2for22

Caribbean Premier League

How often our predictions came true, graded against 38 finished matches in the archive.

Pre-match favourite won
49%
Rerun after the fact 18 of 37
In-play: the side we had ahead won
74%
133 of 180 readings, after 5, 10 and 15 overs of each innings
Projected total after 10 overs
±21 runs
off the final total on average, over 32 innings

Before the first ball

Did the side our pre-match forecast favoured go on to win?

Saved before the first ball
Saving began on 27 September 2026. The first saved forecast is graded when its match finishes.

The true record: taken once both XIs are named and kept unchanged, whatever happens.

Rerun after the fact
49%
Called 18 of 37 matches right

The archive played again now. Our player data already includes most of these matches, so this runs a little flattering.

Who's ahead, as the match is played

After 5, 10 and 15 overs of each innings: did the side our win % had ahead go on to win?

50%
coin toss
66%
5 ov
32
69%
10 ov
32
68%
15 ov
31
77%
5 ov
31
87%
10 ov
30
79%
15 ov
24

The figure under each bar is how many matches were graded there. A chase won before 15 overs has no 15-over reading.

When it said…

Every in-play reading above, grouped by how sure it was. A well-judged win % lands on its tick: of the times it said about 70%, that side should have won about 70% of them.

Said 50–60%
won 64% of 44
Said 60–70%
won 53% of 34
Said 70–80%
won 67% of 33
Said 80–90%
won 88% of 34
Said 90–100%
won 100% of 35

The projected first-innings total

How far the projection was from the total the side actually made.

After 5 overs
±26
runs off, on average
within 10 runs
22%
within 20 runs
47%
No lean either way · 32 innings
After 10 overs
±21
runs off, on average
within 10 runs
22%
within 20 runs
50%
No lean either way · 32 innings
After 15 overs
±13
runs off, on average
within 10 runs
48%
within 20 runs
77%
Tends to run 3 low · 31 innings

Year by year

Tap a row to see it on its own.

Year Matches Pre-matchsaved Pre-matchrerun In-playside ahead won Projectionafter 10 ov
2026 38 — 49%18/37 74%133/180 ±2132 inns

Faded figures rest on fewer than 10 matches.

How this is graded
  • Pre-match forecast. Graded on the percentage the forecast tab showed: since 27 September 2026 the 500 simulated matches weighed for how sure they have proved and combined with each side's team strength, with a player the data does not know stood in as a newcomer; before that, the plain share of the 500 won. It counts as right when the side it favoured won. A 50–50 forecast makes no call and isn't counted.
  • Saved and rerun are kept apart. From 27 September 2026 each forecast is saved once both XIs are named and before the first ball, and it is never changed. Matches before that were played again afterwards, on today's forecast and on player data that already includes most of them, which flatters the rerun a little.
  • In-play win %. Read exactly as the match's own screen shows it, at the end of the 5th, 10th and 15th overs. It counts as right when the side it had ahead (over 50%) won. A reading taken on the ball that ended an innings isn't counted: by then the match was decided.
  • Projected total. The first innings only (a chase has a target instead), against the total the side actually made.
  • Left out. Matches with no winner (abandoned, no result, tied), 1 here. Rain-shortened matches are left out of the in-play grades (5 here), because their readings are now worked out on a length nobody knew at the time.
  • Small samples. A figure resting on fewer than 10 matches is marked, and faded in the table. One lucky match shouldn't read like a trend.