Picking the next game: bets and evidence
A bet, not a fact
Every new game is a bet: the team spends months on a guess that players will love it.
Good studios say the bet out loud, together with the evidence that would prove it wrong. Kai writes Shard Garden's: "We bet that friends will come back to a shared garden every day. If fewer than 35% of playtesters come back the next day, we change the game before spending more."
Cheap evidence first
Evidence gets more expensive as the game grows, so studios collect the cheap kinds first:
- A rough prototypeA day or two of work. Does the core idea feel fun at all?
- A playtestWatch five real players. Do they ask to play again?
- A store pageA Steam page marked "coming soon" collects wishlists: players who ask Steam to tell them when it launches.
- A soft launchRelease in a few countries first, and measure day-1 retention: the share of players who come back the next day.
Quick check
Which evidence costs the studio the least?
- Five friends playing a rough prototype
- A soft launch in three countries
- A finished console version
Show the answer
Five friends playing a rough prototype
A rough prototype takes days. A soft launch needs a working game, servers and store pages, and a console version needs the whole game.
True story: champagne for a stopped game
Supercell, the Finnish studio behind Clash of Clans, released its first game, Gunshine, in 2011. Its next, Pets vs. Orcs, had a soft launch on phones. Players did not come back often enough, so the studio shut it down.
Co-founder Ilkka Paananen brought the team champagne: "I'll just get a few bottles of champagne and I'll just give them to the team." Clash of Clans came out in August 2012.
Stopping a game that misses its numbers early saves the money for the next bet.
Source: Sequoia Capital: Supercell ft. Ilkka Paananen, sequoiacap.com
Write the stop rule first
If the team picks its stop number after seeing the results, it is tempting to move the goalposts: "31% came back. That is nearly 35%!"
Writing the rule down before the test keeps everyone honest. The rule is called a kill rule: the number below which the game is changed or stopped. Games analyst Joost van Dreunen describes Supercell checking each new game's numbers every day during a soft launch.
Quick check
It is fine to choose the stop number after the soft-launch results come in.
- True
- False
Show the answer
False
Once you have seen the results, any number can be made to look good. A kill rule only works if it is written down before the test.
Read the numbers
1Kill rule: below 35%2 3Country A 31%4Country B 29%5Country C 33%
Kai wrote the kill rule before launch. What does the evidence say?
- The bet missed in every country, so the rule says change or stop this version
- Country C is close enough, so the bet passed
- Three countries mean nothing, so carry on as planned
Show the answer
The bet missed in every country, so the rule says change or stop this version
All three are below the 35% the team chose before it looked. The rule does not say the idea is worthless. It says this version should not get the big budget yet.
On the studio board
The studio writes its first decision on the studio board, where every later plan can see it:
Shard Garden's bet: four friends will come back to a shared garden. Kill rule: under 35% day-1 retention in playtests, we change the game before spending more.
Bolt says: Next: the design doc, where the bet becomes a plan.
Fact-checked 12 October 2026
- Sequoia Capital: Supercell ft. Ilkka Paananen sequoiacap.com
- Steamworks: Wishlists partner.steamgames.com