Vishal Jaiswal Builder of CricArcade

There is no studio behind CricArcade and no team. Every game, every player rating, every scorecard page and every guide you read here was built by one person. That has a cost: things take as long as they take, and when something breaks there is nobody else to blame. It also has an upside. Every decision about how a game plays goes through the same pair of eyes, and that consistency is most of what makes a cricket simulation feel right.

My day job is in data. I have spent more than 15 years in analytics, data strategy and AI, building products on real-world data and working out what the numbers are actually saying. That habit runs through everything on this site. Every player rating starts from a real career record, and the 2,537 Test and 4,981 ODI scorecards in the database are real matches, not generated ones.

The first commit went in on 12 June 2026, and it was built around a single question: could an all-time World Cup XI chase 380? Pick your side from real World Cup squads, walk out needing 380, and find out whether your selection holds up ball by ball. That game became The Chase.

Once the question worked in one format, the obvious next step was to ask it in the others. Five days later the T20 game was underway, built on the same foundations but living in a different world: auctions, salary caps and franchise squads. A month after that came Test cricket, the hardest of the three to get right, because five days gives a bad model nowhere to hide.

1. 12 June 2026The Chase launches: draft a World XI and chase 380.
2. 17 June 2026The T20 franchise manager starts, with real auctions and squads.
3. 11 July 2026The Follow-On brings five-day Test cricket, with every Test player since 1877 to pick from.
4. 3 to 10 August 2026Three sites become one. The games move onto cricarcade.com with every saved career carried across, so one account, one squad and one leaderboard work everywhere.
5. Since thenDepth, not just breadth: World Cup careers on a real international calendar, IPL dynasties that can begin in any season from 2008, the Big Bash, a Super Over for tied knockouts, and a searchable archive of real scorecards, players and venues.

Every change to CricArcade is recorded as a commit in its code history, across the three original game repositories and the one that now runs the whole site. The charts below are drawn straight from that history. Nothing is estimated.

It is not a steady line, and it was never going to be. The tall bars cluster around launches, when a new game is being finished, and around the weeks after them, when players arrive and start finding the things I missed.

The gold slice is the biggest for a reason. Merging three games onto one site meant rebuilding the shared parts (accounts, saves, leaderboards, the archive) once, properly, and every game has been improved in that one place since.

The honest measure of a project like this is not how much gets written in a burst; it is whether someone keeps turning up. 84 out of 104 days have at least one change on them. As of 23 September 2026, the site runs on about 141,000 lines of game and site code, backed by 703 automated tests.

A cricket simulation is only worth playing if it behaves like cricket. So before any game is launched, I play it. Not a quick look: many full matches, in every mode, with sensible sides and silly ones. Then I take each part of the match on its own and ask one question of it: is this how a real match would go? The new ball, the middle overs, the chase at the death, a declaration, a collapse, a tail that hangs on. If a part does not read like real cricket, the game does not ship until it does.

Playing is how the problems that numbers alone would miss get caught. A few examples:

• The batter nobody could dismiss.After one change to the IPL game, an all-time squad's Virat Kohli became almost impossible to get out. An ordinary season looked fine; it only showed when I picked the strongest XI I could and played it. It was found by playing, and fixed.
• Too many wickets in T20. Measured against real seasons, early T20 matches were taking noticeably more wickets than real ones do, even though the scoring looked right. It was corrected against the real record, not by feel.
• Every ODI era, not just today's. One-day cricket in the 1970s and one-day cricket in the 2020s are different games. Scoring in each era is checked against what real matches of that era produced, so a 1983 chase feels like 1983.

How the engine does this stays private; it is the part of CricArcade that took the longest to get right. What it is judged against is public: real matches, and whether the cricket it produces looks like them.

However many matches I play, players play more. Most of the fixes since launch started as a message on Discord. One from early September is typical: a few players said their World Cup careers had stopped moving. When I checked, the next-fixture logic was reading matches in the wrong order, and one career had 30 unplayed matches stuck behind a tournament that could not finish. Dozens of accounts were affected. The fix was checked against the affected careers themselves before it shipped, and play resumed where it had stopped.

That loop, report to fix to published note, is how the site keeps improving. More than a third of all commits (248 of 657) are fixes, and everything players notice is written up in plain language in updates.

I use AI tools, and I would rather say so plainly than have you wonder. They are used only to enhance the work: to help me write and review code faster, to check data at a scale one person cannot manage by hand, and to tighten the wording of pages. They do not decide how the games play, and nothing they produce goes live on its own.

The judgement is mine. I play-test the match engine myself, decide what the ratings should be, and read and edit every page before it is published. Figures quoted in the guides and team pages are checked against the scorecards in the archive, and when a source is wrong, it is corrected only against a second source that says otherwise.

If a stat, a rating or a result looks wrong, tell me through the contact page or on Discord. I read everything, and fixes are logged in updates.

Elsewhere: personal site, LinkedIn, GitHub and X.