An equity that ignores the doubling cube, as if nobody could double; the analysis board and checker rollouts value plays this way in money play with the cube in the middle.
An equity that accounts for the doubling cube and who can use it, which is what cube decisions, cube rollouts and game reviews are judged by, and checker plays on the analysis board when the match score, an owned cube or the Jacoby rule matter.
Performance rating: the average equity you gave up per non-forced decision, times 500, where lower is better and under 5 is world class; twice the PR is GNU Backgammon's error rate.
How far the engine looks ahead: D0 judges the position right after the move, and each extra level averages over one more roll, so higher depths are slower and stronger.
Playing a position out many times with the engine, to the end or for a few moves, and averaging the results, which is slower than a search but usually more reliable; most puzzle answers on this site come from rollouts, and the analysis board can run them too.
A correction that takes the luck of each roll out of a rollout's games, so the same number of trials gives a much tighter result without tilting it either way.
An equity that ignores the doubling cube, as if nobody could double; the analysis board and checker rollouts value plays this way in money play with the cube in the middle.
An equity that accounts for the doubling cube and who can use it, which is what cube decisions, cube rollouts and game reviews are judged by, and checker plays on the analysis board when the match score, an owned cube or the Jacoby rule matter.
Performance rating: the average equity you gave up per non-forced decision, times 500, where lower is better and under 5 is world class; twice the PR is GNU Backgammon's error rate.
How far the engine looks ahead: D0 judges the position right after the move, and each extra level averages over one more roll, so higher depths are slower and stronger.
Playing a position out many times with the engine, to the end or for a few moves, and averaging the results, which is slower than a search but usually more reliable; most puzzle answers on this site come from rollouts, and the analysis board can run them too.
A correction that takes the luck of each roll out of a rollout's games, so the same number of trials gives a much tighter result without tilting it either way.