TipKhel · Method
How our model works
Open, simple and graded in public. TipKhel shows probabilities for discussion. They are not promises, and TipKhel has no betting, odds or ads.
01
Ratings from every result
Each cricket team has an Elo rating per format (T20, ODI, Test), and each football club has one per league. Ratings move after every result: beating a stronger side moves you more than beating a weaker one.
02
Context adjusts the base
Cricket reads add toss, venue history (bat-first win rate, average first-innings score) and recent form. Football turns the rating gap and home advantage into expected goals for each side, then into home, draw and away chances.
03
We abstain when it is close
If the stronger side is under 56%, or the data is thin, TipKhel publishes both probabilities and makes no call. Those matches are not counted as hits or misses.
04
Frozen before the start
The published read is frozen at the toss or kick-off. We never edit a prediction after the match starts, so the record can't be cherry-picked.
05
Live estimate during play
In limited-overs cricket, the live win graph updates each over from runs, wickets and balls left, using a resources model shaped like Duckworth-Lewis. It is labelled as an estimate and never replaces the frozen pre-match read.
06
Graded automatically
When a match is settled, the call is marked right or wrong automatically and added to the public record, along with calibration (Brier score). Tip sources are graded the same way.
What a probability means
A 65% read should come true about 65 times in 100, which also means it should be wrong about 35 times. That's why we track calibration, not just hit rate. See every graded call on the accuracy record.