There's a difference between using probability to claim you win and the ability to actually demonstrate a win. As people have already stated, shortcuts work when you can demonstrate that you can yield a neutral or positive result when it comes to the generation of mana/tokens/damage, etc. Stacking the deck is not one of those things - statistics are only going so far. I will attempt to show some examples.
Ex 1. Mill decks can't mill out a deck with an original Eldrazi titan (
Kozilek, Butcher of Truth &
Ulamog, the Infinite Gyre) in it, unless they can manage to do one of the following:
1) Cause the player being milled to draw more cards than they have in their deck, in response to one of the titan's reshuffle trigger.
2) By exiling the titan in response to it hitting the graveyard and the trigger being placed on the stack. (or all the cards with something like
Leyline of the Void)
3) Somehow stealing the Titan(s) from their deck to prevent the anti-mill.
4) The player being milled draws the card(s) in question or you force them to draw into it before continuing the mill.
This is due to the interaction of how the Eldrazi Titans prevent mill strategies. They're not unbeatable, but without a silver bullet answer like the ones mentioned above. Infinite mill potential on a player's turn doesn't really do anything. You could say that "statistically" there could be a scenario where the eldrazi titan(s) are among the last few cards in an opponent's deck, but that doesn't allow you, as the one forcing the mill, to shortcut to that potential outcome.
Ex. 2 (competitive decklist) On my turn, I have
Grand Abolisher,
Captain Sisay, enough mana rocks to generate 3-5 mana and plus a card in hand to cast. I use Sisay to tutor for
Paradox Engine and cast and resolve it since no one can counter anything I do, short of something like
Kheru Spellsnatcher or
Voidmage Apprentice. My turn is my own and no one can interact with me or what I do. I can and likely will then pull out every legendary non-land permanent and play them. The cheap ones net me mana in this example and allow for the ones over the 3-5 mana cost that the rocks/dorks are producing between each spell tutored/cast (nevermind any open lands). Something in the list is bound to give everything haste, such as
Akroma's Memorial.
Selvala, Heart of the Wild and
Kamahl, Fist of Krosa and any of the other pieces I'll draw into will enable infinite mana, infinite pump.
In this example, there is a repeatable set of actions that I can follow, that will net me mana or remain mana neutral, flood the board with legendary creatures/non-lands, where no opponent can react, and that will result in lethal damage regardless of how many creatures are on the board. I've done it in my play group enough during competitive games that if those pieces are there, I tell them what I'm going to do, and they acknowledge the win and we swap decks and shuffle up.
Statistics do not justify shortcutting outcomes because of the potential that exists for it to occur. It's the reason why the new
Rakdos, the Showstopper +
Krark's Thumb doesn't allow you to shortcut to 50-75% of your opponent's creatures being destroyed. An increased chance to get the outcome you want doesn't justify anything.