Value betting in football

Two numbers matter. The market probability is what the price charges you for, corrected for margin. The estimated probability is your own view, from a model, from ratings, or from how money has moved. Value exists only when the second is larger than the first.

value exists when p_estimated > fair market probability

A home win is priced at 2.10. Raw implied probability is 1 ÷ 2.10 = 47.6%; after removing a 4.8% overround the fair probability is 45.4%. If your estimate is 53%, the edge is 53% − 45.4% = 7.6 percentage points.

Expressed as expected value per €1 staked: 0.53 × (2.10 − 1) − 0.47 = +€0.113, or +11.3%.

Flip the estimate to 42% and the same price is a losing bet in the long run — even though the team may well win the match. The probability estimate does all the work; the arithmetic is trivial.

edge = p_estimated − fair probability

A 53% shot loses 47 times in 100. A value bet is a claim about the average of many similar decisions, not about this Saturday. Any single result — win or lose — tells you almost nothing about whether the estimate was good.

This is also why a losing week is not evidence a method is broken, and a winning week is not evidence it works.

A 5% edge is invisible over ten bets: random variation at that scale is many times larger than the edge. Under 100 settled bets, treat any return as an illustration of the process. Several hundred settled bets are needed before a return on investment becomes a serious estimate of edge.

That is why every performance figure on this site is published next to its settled-selection count.

Most value-betting errors are estimation errors dressed up as maths.

• Comparing a model against raw implied probability instead of fair, margin-free probability — this invents edge that is not there.
• Treating a large edge as a strong signal. An implausibly large edge usually means the estimate is wrong or the price is stale.
• Rebuilding the estimate after seeing the price, which quietly turns the market's opinion into your own.
• Counting only the bets that won when reviewing the record.

BetBuddy does not claim to know the true probability of a football match. Its selection signal is market-based: outcomes on high-volume markets whose price has shortened materially from its opening value, subject to a hard minimum price of 1.80 and a maximum of 5.00.

The reasoning is that a well-supported price move on a liquid market carries information the opening price did not have. That is an assumption, and the settled history is published so anyone can check whether it has held.