How CricArcade Decides a Single Delivery Policy & Terms

Every engine in CricArcade, Test, ODI and T20, draws its randomness from one small seeded generator. A seed is just a starting number, and the same seed always produces the same sequence of rolls, which means the same match, ball for ball. That is what makes a saved game reproducible and a leaderboard run verifiable: nothing is decided by the wall clock, only by the seed and your decisions.

Each batter carries how fast they score and how long they survive; each bowler carries how few runs they concede and how often they strike. For every delivery the engine combines both sides into a scoring chance and a wicket chance, in a way that lets each genuinely influence the ball. A great bowler bowling to a great batter is a real contest: neither one alone dictates the outcome, which is why an elite attack still gets hit and an elite batter still gets a good ball.

A player's overall rating tips the balance, tuned so that a higher rating buys genuine survival and not merely faster scoring, which is the real mark of the best batters. How a career becomes those numbers is covered, at a high level, in how real careers become ratings.

Those two base rates are then multiplied by everything that makes cricket situational: the pitch and its wear, the session, the age of the ball, the batter's pace-versus-spin matchup, left-right hand geometry, how long the batter has been in, the era, and the match situation. These effects stack, in a way you can reason about at the crease. A new batter facing the new ball from a seamer on a green day-one pitch has several things tilted against them at once; a set batter on a flat road in the afternoon has them all working the other way.

With the rates settled, the engine resolves the delivery in order. First it checks for an extra (a wide, no-ball, bye or leg-bye), which is independent of the matchup. Then it rolls for a wicket against the wicket probability. If no wicket falls, it samples the runs from a distribution shaped to the batter's strike rate, so a big hitter's share of fours and sixes is genuinely higher than a blocker's.

When a wicket does fall, a second roll decides how: caught is the most common outcome, then bowled, then lbw, with stumpings available only to spinners and run-outs kept realistically rare. Catches are not automatic, either. Every fielder has a fielding rating that sets a drop chance, so a sharp chance to elite hands is put down far less often than the same chance to the tail.

In the modern Test era the engine adds an umpiring layer over that same wicket roll. A small share of genuine lbw and caught dismissals are missed by the umpire, a matching share of not-outs are wrongly given, and some big appeals are correctly turned down. That is exactly the uncertainty the review system exists to resolve, which is why a review prompt in a Test is never a guaranteed wicket: sometimes the finger was right.

Watch the model produce a full innings in The Follow-On, The Chase or the T20 Franchise Manager.