Implement state distribution in result set. #742
Conversation
Closes #738 Add state distribution and normalised state distribution to the result set. Also adds the normalised state distribution to the summary.
| @@ -326,6 +326,26 @@ def build_ranking(self): | |||
| return sorted(range(self.nplayers), | |||
| key=lambda i: -nanmedian(self.normalised_scores[i])) | |||
|
|
|||
| def build_normalised_state_distribution(self): | |||
There was a problem hiding this comment.
Needs an underscore? Seems like we're being inconsistent with the build* functions
| for player in self.state_distribution: | ||
| counters = [] | ||
| for counter in player: | ||
| total = sum(counter.values(), 0.0) |
There was a problem hiding this comment.
Is the 0.0 necessary here? Is it possible for sum to be zero?
There was a problem hiding this comment.
Looks like it wasn't needed. Have removed it.
| @@ -401,6 +423,13 @@ def _update_cooperation(self, p1, p2, cooperations): | |||
| self.cooperation[p1][p2] += cooperations[0] | |||
| self.cooperation[p2][p1] += cooperations[1] | |||
|
|
|||
| def _update_state_distribution(self, p1, p2, counter): | |||
| self.state_distribution[p1][p2] += counter | |||
There was a problem hiding this comment.
Added this and a bunch of others that were missing (whilst doing this I got annoyed at my prior lazy self).
| def _update_state_distribution(self, p1, p2, counter): | ||
| self.state_distribution[p1][p2] += counter | ||
|
|
||
| counter[('C', 'D')], counter[('D', 'C')] = (counter[('D', 'C')], |
There was a problem hiding this comment.
I like the swap but not the line break.... I guess there's no getting around that, but maybe it would be more clear if broken after the = sign
There was a problem hiding this comment.
Could shorten with C, D = actions.C, actions.D at the top of file...
| "CC_rate", "CD_rate", "DC_rate", | ||
| "DD_rate"]) | ||
|
|
||
| states = [('C', 'C'), ('C', 'D'), ('D', 'C'), ('D', 'D')] |
There was a problem hiding this comment.
Could use C, D = actions.C, actions.D
| for i, player in enumerate(self.normalised_state_distribution): | ||
| counts = [] | ||
| for state in states: | ||
| counts.append(sum([opp[state] for j, opp in enumerate(player) |
There was a problem hiding this comment.
Two lines here? One for the sum, one for the append so there's no line break
| counts.append(sum([opp[state] for j, opp in enumerate(player) | ||
| if i != j])) | ||
| try: | ||
| counts = [c / sum(counts) for c in counts] |
There was a problem hiding this comment.
Any change for an integer division error here?
There was a problem hiding this comment.
I don't think so but I've added a property based test to just run a bunch of tournaments and attempt to summarise them...
There was a problem hiding this comment.
What did you mean by an integer division error? (I might have misunderstood.)
There was a problem hiding this comment.
I know what you mean. Am pushing a better hypothesis test and a fix.
| self.cooperating_rating, | ||
| median_wins)] | ||
| summary_data = [self.player(rank, *summary_data[i]) for | ||
| summary_data = [self.player(rank, *(list(summary_data[i]) |
There was a problem hiding this comment.
This is a bit hard to follow... could use a readability upgrade?
| @@ -20,6 +20,10 @@ This tutorial will show you how to access the various results of a tournament: | |||
| - Payoff difference means: the mean score differences. | |||
| - Cooperation counts: the number of times each player cooperated. | |||
| - Normalised cooperation: cooperation count per turn. | |||
| - Normalised cooperation: cooperation count per turn. | |||
| - State distribution: the count of each type of state of a match | |||
There was a problem hiding this comment.
Might be good to explain what we mean by state distribution -- that it's two rounds of play rather than some other amount.
There was a problem hiding this comment.
I've added that in the wording below. Repeated it for both which might be a bit overkill, happy to tweak further :)
- Include underscore for all build methods. - Remove 0.0 from state distribution sum. - 'C'/'D' -> C/D - Single line for swap. - Probability count on one line. - Improve readability of summary measures.
Closes #738
Add state distribution and normalised state distribution to the result set.
Also adds the normalised state distribution to the summary.