A.I. Improvement

Currently the AI will choose colors that none of its troop use (I don’t mean 4-matches) even when it has Spells ready to fire.

I have no problem fixing things like that, but gameplay has to remain casual.

A slider system would not be the end of the world, but if it’s just one increment on the way to a challenging AI, please say no to all the increments.

As a former game designer I always thought that a good game would have a win rate of 75%/25% at normal level for a normal person (3/4).
Playing at easy level would have a win rate of up to 90%/10% (9/10).
Playing at harder level would have a win rate of down to 50%/50% (1/2).
This means that the game is always interesting because you always have the chance to lose games and, if you’re skilled, you can can keep a positive percentage of victories even at the most difficult level.
But, of course, I can understand those who want to satisfy themselves with a 100% victory rate playing in “brainless mode” in front of television. Someone who’s playing to collect cards, not for the challenge imposed by the game, for example.
It’s just not me, I’m the competitive type of person who love and want and search the challenge.
So, I would like to be able to completely customize the defensive strategy that must be used by the AI. But, of course, the defense slider is better than nothing.
Although I consider completely abominable the fact that AI prefers skUlls (or gems matching) to skIlls.

Truthfully, GoW has a fairly low skill cap. Upping the AI to make better choices isn’t going to change that. There are a limited number of things you can do at any given point, and fairly easy to see what potential effect they will have to the current board. Making one wrong move can cost you a match, even with the current AI, as can using a gem spawner and lining up a bunch of match 4/5s for them to take, or getting a bad cascade. It is just fairly easy in general to not make such a move, and then you can also skull stall your way out of it occasionally.

There are thee situations where you need to use skulls to stall in general:

  1. You are trying to dig your way out of a hole due to a bad cascade or just made a bad move in general, or they brought empowered board mod and you didn’t and usage of said board mod would give them the mana to run over your entire team
  2. You have a terrible starting board and can’t get the mana to get started before something like Bone Dragon or Khorvash or Famine or Justice fires and puts you in a hole you have to dig out of. As everyone knows, the stronger teams in GoW are mostly predicated around where a single early cast can cost you the match.
  3. You are purposely using a team that isn’t overpowered or needs longer to get going, because you like some variety occasionally

Better AI that makes it so you can’t stall would make comebacks even less likely due to a single early random cascades or bad move. Yes, this makes it “harder”, but only slightly, and occasionally just randomly. Secondly, it accentuates bad boards, making some of them “unsolvable” versus certain teams. Thirdly, it increases the necessity for your team to be “fast” to beat the others to the punch, reducing your viable options instead of increasing them.

This is simply the wrong way you want to increase difficulty. Meaningful difficulty would provoke more thought by giving you more moves or more varied moves where the outcome is predictable but more things need to be considered fully. Counters to consider before the match, different ways to mod the board inside the match. Increase the skill cap, not narrow the field of play.

The only way GoW will ever become more difficult in any meaningful way is introducing powerful board mod spells with difficult to use patterns and then having some kind of challenge where mastery of said board mod must be used to be victorious. The mechanics of the game would have to be completely and totally changed from the ground up, otherwise (no extra turns, instakills, devour, possibly even the way primary gem swaps occur).

1 Like

I didn’t fully read your post @Mithran, but skull-baiting needs to be a less viable strategy. You are right in that it is a valid comeback strategy, but right now I often don’t even really register what’s on the opposing side. I just bait him with skulls while I fill my team, then start rolling.
That’s no fun.
Either the AI needs to recognize that certain frontline troops make skulls less viable (like a Gorgotha up front) and thus deprioritize skulls vs skills/mana, or something that makes skull-baiting less viable.

As a case in point, one of the most fun games I had last week was where the opponent had a Kerberos and I hadn’t registered it until I left a five match in purple for him and ofcourse he devoured. I managed to scrape a win, with the endgame board being my Dragonmoth (over from Dragonmoth/Krys/TDS/Sylv) vs a Warg (over from his Kerberos) which ofcourse summoned a Dire Wolf on death. He used a Dust Devil to put my Dragonmoth last as first move. Coincidentally it also was the game last week where I came closest to losing.

