ML-POWERED IMPACT ENGINE
XGBOOST + SHAP

Measure the realimpact of cricket.

Not averages. Not strike rates. A single score (0-100) that captures how much a player actually changes the game — weighted by context, pressure, and match stakes.

P×C×W×K0-100
PerformanceContextPressureKnockout

1,169

Matches

768+

Players

71

Knockouts

8

Seasons

METHODOLOGY

Four variables. One score.

Every player's impact is measured through performance, situation context, pressure intensity, and match stakes.

P
PERFORMANCE

Raw on-field output — what the player actually did with bat or ball.

Runs scored & strike rate
Boundaries (4s & 6s)
Wickets & economy rate
Dot balls bowled
C
CONTEXT

Phase weights differ by role. A powerplay wicket is gold for bowlers; death-over runs are gold for batsmen.

Batting: PP 1.0x, Mid 1.15x, Death 1.4x
Bowling: PP 1.3x, Mid 1.0x, Death 1.5x
Chasing bonus for batsmen only
Defending bonus for bowlers only
W
PRESSURE

Clutch moments. High RRR, few wickets left, and low-total defense raise the stakes.

High required run rate
Few wickets remaining
Defending low totals
Late match stage
K
KNOCKOUT

Playoff performances matter more. Finals and eliminators carry higher weight.

Qualifier 1 & 2
Eliminator matches
Finals
1.15x multiplier
OUTPUT
P×C×W×Knormalized 0-100
ROLLING 10 INNINGS | RECENCY WEIGHTED | ML-CALIBRATED
EVOLUTION

From guesses to data-driven weights.

We started with hand-tuned rules. Then trained an XGBoost model on 1,169 real IPL matches to discover what actually predicts winning.

v1

Rule-Based (Before)

Hand-tuned coefficients

Battingruns x 0.5 + SR x 0.2 + 4s x 2 + 6s x 3
Bowlingwickets x 15 + dots x 1 + eco bonus
Phase weightsSame for bat & bowl
Death overs1.5x (both)
Middle overs1.2x (both)
High RRR1.4x when RRR > 10

Educated guesses - no validation

v2

ML-Trained (After)

XGBoost learned from 1,169 matches

BattingWeighted by 21 features
BowlingWeighted by 18 features
Phase weightsSplit by bat vs bowl
Death (bat/bowl)1.4x / 1.5x
PP (bat/bowl)1.0x / 1.3x
High RRR1.08x (lower than guessed)

0.78

Batting R²

0.60

Bowling R²

200

SHAP Players

Knockout Aware

Playoff finals and eliminators carry 15% more weight. A 50 in the final is not the same as a 50 in a league game.

71 KNOCKOUT MATCHES

Defending Pressure

Bowlers defending low totals with tight economy get a pressure bonus. Keeping it tight at 140 defense matters.

ECONOMY-BASED BONUS

Fully Explainable

Every score is broken down with SHAP. See exactly which stats pushed a player's impact up or down.

200 PLAYERS WITH SHAP

Plug & Play ML

Standalone Python template with CLI batch mode and REST API. Feed CSV or send JSON — get impact scores anywhere.

CLI + REST API
IQ
CricketIQ Impact Intelligence
1,169 Matches · 768 Players · 71 Knockouts · 2017-2024