How our cricket predictions, player ratings and rankings work

Where do the figures come from?

How are cricket player ratings worked out?

• Every ball is valued by how much it changed the batting side's expected total, given the balls left and the wickets down. Tests have no limit on balls, so there it's the innings and the wickets down. In men's T20s the expected total is also set to the competition and the ground.
• That value is shared between the batter and the bowler. Wides and no-balls fall on the bowler alone, and a run-out isn't credited to the bowler.
• A batter is measured against par for the moment: an average batter in the same over, with the same wickets down, at the same point of their own innings. An opener gets no credit just for batting in the powerplay, and a No. 9 who walks in at 150 for 7 isn't marked down for it.
• Batting is split into runs and dismissals, and bowling into runs conceded, extras and wickets. Each part has its own memory, set for each format. In T20s and Tests a bowler's scoring rate settles within a couple of seasons, while wickets swing so much from match to match that they need years of evidence.
• A player with a short record starts at the level of a typical newcomer and moves towards their own record as the balls build up.
• The opposition counts. Runs off a bowler we rate highly are worth more than the same runs off a weak one, and the wicket of a top batter more than a tailender's.
• Competitions differ in strength. We measure each one by how players who move between them do against their ratings, so fifty in a weak league counts for less than fifty in the IPL. International sides are split the same way: Full Members, other ODI sides, and the rest.
• Scoring rises over the years, so every ball is set against the scoring of its own era.
• Each format draws on all of its cricket. T20 ratings take in internationals, the leagues and The Hundred, where a ball counts as 1.2 T20 balls so the two powerplays line up. One-day ratings take in ODIs and the domestic one-day cups, and first-class ratings Tests and county, state and other first-class cricket. Men's T20 ratings go back to 2012, the rest to 2014.
• A rating is published as runs a match above an average player. A batter's is the rating per ball times the balls a batter in their usual position faces; in Tests it's their own balls a match, settled towards their position's over their first few Tests. A bowler's is the rating per ball times a fifth of the side's balls. Half a standard error comes off first, so a player with a few big matches needs a higher rating to rank above one with years of them.

How do the cricket rankings work?

• Men's T20 players are ranked on the same ratings that drive our previews. Men's and women's one-day cricket, men's first-class cricket and women's T20s are rated by the same method, refitted on each format with its own par and its own league strengths. Women's Tests are too rare to rate anyone on.
• An international list takes a player who has played an official international in the format in the last year, with enough of a record over the last two: 10 matches batted (6 in first-class cricket) or 240 balls faced for a batter, and 240 balls and 3% of the side's balls for a bowler. They also need a record against Full Member sides over the last three years.
• Men's T20, one-day and first-class rankings open on every player who has played the format in the last year, capped or not. They need the same record over the last two years, and a record over the last three in domestic cricket or in internationals with a Full Member side.
• Each league and domestic competition we cover has its own list, on the same ratings: its clubs' squads and everyone who has played in it over the last two years, with the same record over those two years as any other list.
• Each international list can also be rated on Tests, ODIs or T20Is alone, by the same method, with each player's record in those matches beside them. Ratings on all first-class cricket pick out next year's best Test players better than Tests alone do; in ODIs and T20Is the two come out about level.
• Men's batters are ranked against an average top-seven batter, allowing only for how much harder their slot is for the same batter, and over the top seven's balls a match, so a No. 8 is measured on the same scale as an opener. A player who bowls stays on a men's batting list if they usually bat in the top nine in T20 and one-day cricket, or the top eight in first-class cricket. Women's batters are ranked against par for their slot, with bowlers listed down to No. 7.
• The all-rounders are the players in both a list's batters and its bowlers who usually bat in the top seven and bowl at least a twentieth of their side's balls, or bowl a seventh of them and bat anywhere. They're ordered by what they add to a strong side: for each of the format's eight strongest international sides, the runs a match between the best XI it can pick with them and without them, averaged. An XI needs a keeper, at least four frontline bowlers and no more than four specialist bowlers, and bats its best batters first, wherever each player's own side bats them, so the figure counts the batter or bowler an all-rounder makes room for as well as their own runs and overs.
• Every sides list is in runs a match above an average Full Member side, so most Associates sit below zero. In men's and women's T20s and women's ODIs, a side's strength is the average rating of the XIs it has fielded over the last two years. An XI from six months ago counts half as much as the latest one. Men's one-day and first-class sides are ordered by their chance of beating an average Full Member side on a neutral ground, from the same models as our ODI and Test previews, and shown as the runs a match that give a side that chance.
• The ICC's own rankings sit beside ours. They play no part in the ratings.

