Every ball counted

How Our Numbers Work

Most cricket sites quote you a season average. We count the deliveries. Here is what that means in practice.

The data pipeline and the models described here were built by Deeptanshu Pandey, who maintains them.

3.8M+
deliveries in our database
7,600+
matches, ball by ball
Every format
with its own span, stated

We start from the ball, not the scorecard

A scorecard tells you a batter made 42. It does not tell you that 31 of those runs came against spin, that he faced 19 dot balls in the powerplay, or that he has been dismissed by left-arm seam four times in his last eleven innings.

Our database holds every delivery of every match we cover: Test cricket, one-day internationals, T20 internationals for both men and women, the IPL, the Big Bash, the Pakistan Super League, every T20 World Cup and more, with new competitions added as we grow. That is where every number on this site comes from. Nothing here is copied from a summary table, and nothing is estimated from a career average.

What the model reads before a match

For each player in a named squad, our model works through their full delivery-level record and builds a picture from several angles at once:

  • Recent form, weighted so that the last few weeks matter more than a season two years ago.
  • The opposition attack, and how this batter has historically handled that kind of bowling.
  • The venue, because a ground's scoring behaviour is a real and measurable thing.
  • Role and position, since an opener and a finisher are being asked different questions.
  • Match context, including the format and what a competitive score looks like in it.

How those inputs are weighed against each other is our own work, and it is the part we keep. What we will always tell you is what went in and what came out.

What the probabilities actually mean

We publish the probability of a batter finishing in each scoring range rather than a single predicted score, because a single number is the least honest way to describe a T20 innings. A batter who makes 8 one night and 78 the next has not contradicted himself. He has done exactly what the distribution said he might.

So when we show a scoring range at 30 per cent, we are saying this is roughly how often an innings like this one ends up there. Over a season those proportions are what to judge us on, not any single night.

What we will not do

  • We do not publish betting advice, tips or odds, and we never will.
  • We do not claim an accuracy percentage. Anyone quoting one is selling something.
  • We do not invent a number when the data is thin. We say the sample is small.
  • We do not present a partial record as a complete career. Where our data starts, we say so.

Where we are still growing

Our ball-by-ball coverage of Test and one-day international cricket runs from 2005, IPL from 2013, men's T20 internationals from 2017 and women's T20 internationals from 2020. Every figure on the site is a figure over the matches we hold, and where that differs from a published career total we say so rather than round the gap away.

One gap is worth naming directly. Matches involving Afghanistan's men's team are withheld from the public ball-by-ball archive our Test and one-day records are built on: 13 Tests, 155 one-day internationals and 158 T20s. They are absent throughout, which is why some players' Test records here stop in November 2025. The matches happened and are on the scorecard everywhere else; we simply do not hold the deliveries, so we do not count them.

Live scores are currently offered for the competitions we can maintain properly. We would rather cover fewer matches well than list a fixture we cannot keep accurate.

Everything on CricNerd is free, and it is built for people who want the numbers behind the cricket rather than the noise around it.

See the current predictions →About CricNerd →