Fairshare between two and more
Add an extension showing the relative fairness correlation of this selection algorithm. An absolutely fair algorithm would have a correlation of 1 for each person (no person has an advantage or disadvantage due to an algorithmic artefact.) This algorithm is fair, and gets better the more iterations are done.
A lower correlation factor corresponds to an advantage, higher to a disadvantage, the closer to 1 it is, the fairer the algorithm. Absolute best possible advantage correlation is 0. Absolute worst is 2.
sub fairshare (\b) { ^∞ .hyper.map: { .polymod( b xx * ).sum % b } }
.say for <2 3 5 11>.map: { .fmt('%2d:') ~ .&fairshare[^25]».fmt('%2d').join: ', ' }
say "\nRelative fairness of this method. Scaled fairness correlation. The closer to 1.0 each person
is, the more fair the selection algorithm is. Gets better with more iterations.";
for <2 3 5 11 39> -> $people {
print "\n$people people: \n";
for $people * 1, $people * 10, $people * 1000 -> $iterations {
my @fairness;
fairshare($people)[^$iterations].kv.map: { @fairness[$^v % $people] += $^k }
my $scale = @fairness.sum / @fairness;
my @range = @fairness.map( { $_ / $scale } );
printf "After round %4d: Best advantage: %-10.8g - Worst disadvantage: %-10.8g - Spread between best and worst: %-10.8g\n",
$iterations/$people, @range.min, @range.max, @range.max - @range.min;
}
}
Output:
2: 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 0, 1, 1, 0, 1, 0, 0, 1, 0, 1, 1, 0, 0
3: 0, 1, 2, 1, 2, 0, 2, 0, 1, 1, 2, 0, 2, 0, 1, 0, 1, 2, 2, 0, 1, 0, 1, 2, 1
5: 0, 1, 2, 3, 4, 1, 2, 3, 4, 0, 2, 3, 4, 0, 1, 3, 4, 0, 1, 2, 4, 0, 1, 2, 3
11: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 0, 2, 3, 4
Relative fairness of this method. Scaled fairness correlation. The closer to 1.0 each person
is, the more fair the selection algorithm is. Gets better with more iterations.
2 people:
After round 1: Best advantage: 0 - Worst disadvantage: 2 - Spread between best and worst: 2
After round 10: Best advantage: 1 - Worst disadvantage: 1 - Spread between best and worst: 0
After round 1000: Best advantage: 1 - Worst disadvantage: 1 - Spread between best and worst: 0
3 people:
After round 1: Best advantage: 0 - Worst disadvantage: 2 - Spread between best and worst: 2
After round 10: Best advantage: 0.99310345 - Worst disadvantage: 1.0068966 - Spread between best and worst: 0.013793103
After round 1000: Best advantage: 0.99999933 - Worst disadvantage: 1.0000007 - Spread between best and worst: 1.3337779e-06
5 people:
After round 1: Best advantage: 0 - Worst disadvantage: 2 - Spread between best and worst: 2
After round 10: Best advantage: 1 - Worst disadvantage: 1 - Spread between best and worst: 0
After round 1000: Best advantage: 1 - Worst disadvantage: 1 - Spread between best and worst: 0
11 people:
After round 1: Best advantage: 0 - Worst disadvantage: 2 - Spread between best and worst: 2
After round 10: Best advantage: 0.99082569 - Worst disadvantage: 1.0091743 - Spread between best and worst: 0.018348624
After round 1000: Best advantage: 0.99999909 - Worst disadvantage: 1.0000009 - Spread between best and worst: 1.8183471e-06
39 people:
After round 1: Best advantage: 0 - Worst disadvantage: 2 - Spread between best and worst: 2
After round 10: Best advantage: 0.92544987 - Worst disadvantage: 1.0745501 - Spread between best and worst: 0.14910026
After round 1000: Best advantage: 0.99999103 - Worst disadvantage: 1.000009 - Spread between best and worst: 1.7949178e-05