Rankings

How are T20 win chances worked out?

• A regression on the gap between the sides: our player ratings, batting and bowling phase by phase set against the competition's scoring, team Elo, recent form, rest days and each side's experience of the host city.
• A neural network that reads both XIs player by player.
• A ball-by-ball simulation that plays the match out hundreds of times, half with each side batting first, under the competition's own rules, including an Impact Player where the league has one.

Match previews

What is a par score, and does batting first matter?

• Par is the first-innings total with an even chance of being defended. In a T20 league it starts from the competition's recent first innings, moves by how the ground has scored against what each competition there expected, and adds the margin a model of every men's T20 since 2012 says a total needs to be defended half the time.
• Since 2016, a first innings at par in the leagues we preview has been defended 48% of the time: 42% in the IPL and 43% in the PSL, where chasing sides win more often, 46–48% in the BBL, BPL and CPL, and 55% in the T20 Blast.
• In a T20I or an ODI, par starts from first innings between Full Members, with recent years counting most, moves by how the host country has scored lately and by the ground's own record, and adds the margin a total needs in that format. In a Test, it starts from first innings in the host country since 2015 and moves by the ground's own record.
• In a T20 league preview, each side's chance batting first or chasing comes from the blend, moved by how often chasing sides have won in that competition. The simulation replays each order thousands of times for the likely first-innings total and the likelier result's winning margin, adjusted to the ground.

Grounds

How are ODI win chances worked out?

How are T20 international win chances worked out?

How are Test win chances worked out?

How are series winners predicted?

How are series predicted before the squads are named?

What else goes into a match preview?

• In a T20 league preview, each side's batting and bowling standouts are its highest-rated players whose rating rests at least half on their own record and who have played in the last year. Weak links are players rated a run a match or more below average in the weakest batting or bowling of the two sides.
• In an international preview, the standouts are each XI's leading run-scorer and wicket-taker in recent matches of the format.
• A batter's expected runs in an international preview start from their recent average and their recent runs a match, scaled to what their place in the order usually makes. In Tests and T20Is the figure then moves with the opposing attack: up against bowlers who have been going for more runs and taking fewer wickets than most, down against a stingy one. Previews are set before the toss, so every figure assumes either side could bat first. Batting first is worth about a tenth more to the top seven, and chasing about a tenth less.
• The sides board sets each XI's batting against the other's attack in three stages of the innings: overs 1–6, 7–15 and 16–20 of a T20, 1–10, 11–45 and 46–50 of an ODI, and in a Test the new ball (overs 1–20), the old ball (21–80) and the second new ball. Each stage is rated on its own balls, the same way as the player ratings, and each player counts for the balls they usually face or bowl in it. A bar shows how many runs an over the batting side is ahead of the attack, compared with an average side meeting an average attack: an average Full Member side in an international, an average side in the league in a T20 league.
• Recent form is a player's last ten matches: men's T20s, The Hundred included, in a league preview, and matches of the format in an international.
• The weather board shows the forecast once a match is within 16 days. Before that, a league preview shows past days at the ground around the same date, at the match's hours, each labelled with its date.
• A T20 league preview's take is drafted by an AI language model from the figures on the page. A draft that quotes a number it wasn't given, contradicts a player's rating or names someone outside the match is sent back with the problems listed, up to three attempts in all. An international preview's take is written by us, and shows only while the call and par it was written for still hold.

How are the most exciting matches picked?

Records

How are a player's strengths and weaknesses picked?

Players

How are a team's strengths and weaknesses picked?

Teams

How accurate are our cricket predictions?

• In the T20 leagues we preview, 861 of those games, it picked 58.9% of winners, with a log loss of 0.665. League favourites won more often than our chances said.
• In the 908 T20 internationals it picked 75.4%, with a log loss of 0.479.
• On the 655 league games with a Betfair price 25 minutes before the start, Betfair's log loss was 0.676 against our 0.671.
• On the 231 T20 internationals with a Betfair price 25 minutes before the start, Betfair's log loss was 0.516 against our 0.519.

Track record