How We Predict & Grade

1. Where the data comes from

Fixtures, squads, live scores and scorecards come from third-party cricket data providers (CricAPI / cricketdata.org and Highlightly). Venue records, head-to-head histories and player form are then derived from the scorecards we have collected ourselves over time, not from memory or guesswork.

2. The anti-invention rule

Every player we name must come from a real squad or playing XI observed for that specific match. The model is given that list and is not permitted to add anyone to it — no famous names dropped in because they "should" be playing. When a confirmed XI is published at the toss, we regenerate the prediction against it.

If no genuine squad is available for a fixture, we publish the fixture with no prediction. We would rather show nothing than show something invented.

3. Building the XI

The model receives the squads, each player's recent fantasy-point average, venue scoring patterns, head-to-head record, weather and the match format. Format matters a great deal: in a T20 strike rate and economy carry bonuses, while in a Test or County Championship match bowlers bowl in both innings and a five-wicket haul routinely outscores a century. The selection strategy changes accordingly.

4. How grading works — the strict part

After a match ends we fetch the full scorecard and score every player who took part using the Dream11 points system for that format:

• T20 — runs, boundary and six bonuses, 30/50/100 milestones, 25 per wicket, plus strike-rate and economy bonuses.
• ODI — 50/100 milestones, 25 per wicket, wider strike-rate and economy bands.
• Test / First-class — 16 per wicket, 50/100 milestones, and no strike-rate, economy or maiden points.

We then rank every player by points and take the real top eleven. Our prediction scores one point for each of our eleven that appears in that list.

This is deliberately a harsh bar. A player can be picked, play the whole match and still count as a miss if ten others out-scored him. We grade against the actual top eleven by fantasy points, not against the playing XI — which is why our published average looks lower than the numbers other sites advertise. It is the honest measurement, and it is the same bar applied to every match.

5. What we publish

Every graded match appears on the accuracy page with its score and the players we got right, and each has its own review page showing our XI beside the actual top eleven. Nothing is removed for looking bad.

Where a match cannot be graded — because no data provider publishes a per-player scorecard for that competition — we mark it as having no result available rather than guessing at one or quietly dropping it.