Day-and-Time Backtest Filters: What’s Genuinely Useful and What’s Misleading

The pitch contains a genuinely useful idea wrapped around a demonstration that, on closer inspection, commits the very error it claims to prevent. This article separates the two so you can keep the insight and discard the spin.

1The Core Idea Being Sold

The argument runs as follows. When you trade is correlated with which matches you can realistically trade, because leagues around the world kick off at different local times. A trader who only sits down on weekday evenings will never bet on a morning Asian fixture, no matter how good it looks in a backtest. So — the claim goes — a backtest run across every league at every hour misrepresents what that trader could actually have done. Filtering the historical sample down to the specific days and kickoff times the trader is available is presented as the fix.

Alongside this sits a second observation: that a time filter is “secretly a league filter,” because narrowing the clock automatically narrows the pool of leagues in play. A spectacular before-and-after example is then shown, in which a strategy’s strike rate leaps from roughly the mid-40s to the mid-90s once the filter is applied.

2What’s Actually True

Your trading hours really do constrain your league universe

This part is sound and worth taking seriously. If you can only trade at certain times, then matches outside those windows are not part of your real opportunity set. A backtest that silently includes them is measuring a fantasy schedule, not your schedule. Restricting the sample to the hours you can genuinely act on makes the result more honest, not less.

Leagues genuinely behave differently

It is also true that competitions have distinct statistical “personalities.” Home-win rates, goal frequencies, and scoring patterns vary meaningfully from one league to another. Mixing wildly different environments into a single average can hide that variation, and being aware of it is good practice.

Time and league exposure are linked

The connection between kickoff time and league composition is real. Filtering by hour does reshape which competitions dominate your sample. As a way of aligning a backtest with lived trading reality, that mechanism is legitimate.

3What’s Wrong — and Why It Matters

The headline demonstration is not credible

The showcase moment — a strike rate jumping from around the mid-40s to the mid-90s purely by deselecting one weekday and trimming kickoff times — should set off alarm bells. For a market like “both teams to score,” the outcome depends on whether two teams find the net. The day of the week and the clock time of kickoff have no plausible causal mechanism strong enough to swing that outcome by fifty percentage points. A change of that magnitude from a time slice is the signature of an artifact, not an edge.

It quietly recreates the bias it warns against

The presentation opens with a correct lecture: that traders deceive themselves by dropping the leagues that hurt a strategy and keeping the ones that flatter it — “back-fitting” the data to the result they wanted. Yet the demonstration does precisely this through a different door. Deselecting days and narrowing time windows until the number looks extraordinary is selection by another name. Wrapping it in the language of “realism” does not change what it is.

Shrinking the sample is treated as a feature, not a warning

In the example, the filtered sample falls to a fraction of the original. A smaller, heavily sliced sample is more vulnerable to noise and coincidence, not more trustworthy. An honest tool would flag that the result has become statistically thin. Presenting a near-perfect strike rate over a shrunken sample as proof of a hidden edge gets this backwards.

The supporting statistics are unsourced and conveniently round

Specific figures are quoted to dramatize the point — a league with home wins near 29%, another near 54%, an over-1.5 rate around 96% versus around 50% elsewhere. These arrive without any source and land on neat, persuasive numbers. Real league rates cluster more tightly than these cherry-picked extremes suggest. Treat such figures as illustrative storytelling, not as established fact.

The framing is marketing, not analysis

Language about being in “a completely different category,” about rivals “chasing” and “catching up,” is promotional positioning. There is nothing wrong with a company promoting its product, but it is a signal that the demonstration was built to impress rather than to inform — and the statistics bear that out.

4How to Use the Idea Responsibly

The underlying principle is worth keeping. The cure is to apply it without smuggling bias back in:

5Bottom Line

The insight is real: restrict a backtest to the matches you could actually have traded, and you get a more honest picture. But the demonstration used to sell it is statistically implausible and is itself an example of the data-shaping it claims to cure. Keep the principle. Be deeply skeptical of any “filter” that turns a coin-flip strategy into a near-certainty — especially when the sample shrinks and the explanation is enthusiasm rather than cause.