Distribution · statistics

Game::Stats

Divergent

Descriptive statistics and discrete probability as five small classes — an append-only sample, its mean and variance, covariance and correlation, and a Bayes helper.

Version: 0.2.11 cpan:HOLYGHOST
Depends: nothing outside the core
License: GPL-3.0
Its own test suite: 2 files, green
Checked: 2026-09-15 against Raku++ 3.28.0 and Rakudo 2026.08
Where it lives: raku.land · source

Install it #

$ rakupp install Game::Stats

zef install Game::Stats writes the same store; either installer leaves the module usable by both engines.

What it is for #

The first statistics any simulation needs: collect numbers, then ask for their total, their mean and their spread; given two such collections, ask whether they move together. The distribution adds a discrete-probability class on top, with the conjunction, disjunction, conditional and Bayes operations over an indexed probability vector.

Despite the name there is nothing game-specific in it — no dice, no random number generator, no sampling. It is a small statistics library.

Samples and their moments #

File

use Game::Stats::Population;
use Game::Stats::DistributionPopulation;

my $p = Game::Stats::Population.new;
$p.add($_) for 3, 1, 4, 1, 5;
say $p.population.join(',');
say $p.nth(0), ' ', $p.nth(4);

my $d = Game::Stats::DistributionPopulation.new;
$d.add($_) for 0.1, 0.2, 0.3, 0.4;
say 'sum       ', $d.GeneratedNumber;
say 'mean      ', $d.Expectance;
say 'variance  ', $d.Variance;
say 'cumulative to index 2: ', $d.Cumulative(2);

Output

3,1,4,1,5
3 5
sum       1
mean      0.25
variance  0.016667
cumulative to index 2: 0.3

Population is the append-only container and DistributionPopulation the subclass that computes over it. GeneratedNumber is the sum, Expectance the mean, and Cumulative a running total up to an index — which is what you index into when turning a uniform draw into a categorical one.

The probability class works over a vector of probabilities by index:

File

use Game::Stats::Probablity;

my $P = Game::Stats::Probability.new(xpop => [0.2, 0.3, 0.5]);
say (^3).map({ $P.P($_) }).join(' ');
say 'and      ', $P.Pand(0, 0.5);
say 'or       ', $P.Por(0, 1, 0.5);
say 'cond     ', $P.CondP(0, 0.5);
say 'bayes    ', $P.Bayes([0, 1, 2], [0.9, 0.5, 0.1], 0);

Output

0.2 0.3 0.5
and      0.1
or       0.4
cond     0.5
bayes    0.473684

The one thing to know #

The distribution is named Game::Stats and there is no Game::Stats unit to load, so use Game::Stats; fails. Each of the five units must be loaded by its own name — and the probability one is misspelled in the file name while the class inside it is not:

File

say (try { EVAL 'use Game::Stats; 1' }) // 'use Game::Stats: not a unit';
say (try { EVAL 'use Game::Stats::Probablity; 1' }) // 'as shipped: no';
say (try { EVAL 'use Game::Stats::Probability; 1' }) // 'spelled right: no';
say (try { EVAL 'use Game::Stats::Probablity; Game::Stats::Probability.^name' })
    // 'the class: no';

Output

use Game::Stats: not a unit
1
spelled right: no
Game::Stats::Probability

So you write use Game::Stats::Probablity; — no second i — and then refer to Game::Stats::Probability with it. The two spellings are never the same one at the same time.

Where the two engines differ #

Nothing here differs between the engines: both produce the same numbers, including the wrong ones. The badge on this page is amber for a different reason, which is that some of those numbers are wrong on both.

Covariance flattens its two populations into a single list and then walks that list in pairs, so for x = 1,2,3 and y = 4,5,6 it pairs (1,2), (3,4) and (5,6) instead of (1,4), (2,5) and (3,6). The result is −0.667 where the textbook answer is +0.667 — wrong sign and wrong value. Correlation is built on it and inherits the fault: correlating a variable with itself, which is 1 by definition, gives 0.333.

Variance divides by n−1 while Expectance divides by n, so the two conventions are mixed even before that. And a division by zero yields a rational with a zero denominator rather than an error, which then throws when you print it.

The sample container, the mean, the cumulative totals and the probability class are all sound. Treat the covariance and correlation pair as unusable and compute those two yourself.