Post by chriscrawford on Apr 25, 2016 20:33:55 GMT -8
The time has come for me to devise the algorithms that decide whom to attack and what auragon to use. This will be complicated.
The primary goal is to figure out the attack that is most likely to win. I see several possible strategies here.
One is to attack the opponent who is most likely to have zero auragons of one color. That requires only searching for anybody with a perceived auragon count of zero and the highest certainty. But this strategy may fail if the algorithm finds nobody with a perceived auragon count of zero. And the algorithm might decide that the certainty is too low to make this choice advisable.
The second best algorithm is to examine all three perceived auragon counts and look for a big differential. If one opponent has a great many auragons of one color, then he is likely to use one of those, so the attack should go against that color.
These two ideas merge into a single concept: find the extrema and act accordingly. If you are confident of the most or least, then attack based on that. But what if neither of these works well? Is there some way to evaluate all three perceived auragon counts and act accordingly? I don't see any such strategy. I'm sure that there is a formally correct way to handle it, but it'll be a mess.
OK, so the attack algorithm is to find either the lowest auragon count or the highest auragon count (weighted by certainty). But now that decision should be weighted by the personal relationship. That is, the character will need to balance the probability of success against other factors, such as how much he likes the candidate and how fearful he is that the candidate is close to winning.
And what about the defense algorithm? In that case, it's a simple calculation of the relative perceived values of auragon counts. Find the highest count color and assume that it is the attacking color.
Time for bed. I'll sleep on it and decide in the morning.
The primary goal is to figure out the attack that is most likely to win. I see several possible strategies here.
One is to attack the opponent who is most likely to have zero auragons of one color. That requires only searching for anybody with a perceived auragon count of zero and the highest certainty. But this strategy may fail if the algorithm finds nobody with a perceived auragon count of zero. And the algorithm might decide that the certainty is too low to make this choice advisable.
The second best algorithm is to examine all three perceived auragon counts and look for a big differential. If one opponent has a great many auragons of one color, then he is likely to use one of those, so the attack should go against that color.
These two ideas merge into a single concept: find the extrema and act accordingly. If you are confident of the most or least, then attack based on that. But what if neither of these works well? Is there some way to evaluate all three perceived auragon counts and act accordingly? I don't see any such strategy. I'm sure that there is a formally correct way to handle it, but it'll be a mess.
OK, so the attack algorithm is to find either the lowest auragon count or the highest auragon count (weighted by certainty). But now that decision should be weighted by the personal relationship. That is, the character will need to balance the probability of success against other factors, such as how much he likes the candidate and how fearful he is that the candidate is close to winning.
And what about the defense algorithm? In that case, it's a simple calculation of the relative perceived values of auragon counts. Find the highest count color and assume that it is the attacking color.
Time for bed. I'll sleep on it and decide in the morning.