What Is a Backtest? A Real One, and What It Can't Tell You

A backtest replays a betting rule on matches that have already been played, at the prices that were on offer before kickoff, and adds up what it would have won or lost. It is the cheapest way to test an idea, and the easiest way to fool yourself. This article walks through one real backtest from start to finish, and then through everything it cannot tell you.

The rule

Captain Hindsight's first strategy, the one this site launched with:

**In the Eredivisie, back every Over line (over 2.5 goals, over 3.5, and so on) priced between 1.50and 2.50 at Toto before kickoff. €10 flat on every one.**

A rule like this has to be precise enough that a computer can apply it without judgement: which competition, which market, which price range, which bookmaker, what stake, and what to do when several lines of the same match qualify (here: bet all of them).

The result

Replayed on every Eredivisie match from 7 August to 31 October 2026 (95 matches, all with a Toto price before kickoff):

By line:

On paper, a strategy that made 18.5% on its stake in three months. Here is why that sentence alone is worth very little.

What a backtest cannot tell you

1. Whether it was luck

122 bets is a small sample. At odds around 2.00, chance alone moves a 122-bet ROI by roughly ±18% either way. This result sits just outside that band, which is interesting, but not proof. See How many bets before a result means something?.

2. Whether it still works

The result was not spread evenly. The first 71 bets (to 15 September) made +€220.80; the 51 bets after that made +€4.80. Almost the whole profit came from the first six weeks. See The hot-start trap.

3. How many bets it really was

Several lines of the same match usually win or lose together: if a game ends 1-0, Over 2.5 and Over 3.5 both lose. 122 bets were placed on only 95 matches, so the real number of independent outcomes is smaller than the bet count suggests. See Why Over 2.5 and Over 3.5 in one game aren't two bets.

4. How many rules were tried before this one

This rule was found by looking at the data. If you try enough rules on the same matches, one of them will look brilliant by accident. See Overfitting.

5. Whether you could actually have placed the bets

A backtest assumes you got the price. In reality, prices move, accounts get limited and bets get refused. Our backtests use one bookmaker's real pre-kickoff prices (Toto) and never the best price across several books, because picking the best afterwards is exactly the kind of hindsight a backtest must avoid.

So what is a backtest good for?

Killing bad ideas cheaply. Most rules lose once you replay them, usually by roughly the bookmaker's margin, and finding that out costs nothing. A rule that survives a backtest has earned the next step, which is not real money but paper betting: applying it to matches that have not been played yet.

That is the whole joke behind Captain Hindsight: in hindsight, every strategy is obvious. The honest version of a backtest is the one that shows you, in the same breath as the headline number, all the reasons it could be wrong.

The honest checklist

Before you believe any backtest, including ours, ask:

• How many bets, over how long? Under 30 bets, ignore the result entirely.
• Is the ROI outside the luck band for that number of bets?
• Did it work in both halves of the period, or only in one?
• How many bets were in the same matches?
• How many rules were tried to find this one?
• Whose prices, taken when? One book's real prices, or the best of several picked afterwards?

Source: Edge Terminal backtester, rule leagues=eredivisie&class=totals&outcomes=over&min=1.5&max=2.5&stake=10&per=all, Toto prices before kickoff, re-run 3 November 2026 on all Eredivisie matches played 7 August – 31 October 2026 (95 of 95 priced). Past results only; nothing here is a tip.