I cant respond because it would probably depend on a few other factors.
If the deck contains creatures/supports, the chances of being steamrolled on turn 1 is relatively low, I assume. Of course the AI receiving perfectly elegant cascades is a variable.
As previously mentioned, is a significant AI cascade necessary to fuel the AI turn 1 kill?
Does the combo starting 1-2 card’s mana cost equal more than 1-2 matches worth of mana? (On color or off color?)
Additionally, the AI seems to casts cards based on casting cost. Therefore, the cast sequence in order to achieve that turn 1 win would have to directly correlated to the deck’s cards casting cost. In other words, the cast sequence would have to match an exact descending casting cost sequence. There are a few anomallies Ive seen which is stored cards/rarity possibly skewing the casting order at times by the AI. Ultimately, the turn 1 kill would have to occur with as little hand casting order/exile manipulation by the AI. Is this the case?
Furthermore, the AI would have to be capable of selecting the right choices if it prompts for selections, i.e. digging for cards. Therefore another variable that could hinder the AI side turn 1 kill.
How many combo pieces are required for the AI turn 1 kill?
Can the deck achieve secondary objectives with relative ease? Can it achieve a broad spectrum of objectives like cycling was capable of? Can the deck with a slight modication for objectives achievement still retain an AI turn 1 kill capability?
What is the consistency at which it achieves a turn 1 kill? What is the expected consistency for achieving a turn 1 kill by the AI?
Does the deck have troll characteristics? Similar to Naru-plicate? (Now what constitutes a troll loop deck is sometimes kind of grey.. like the kiora gem conversion, path of discovery, certain sunbird decks. While others like naru-plicate feel a bit more troll-y if its specifically built to enable that combo and if the player purposely leaves that deck for others to fall victim to. It seems kind of weird but if a player used a troll-ish deck and then changed it out so others dont suffer, i would feel a little more inclined to leniency - though it doesnt mean much, as not everyone is inclined to manners.. but i digress).
I dont doubt that you have some terrific decks, including turn 1 kill decks, as you are a great deckbuilder. But, i believe, its a little hard to make a good selection on this topic due to the variables.
Edit: re-read the initial post.. turn 3-4 kill not turn 1 kill**