1 Like

I must wholeheartedly disagree with this point. It’s mostly true for simple board-control mechanics (swapping gems) and even gem transformers - but even then there’s a chance for a board where you swap gems in one corner to make a long cascade ending with 4+match in another that’s very difficult (or rather tedious and time-consuming) to spot by a human, but fairly easy for an AI (a well-programmed one of course). When you add gem destroying/removal (all gems of a colour and such - to get a deterministic result) you get a board that’s mostly rubbish to human eye, but can be easily analysed by AI. AI that is really good at board-control would make this game rather unfun bordering on unplayable.
Using non-board-control (or non-deterministic) spells is a different matter of course, much more difficult to program efficiently, but there are still many troops that would be available for such improved algorithm.

1 Like

I don’t disagree with any of this, but the larger point is that even this would not make the game more difficult in the right way. Two “perfectly” played games are always going to be determined in large part by RNG, with bias to the player that moves first. While it is an extreme case that I don’t think we will be seeing here, having an AI that plays “perfectly” makes at least some matches straight up unwinnable because of this. As you stated, it would be extremely unfun.

However, in the end, better AI accentuates your disadvantage state, and you begin in an advantage state. In most cases, using a meta team or a counter team to what you are fighting, you can simply ride this out all the way to the end, without needing to skull bait. If you do find yourself in disadvantage, using skull bait to dig yourself out of a hole is one way you can make a comeback. Some teams only work because skull baiting is a thing. A huge majority, actually, if you happen to be going up against any “meta” defense.

My larger points was “smarter” AI in the way that is on consoles does not increase your options like some people think it will. It will not increase variety on defense, though the primary defenses used may be shifted (which will of course invite endless complaints and cries for nerf). It makes it less likely for a player with lower stats to be able to combat a higher one, even if they play “perfectly” - especially toxic to a beginner’s foray into PvP. It reduces the number of viable team options. It doesn’t provoke any additional thought to your moves in most cases - boards are still the same, turns still pass the same way, there are still the same board mod options.

It will still be a game of turn management, a race to get your damage in, and you still have a head start. And thats how it should be. Its a puzzle to be solved. Its only fun as long as the puzzle is solvable. Having the puzzle be basically free no matter what choices you make isn’t exactly fun, but that isn’t where we are at now in terms of relative difficulty. Even in the most basic of PvP games versus any team with similar stats that is even semi-functional together, you still have to swap the right gems or cast the right spells in the right location to win, its just usually really obvious what that move is. It is very easy to take a loss from just a single wrong move, even with todays AI, especially for lower level players that can be one-shot with many many spells, but it is also very easy to see the correct move. Some things require puzzles to be solved in a certain way (Wraith, Khorvash, manticore, great maw, the deathknight class, bone dragon/courage, famine, even queen mab etc.) and even then its nothing but complaint after complaint when they keep showing up over and over because you have to use a very small set of troops in a very specific way to solve the puzzle.

tl;dr The game is easy because the game is easy. Making AI better will not increase the level of “skill” required to win, but it may increase the number of unwinnable boards with good teams (slightly) and make a number of sort-of decent teams completely unviable because it becomes impossible to win the mana race without skull baiting.


The console AI makes this a less viable strategy. This is both because players will actually reduce their defence team’s tendency to take skulls (by moving the slider) and because the human player will be uncertain about whether their opponent has moved the slider. Whether the slider has been moved or not, you can’t be certain about what the AI is going to do at any given time. Even at level 1020, it still makes some sub-optimal moves, such as taking a 4-match as a 3-match, or ignoring a colour it needs to take one that it doesn’t. This tendency to occasionally make a ‘wrong’ move seems to be unrelated to the customization options.

The sliders also seem to have some effect on spell-casting, such as choosing which colour a Guardian will transform, and it seems to be better at handling mana transformers like Valkyrie or Giant Spider than the old version. Even with transformers, it will sometimes still make a less than optimal choice.

Overall, I think it makes for more fun than the too-predictable PC/Mobile AI.