IPL men · first innings · 2023–2025 evidence
Choose what to review.
Keep the uncertainty.
Compare observed bowling profiles with a regularized estimate. Keep up to three cases and the questions you want to investigate.
Fixed historical cohort. The 2026 test is separate. Match context and sample size limit what these signals can tell you.
Loading the published evidence snapshot…
The evidence could not be loaded.
Profiles · overs 1–6
| Player / evidence | Legal balls matches | Dot % raw → reg. | Boundary % raw → reg. |
|---|
No matching profiles
Try another name, phase or evidence filter.
Observed and regularized
○ Observed● Regularized│ Reference
| Outcome | Observed | Regularized | 95% band |
|---|
Difference from the reference
Percentage points, not relative percentage change. A band spanning zero does not separate the profile from this reference. It does not prove equality. No pairwise player comparison is calculated.
Sensitivity & evidence checks
Contributing matches
Source identifiers help you locate footage you have permission to review. No video viewing is implied.
Your working selection
A short list of review questions.
Choose up to three cases yourself. Notes stay in this tab; export a file to keep them.
No cases selected. Open a player above, inspect the evidence, then add a review question.
Selection is for investigation. Limited or uncertain cases retain their warning in every export. A review question is a prompt, not a finding from footage.
Method & evidence
A small model.
A bounded conclusion.
Probabilities describe recorded outcomes in this cohort, not intrinsic talent or the causal effect of choosing a bowler.
What the profiles contain
Dot = zero total runs on a legal delivery. Boundary = a batter four or six, excluding non-boundary flags. Other = the remaining legal deliveries. Illegal deliveries are counted separately. These measures are not economy or credited wickets.
Held-out 2026 check
Retrospective evaluation, separate from the 2023–2025 profiles. Lower Brier is better; it is not an accuracy percentage.
Limits, coverage and provenance
How to read the evidence states
Limited evidence: fewer than 60 legal deliveries or five matches. These are operational thresholds, not a statistical guarantee. No history means a phase fallback, not a personal estimate.
Uncertain: a reference contrast spans zero, or its direction changes after removing one match or the dominant venue. Reviewable: passes those stated checks; it still does not recommend a lineup. Unavailable checks remain visible.
Intervals are conditional on the fixed model and cohort. They are not probabilities of player superiority. Inspecting multiple profiles is exploratory.