% | 27.0% | 27.0% |
| After corrections | 46.5% | 27.0% | 26.5% |
| Fair market price | 44.0% | 28.0% | 28.0% |
| Final (0.9 market + 0.1 ours) | 44.25% | 27.90% | 27.85% |
What the example leaves you with
Look at how the number moved: from 46.0 to 46.5 after the corrections and to 44.25 after the comparison with the market. The stories around the match, the motive, the derby, the history of the two teams, changed almost nothing. The market did most of the work. That is the right result when you have nothing the market does not know. And note this: even if the decision had been "play", it would not be a bet that "must" win. With a probability of 44.25% you lose 55.75% of the time. Only over a large number of bets does it show whether the threshold and the price work.