[cowtools]

With a normal bot, you can only run one boss-fight at a time and you are subject to network latency and imprecision of your actions.

In this blockchain scenario, couldn't you run millions of different simulations of the boss fight on your own computer and publish the one in which you get the most loot, take the least damage, produce the most compact proof, or whatever?

In any case, the "proof-of-victory" as I'll call it is a useful technique for a provably-fair gambling system. For example, you could have players gamble over the outcome of some turn-based game (with provably-fair randomness) like poker, or nethack. If you can implement a time limit, it would also work for something like chess.

[lmm]

> couldn't you run millions of different simulations of the boss fight on your own computer and publish the one in which you get the most loot, take the least damage, produce the most compact proof, or whatever?

Sure, if your computer is up to running the boss fight that many times over. I assume the game would only let you run each "challenge" once (and I guess might charge you a tiny fee to enter the boss fight) and most random inputs would lose; in that case they just have to make it more expensive to do that than to make currency by mining.

[cowtools]

> Sure, if your computer is up to running the boss fight that many times over.

The boss fights have to be very CPU and bandwidth-efficient because every other computer on the network will have to verify many of them. One exception could be that if the probability of winning is low enough with most seeds, you could just make it so that you don't need to broadcast your game (you only broadcast your "proof-of-victory" claim the reward, so if you get a losing seed you just have to broadcast a transaction that pays for your "entry ticket"). That way you could make it difficult to simulate many games while keeping the network performant.

Actually, maybe that's not right because you might be able to make a program that checks to see if you got a lucky seed, so it will just effectively add an annoying roulette-style gambling element of getting good RNG. I don't know.

Anyways, I think a somewhat-fun single-player game may work if it's something like this:

(1) You pay a fee to play a provably-fair deterministic dungeon-crawler type game like nethack.

(2) When you die, you get an NFT for every item you earn. There are massive network fees that prevent you from easily trading the NFT on an open market. So it's preferable to get them through the singleplayer game.

(3) The NFT is used in some hearthstone-like multiplayer card game that encourages you to collect a variety of cards. The "rarer" NFTs are not explicitly better than the common NFTs, they just allow for more exotic gameplay situations so they will simply be more desirable because they are more fun. This encourages you to play the single-player game many times to collect a wide variety of items.

(4) Optional: implement competitive (tournament pool) gambling on top of the card game.

Making a bot to do the singleplayer game would be difficult/undesirable here because:

- you would have to program it to understand the user's different tradeoffs between going to different parts of the dungeons and getting different types of loot, which is sort of subjective decision making.

- trading rewards between players is infeasible due to fees, and there's no trustworthy way to "buy an account" off of a botter.

- the reward is mostly cosmetic and has sentimental value: you can just fork the card game to play it without the NFTs, even to gamble.

- the singleplayer game is somewhat fun in itself, and has a skill cap to prevent bots from stomping actual players (Sort of like TF2's MvM)

I think a crypto game like this could work if it was open source, 100% of fees go to miners, and the development is funded-by-donation. But many crypto games are pretty scammy so they don't pan out.