After pouring through reams of data I finally understand how the Sift works. Both perturbation methods and quantum loop gravity failed to reproduce the complexity, so in desperation I resorted to empirical methods. I was able to reproduce the Sifting action to eight decimal places (Friday afternoons notwithstanding). I give you the formula:

Odds of being sifted = [(Catchiness of title) * (Quality of video) * (Number of requeues +1)^0.5 - (Number of people who don't "get it") - (Ant Irritation factor) + (Did dystopianfuturetoday fall off his chair laughing and knock himself unconscious?)] / 10

I think you'll agree that it's a thing of beauty. I welcome any and all tests of its validity.