My viewpoint is probably a strange one, because I am not a pure PvP or PvE player.
1. I like difficult games but my first PvP online game was chess. In chess, the AI is definitely stronger than any human because the concept of advantage can be expressed mathematically, which allows AI to learn from the past and analyze with precision. Google's latest AI is so advanced now it can self-learn without seeing any human games. In chess, there doesn't exist a PvP and PvAI distinction. I think a game like Civ could definitely benefit from machine learning.
2. My second hobby as a gamer was not a digital duelist, but a modder/designer. My dream has always been simple, make (or remake) a complex, turn-based PvP videogame where PvP and PvE would have no distinction, like chess. Obviously, this means the AI and player should play under the same ruleset. Obviously, in games of reaction times and micromanagement (like Starcraft or Dota2 or Street Fighter/Super Smash Brothers), this is clearly unfair, but I always liked turn based games more due to my first game being chess. I wish to go one further than chess, and try to minimize as much as possible the impact of opening theory (meaning a veteran's advantage is only in the experience afforded by game sense rather than repetition or memorization of specific openings or endings). This means game mechanics should be intuitive, strong play should be obvious even to new players, and the best move should be objectively so and not contingent on whether or not the opponent is a player or an AI or a noob or a veteran or whether or not the move is "safe" or "risky".
3. Of course, my last sentence is seemingly impossible. So long as a game has interest, people and AI will do analysis and summarize their experiences. This means there will always be a "meta" or an "opening theory". Again, chess provides many good answers. One is a "deep" game, or one where any attempt to create general principles to understand the game will always yield counterexamples where playing "by the book" leads to losses. This means even if the game rules are not patched, the meta will be in flux for a very long time. Second is the idea of "checkmate", or that (all) the win condition(s) should be discrete and discontinuous. This makes machine analysis hard and forces players to constantly pay attention because even if the game contains numbers or easily quantifiable or continuous elements, these things are only correlated to victory but do not directly cause it (i.e. the enemy can comeback if not completely squashed). Furthermore, it encourages asymmetrical play rather than the boring match-up of strength vs. strength and allows for players to make sacrificial play that can essentially give up continuous and quantifiable (often long-term) advantages for the immediate win. Third is the idea of "draw" and "stalemate", where both players can agree to a tie if it's obvious neither can force victory and the attacker can blunder away a victory and be forced to settle for a draw if he is careless. This is helpful because it creates more space for game mechanics and incentivizes good defense (because you no longer need to ensure parity with the attacker, you just simply need to make his life sufficiently difficult). Of course, as children of the computer era, we cannot hope such a game will forever resist the analysis of keen minds and computers. This is where I think DotA2 under IceFrog has the right approach, though of course, the profit motive of Valve restrains his hand. Namely, game depth must be maintained in an environment of iconoclastic patching. What do I mean by iconoclastic? I mean that no game mechanic is sacred and that whatever is currently most dominant in the meta is what will be nerfed the hardest in the next patch. Players should not grow too personally attached to any strategy or skill emphasized by the current patch because it will be gone in X months. Nor should players constantly expect any given faction/hero/playable character or playstyle/item/unit to be permanently enshrined or retain its current role. Veterancy must be constantly re-proven and re-earned. The game should actively punish the kind of "specialist gamers/one-patch wonders" who refuse to learn or adapt beyond a comfortable subset of the skills needed to be top-tier, preferably by never creating such imbalanced, over-centralizing things to begin with.
4. IMO, the whole point of PvP, besides the competitive drive and stress relief, is to practice mastery. Mastery is the ability to rationally analyze the cost-benefits of a situation, come up with the best choice, and then execute it with discipline, under all conditions. By definition, that is risk-neutral, not risk-averse. This is why I find the objection to RNG to be ridiculous.
Of course, many existing PvP games are not of this mold, which probably explains the opposition I get from many PvP players. They mostly play "systematic" games where building up small advantages is the only way to play. Civ and Advanced Wars certainly are of this type, as is Pokemon VG, and most TCGs. In those games, the winner/loser of the game is known before the final victory, usually at the halfway point of the game, or sometimes at the start. Every opening is usually some form of delayed gratification that leads to some power spike and it's usually strength vs. strength or strength vs. direct counter, hence the creation of tier lists. Stratego or 军棋 is more fun that that because at the least, there's more mind-games involved, and the best unit in the game actually dies to the worst unit in the game. Maybe it's just me, but even PvP games are supposed to be an escape, and there's no fun when the same optimal strategy in real life is also the best (or only) strategy in video games, especially when creating the optimal fighter required grinding IRL that was very esoteric or expensive (in time or money) for PvE players. Quite often, my modding suggestions make these players feel like I'm taking their fun away, even though many times, their idea of "competitive fun" was based on the banning of in-game mechanics designed by the creators to check their "setup and sweep" style. (To be fair, I agree that some of the attempts by creators to check the dominant style were very poorly implemented or did not work as intended.) That's why I made this thread, for those of you who enjoy playing such games, what do you get out of it?
