Every toss statistic you have read is built on a sample too small to use

Venue-level toss records look authoritative in a graphic and collapse the moment you count how many matches they contain.

Count the matches. Go on, actually count them.

A commentator says teams batting first have won sixty per cent at this ground. It sounds like a finding. Then you look, and the number rests on fifteen matches spread over eleven years, played under three different curators, two ball manufacturers and a rule change about fielding restrictions. Nine wins and six losses. Flip a fair coin fifteen times and you will get nine heads roughly one time in five without anything strange happening at all.

That is the whole problem, and it does not go away by adding decimal places.

Toss advantage is a genuine effect. Dew changes what a ball does in the second innings. A surface that seams in the first hour is a different surface at tea. None of that is in dispute. What is in dispute is whether we can measure the size of the effect at the level of one venue, which is the only level anyone actually cares about, because that is where the decision gets made.

The maths is unkind here. To detect a modest edge with any confidence you need hundreds of trials. Most grounds do not host hundreds of comparable matches in a decade, and comparability is the killer. Pool enough games to get statistical power and you have pooled across eras where the pitch preparation changed, which means you bought precision by measuring something that no longer exists.

There is a second trap underneath. Captains are not randomising. They read conditions and choose, so the outcomes you observe are contaminated by their judgement. If captains at a ground correctly bowl first whenever the sky is heavy, the record will show chasing teams winning, and you will credit the venue for something the weather and the captains did.

I would rather see the raw counts printed on screen. Not the percentage. The counts. Eleven and seven, with the years beside them. Let people see the sample before they see the conclusion drawn from it.

Broadcast will not do this, because nine from fifteen looks like nothing and sixty per cent looks like insight. The graphic exists to sound confident during a coin toss that lasts forty seconds.

If a number cannot survive being shown as a fraction, it was never strong enough to base a decision on.