Randomly Generated Number favours the high numbers?

I have created a function which picks up a unique pool_id on a random basis:

function pick_random_pool($res)
{
    // getting random number between 0 and 1 with 9 decimals
    $rand = mt_rand(0, 1000000000) / 1000000000;

    $pick = 0;

    $pool_id = 0;

    for ($ii = 0; $ii < count($res); $ii++) {
        // if a randomly generated number is bigger keep updating picked id to the last one
        if ($rand > $res[$ii]->probability) {

            $pick = $res[$ii]->probability;
            $pool_id = $res[$ii]->ip_pools_probability_id;

            // if a randomly generated number reaches a number  more than itself, it stops
        } else if ($rand < $res[$ii]->probability) {
            // now we need to figure out which of two candidate numbers is the closest to a randomly generated number
            if (($res[$ii]->probability - $rand) < ($rand - $pick)) {
                $pool_id = $res[$ii]->ip_pools_probability_id;
            }

            break;
            // if by some chance a randomly generated number is exactly the same as a pool's probability
        } else {
            $pool_id = $res[$ii]->ip_pools_probability_id;
            break;

        }
    }

    return $pool_id
}

It looks like it works fine however there is a serious issue. It picks up the biggest number from $res object approx. 90% of the time. For instance, for an array:

Array
(
[0] => stdClass Object
    (
        [ip_pools_probability_id] => 301
        [ip_pools_id] => 12
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.007166877
    )

[1] => stdClass Object
    (
        [ip_pools_probability_id] => 325
        [ip_pools_id] => 13
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.008621843
    )

[2] => stdClass Object
    (
        [ip_pools_probability_id] => 349
        [ip_pools_id] => 14
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.008984716
    )

[3] => stdClass Object
    (
        [ip_pools_probability_id] => 373
        [ip_pools_id] => 15
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.009268012
    )

[4] => stdClass Object
    (
        [ip_pools_probability_id] => 397
        [ip_pools_id] => 16
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.010412571
    )

...

[20] => stdClass Object
    (
        [ip_pools_probability_id] => 61
        [ip_pools_id] => 2
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.094612022
    )

[21] => stdClass Object
    (
        [ip_pools_probability_id] => 85
        [ip_pools_id] => 3
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.097033897
    )

[22] => stdClass Object
    (
        [ip_pools_probability_id] => 133
        [ip_pools_id] => 5
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.098855823
    )

[23] => stdClass Object
    (
        [ip_pools_probability_id] => 109
        [ip_pools_id] => 4
        [time_start] => 12:00:00
        [time_end] => 13:00:00
        [probability] => 0.099785568
    )

)

it will pick up the 23rd element approx. 90% of the time. Why it is not evenly distributed? Where is the flaw in the code?

Thanks in advance.