Every versus game hides a quiet question. Is this fair? When both players are equally good, who wins — the better strategist, or just the one who moved first? We do not guess at this. Before an original ships, we simulate thousands of matches and measure. Here are the real numbers, and what they changed.
The first-move problem
Turn-based games love the player who starts. Our first version of Tilt proved it: in even matches, the first player won 74 to 79 percent. We threw that whole design away. The Tilt you can play today came from the redesign, and it measures close to even.
The obvious fix that did nothing
1–2PCasualButton
Closest stone to the button takes the end — and the last stone alternates.
▶ Play NowButton is our flick-curling game. In testing, the side that started with the hammer — the last stone of a round — won 74 percent. We blamed the big multi-point rounds. So we tried capping every round at one point. The cap barely moved the number.
The real fix was smaller: alternate the hammer every round, no matter who scores. First-mover wins fell from 74 percent to 49.1. Good thing the cap failed, too. Big rounds are the reward for placing your stones well. Cut them and you cut the skill.
The advantage everyone "knows" — that was not there
Spill is a territory duel. Games like it are famous for first-move advantage. So we planned a catch-up rule, called komi: free points for the second player, like in Go. Then we measured. Across three boards, four stages and three play styles, the first player won between 42 and 53 percent. No edge at all. Spill shipped with no komi, because the folklore was wrong. Only the measurement could tell us that.
When players bring different brains
Sleight is a memory duel, and raw memory duels are brutal. We measured a plain version first. One extra item of working memory won 69.5 percent of games. Three extra items won 96.6. You cannot ship that to two people on one couch. Memory span is something you arrive with, not something you can practice by Friday.
That is why Sleight's board shuffles itself. When tiles trade places, a big memory becomes a big list of things to re-track. With the shuffle on, the one-item gap drops to 54.9 percent. And someone who does not watch the board still loses 97.9 percent of the time. Attention decides the game now, and attention is a choice.
Reflexes should bend the game, not settle it
Wake is a racing duel where the fastest seat is second place. We measured what actually decides a match. Reacting 0.3 seconds late costs about 2 points of win rate. Picking the wrong strategy costs between 10 and 50. Decisions matter roughly ten times more than reflexes. That ratio is on purpose, and we tuned the game until it was true.
What "fair" means to us
Not a coin flip. A perfectly even game with no skill in it is just dice with extra steps. We aim for three things. The starting seat should not decide the match. The better strategy should win clearly. And a gap in reflexes or memory should bend the result, not settle it. When a design cannot reach all three, we kill it — those stories are here.
Want more of the how? Read how we build our originals, and how our CPUs play fair.
