by rbelikov
DbZer0 wrote:
FYI, in 4892 games, the current win-ratio is:
Corporation Victories: 2365
Runner Victories: 2527
Statistical analysis:
Assuming balanced sides, expected mean runner or corp victories = 4892/2 = 2446. Actual victories are 81 higher for runner, 81 lower for corp.
Expected standard deviation (1 sigma) from 4892 games = sqrt(4892/4) = 35. Thus, we have a 81/35 = 2.3 sigma deviation. (Given balanced sides, the chances of this happening are about 3%.) So, depending on your threshold of statistical significance, this may or may not be statistically significant. (Let's call it marginally significant.)
In terms of percentages, 1 standard deviation is 35/4892 = 0.7%. Therefore, with 1-sigma error bars, the measured win rates are:
Corp: 2365/4892 = 48.3% +/- 0.7%
Runner: 2527/4834 = 51.7% +/- 0.7%
Hollis brings up some excellent points about bias. In particular, I think that having games with players of enequal skill skews the win rates towards 0.5 (because the stronger player will win both with corp and runner). Even so, I expected more of a deviation from 50%, and am deeply impressed that an asymetrical game has this level of balance.