Probability

[Digger3000]

Let's say I have something like this: $colors=array("red", "blue"); And I want to randomly echo one of the things in the array. But let's say I want it to be more likely for it to be red than blue. I guess I could put red into the array twice, but is there a quicker way? I'm very, very lazy.

[CFMaBiSmAd]

The only easy way, without actually writing out the probability equations using PHP code, is to create an array with the correct number of entries with the "weight" you want -

PHP Code:

$colors=array("red", "blue"); // gives one out of two chance for red
$colors=array("red", "red", "blue"); // gives two out of three chance for red
$colors=array("red", "red", "red", "blue"); // gives three out of four chance for red 


You could shorten the data structure to a color/weight pair, but you would need to add code to produce the expanded array with all the entries. To get a computer to randomly pick from the weighted possibilities, you need a "pool" of the available choices to pick from.

You can use the shuffle(...) or array_rand(...) function instead of a random number to actually generate a random entry.

[Digger3000]

I made up a solution that seems like it could work. What do you think?

PHP Code:

$colors=array("2|red", "1|blue");
$newcolors=array();
foreach($colors as $key => $value) {
$split=explode("|", $value);
$probability=$split[0];
$thecolor=$split[1];
$i=1;
while($i <= $probability) {
$newcolors[]=$thecolor;
$i++;
}
} 


[CFMaBiSmAd]

Try it and see if it produces the array you expect. That's where writting, testing, and debugging code gets fun, getting code to do what you want.

[Digger3000]

I did print_r($newcolors); and got this: "Array ( [0] => Red [1] => Red [2] => Blue )" So it seems to work.

[ralph l mayo]

Just another method you can use if it suits you... it simulates an array with duplicates determining priority instead of actually making one, but I don't know whether it's faster or not.

PHP Code:

function array_weighted_rand($ary)
{
    # Pick a random index as if the array was arranged so that it actually
    # had more entries of the values with higher probabilities
    # eg. for values = [ red, blue, green]
    #     and probabilities = [ 2, 1, 3]
    # assume something like:
    # 0 => red
    # 1 => red
    # 2 => blue
    # 3 => green
    # 4 => green
    # 5 => green
    $prob_idx = mt_rand(0, array_sum(array_values($ary)) - 1);
    
    # Find where the selected index maps to the real indices by calculating
    # a running sum on the probabilities and returning whenever the index
    # is exceeded.
    $running_sum = 0;
    foreach ($ary as $value=>$prob)
    {
        $running_sum += $prob;
        if ($running_sum > $prob_idx)
            return $value;
    }
}

$colors = array(
    'red'   => 2, 
    'blue'  => 1, 
    'green' => 3
);
$picks = array(
    'red'   => 0,
    'blue'  => 0,
    'green' => 0
);

for ($idx = 0; $idx < 100000; ++$idx)
{
    ++$picks[array_weighted_rand($colors)];
}
print_r($picks);