In 1-game matches, or eg in a 7-game match that's tied 6-6, you care only about winning, not how many points you win.
I added the ability to Player 2.1 to optimize for that instead of optimizing for best full equity.
As expected, it increased the win %-age a little at the expense of equity. I experimented with this in 30k-game matches:
Optimizing for any win: score +0.3497ppg, 66.82% chance of a win.
Optimizing for equity: score +0.4686ppg, 66.01% chance of a win.
So optimizing for any win does improve the odds a little (around 0.8%) but dramatically reduces equity.
I added the ability to Player 2.1 to optimize for that instead of optimizing for best full equity.
As expected, it increased the win %-age a little at the expense of equity. I experimented with this in 30k-game matches:
Optimizing for any win: score +0.3497ppg, 66.82% chance of a win.
Optimizing for equity: score +0.4686ppg, 66.01% chance of a win.
So optimizing for any win does improve the odds a little (around 0.8%) but dramatically reduces equity.
No comments:
Post a Comment