How our cricket research works

Who and how

This is an independent Reddy News data section. Match tables are assembled automatically from the live cricket API already used on this site. Explanatory content and derived measures are clearly separated from provider facts. It is not an official page of the teams, competitions or any betting operator.

Sources and freshness

Our API proxies return live scores, date fixtures, match overview, information, summary, team lists and innings scorecards from the connected LiveScore public feed. The page displays the retrieval time in IST. The initial HTML includes a saved snapshot; JavaScript refreshes live sections every 30 seconds and retains existing data when a request fails.

Recent form and head-to-head

Only earlier completed records in the same format group are used. T10, T20 and 100-ball formats are not combined. First-class and Test cricket are separate from one-day form. W, L, D, T and NR mean win, loss, draw, tie and no result; a question mark means the result note could not be classified safely.

Venue summaries

A first-innings median requires at least eight completed scored matches at the same venue in the same format group. We publish the sample count. No all-time record is claimed from this limited dataset, and no batting-first or toss-win percentage is calculated without confirmed toss and result history.

Team ratings and winning estimates

RN-Elo-v1 is an original experimental Reddy News calculation, not an API-provided prediction, official ranking or betting rate. Every team starts at 1500 in each format. Results move points using K=24 for T20/T10/100-ball, K=28 for one-day and K=30 for first-class/Test. No home-side bonus is added because neutral venues are not reliably identified. Ties, draws, no-results and notes whose winner cannot be safely identified do not update ratings.

Before the result update, raw chance = 1 / (1 + 10^((opponent rating − team rating)/400)). The deviation from 50% is multiplied by min(1, the smaller team sample / 20). A winning bar requires at least 3 earlier decided results for each side in the same format. This is a low minimum for an exploratory display, not evidence of statistical reliability; fewer than 20 remains low-confidence. Test matches receive no two-way percentage because draws require a separate model. No live in-play chance or toss winner is inferred.

Historical coverage and chronological testing

The date API allows a 30-day window. Our initial capture requested the 29 completed dates before 8 October 2026 and stores only returned completed results. Newly completed matches already returned by live refresh are included on subsequent calculations. This is captured-sample history, not career or all-time coverage. Missing or unclassifiable outcomes are excluded and team aliases are normalized conservatively without merging A, youth, women's or franchise teams.

For limited-overs replay, all estimates on a UTC date are calculated before that day's results update the ratings. Fixture estimates use only records from earlier UTC dates, conservatively excluding same-day results. This prevents same-day result order from leaking into a pre-match estimate. First-class ratings describe completed records only and are not used for winner probabilities. The replay is an exploratory diagnostic on a small window, not a long-term out-of-sample validation or independently verified backtest.

Brier is the mean squared probability error; lower is better. Log loss penalizes confident errors. A 50/50 model scores 0.250 Brier and 0.693 log loss. Favourite accuracy is compared with the first-listed-team baseline, with half credit for equal chances. Calibration compares average probabilities with actual outcomes within bands. Headline performance figures are withheld below 30 evaluated estimates, and every small calibration sample remains unreliable even if its outcome percentage looks impressive. The model has not established a predictive edge.

Model limits

Ratings use result notes only. They do not know injuries, pitch, weather, toss, playing XI or ball-by-ball state. A strong rating in one competition may not be comparable with a disconnected competition. Published probabilities are experimental estimates, not calibrated likelihoods or guarantees. Match timing and source coverage can change; the interface shows its data timestamp.

Player and scorecard limits

Available team lists are marked unconfirmed unless the provider explicitly confirms the XI. Scores use the provider’s displayed values. Cricket overs are converted to legal balls before calculating runs per over. Player pages contain only contributions found in the returned records; absent portraits, biographies or career statistics stay absent.

Corrections and accountability

For a mismatch between source records and a page, please use our corrections process. Read our editorial standards and return to the analysis hub.