T20 method

How the T20 ratings are made

Strike rates and economy rates treat every ball alike. A run a ball in the powerplay of a chase of 140 and a run a ball at the death of a chase of 220 are recorded the same way, and they are not worth the same. A wicket in the first over and a wicket with the game already gone go into the same column. Fine Margins rates T20 players in the one thing every ball changes: the chance of winning. This page says how, how it was tested, and where it falls short.

One currency: win chance

Every ball in a T20 match moves each side's chance of winning. A dot ball in a tight chase moves it a lot; a single with 60 needed off 12 barely moves it at all. Each player's rating is how much he moves that chance compared with an average IPL player doing the same job: per innings for a batter, per match for a bowler. A rating of +2 means about two points of win chance more than an average IPL player would give his side.

What the model learns about each player

The model works ball by ball. For every player it learns how often each outcome happens when he is involved: dots, ones, twos, fours, sixes, wides, wickets and run-outs. It also learns how he changes with the situation. That covers the over, the powerplay, wickets in hand, whether his side is chasing and what the required rate is, how long he has been in, the stand he is part of, how long his spell has run, and the type of bowler or the hand of the batter he faces.

A batter who starts slowly and then accelerates, or a bowler who is strong in the powerplay and weak at the death, is described as he is, not averaged into one strike rate. Those habits are then played through a match simulator in a standard role, against average opposition, to turn them into win chance.

One scale across 18 competitions

A run in the IPL is harder to come by than a run in most leagues, and a run against associate sides is easier. The ratings put every competition on one scale using the players who appear in more than one of them: when the same batter plays the BBL, the CPL and the IPL, the gaps between his numbers in each say how strong each league is. With enough such players the strength of every league is estimated together, and every rating is stated against an average IPL player.

The competitions covered are the IPL, BBL, PSL, CPL, SA20, ILT20, MLC, the Blast, the Bangladesh, Lanka and Nepal Premier Leagues, the Mzansi Super League and South Africa's domestic T20, New Zealand's Super Smash, India's Syed Mushtaq Ali Trophy, and men's T20 internationals between Test nations, mixed fixtures and associate nations.

Small samples and ranges

A player with 250 balls behind him has shown much less than one with 5,000. The model is built so that a short record moves a player only part of the way from the average for his league and his kind of player; the longer the record, the more it is taken at face value. Every rating carries a range: the band within which his true level lies with 89% confidence. A wide range is the model saying it is not sure, and it is shown so the number is read with no more confidence than it deserves.

Players need at least 500 balls in a role to appear in the public rankings. Ratings built on 200 to 499 balls held up in testing too (see below), and Fine Margins uses them in scouting work, marked as thin samples. Below 200 balls a player counts as an average player of his league.

This season, not the whole career

The headline rating is a forecast of the coming season. Recent seasons count for more than old ones, and the forecast is tuned on past seasons so that it does not promise more than past forecasts delivered. A career rating, which weighs every season alike, is kept alongside it.

How well it works

The ratings were tested the hard way. For each season from 2021 to 2026, the model was fitted using only the matches played before that season, every player was rated from it, and the ratings were then checked against what the players actually did in that season.

Players rated onBattingBowling
500 balls or more0.430.34
200 to 499 balls0.520.18

Correlation between the rating before the season and the value delivered in it. 1 would be a perfect forecast; 0 none at all.

In every group the ratings beat a player's own past numbers as a guide to his next season. Players rated on short records were not overrated: on average they did slightly better than their ratings said. The weak spot is clear too. Bowlers with short records are hard to read, and their ratings are a lead to follow up rather than a verdict.

A season of T20 cricket is short, so even a perfect rating could not predict a player's next season exactly: luck in a few innings decides a lot. The numbers above are best read against each other, and against the alternatives, rather than against 1.

From players to teams

A team's strength is the value of its best XI under the competition's rules: seven batting places and five bowlers, a wicketkeeper, and the limit on overseas players. The XI is chosen by the numbers, not by name, which is why a famous player can miss it. Players without a rating count as an average player of the league.

What it doesn't do

It cannot rate what the data does not hold. The ball-by-ball records cover the competitions above, and not, for example, Pakistan's National T20 Cup or most domestic T20 outside India, South Africa, England and New Zealand. A player whose cricket is mostly there is either unrated or rated on a small part of his game.

It barely measures fielding and keeping. The data records catches and run-outs, not the chances that went down or the runs saved in the ring. Captaincy, fitness and how a player suits a particular ground are not in the base rating either.

And it sees outcomes, not technique. A four off the edge and a four through the covers count the same.

Data and attribution

The ratings are built from ball-by-ball data published by Cricsheet. Fine Margins is an independent project and is not affiliated with any cricket board, league or franchise.