I have said that for a long time now and it just went ignored.
That’s an awesome idea. It would end the conspiracy theories and add much needed strategy to the game.
1 Like
Let me get this straight. Are you suggesting there’s code in the client software like dropGems(extraTurnChance) that sets the value higher for the AI than for the player?
Of course the “possibility” of what you suggest exists, it’s all code; and the same facility responsible for AI logic (GemsOfWar.exe) is responsible for the RNG. The potential for collusion between these two components is ever present and has been since the game was created; it is also the case in any other player-vs-CPU game ever made.
But you’re suggesting the devs are deliberately coding paths that allow the AI to win when it “wants” to? Are we talking some sort of Skynet out-of-control experiment in machine learning here?
My take is that you’re seeing the increased pendulum swings due to the changing nature of the game mechanics (AI doesn’t always take Skull bait, and 4-matches can now Mana Surge), seeing the AI cast more often than it used to, and interpreting that as a malicious crusade instead of side effects of a changing rule set.
10 Likes
What I’m suggesting is the devs can and likely do tweak that very probability as a way of making the AI “smarter.” It’s an obvious and easy way to adjust how often the AI wins and how “challenging” the battles are. I use “quotes” because I don’t actually think it reflects “smarts” or “challenge” at all, and is simply turning a dial that adjusts the coin-flip nature of the thing. I’m also not talking about a skynet AI that “wants” to do anything - don’t that that too literally. I’m just talking about algorithmic decisions and probabilities.
There are two ways the devs can adjust the win/loss rate of the game.
- The “Visible” decisions - They can adjust how “smart” the AI is with regard to making matches of gems and skulls on the board and when it casts spells (or not)
- The “Invisible” decisions - They can adjust how the initial board lays out, and more importantly, they decide what gems fall out of the sky. That can dictate the course of the battle.
I see no reason why they’d limit themselves to only adjusting #1, and not also using #2 as a way to change how the battle proceeds and its likely outcome.
They’ve gone on record multiple times as saying they explicitly do not do what you suggest they do, aside from some combo breaking scenarios where they adjust a cascade chance the other way. I suppose it boils down to a matter of trust. You’ll not get the source code, so in the absence of trust, all you can do is gather objective data and try to prove your case empirically (not anecdotally).
I like the new AI.
I’ve previously gone on record to say that the matches were too easy. Through skull-baiting (and having either a throw-away troop or one with a defensive trait up front), you could easily win before the enemy would’ve charged or fired a single troop.
Finally the opponent is also using its spells regularly. Finally we need to start taking that into account.
Which is not to say that the game is fine as it is - it needs less RNG.
If you wish to remove the conspiratorial part of the argument, then it’s fine - perhaps the program hands out such things equally to both sides of the battle. I certainly can’t prove otherwise, as you say.
However, as I was also saying, regardless of the “bias” argument, the game is marching onwards towards more of a “win goes to the first side charged” dynamic, which would be closer to a coin flip and less and less reliant on strategy, since the strategy is basically nothing more than “charge up first or die”.
Not every battle is like that, but most of mine now are. My teams are optimized so that if I can charge up, I can almost always loop the team over and over and kill the Defense fast. And now, more and more, i’m going against meta teams that are also optimized, so that if they charge up first, I get wiped out quickly.
Yes, there are some battles that are more extended, hand control back and forth, and require more strategy and planning. But more and more it seems it all just boils down to who charges up quicker (even if I still think the AI biases itself, even if I can’t prove it!)
3 Likes
This part I agree with wholeheartedly. The game used to be slow and methodical, and with every update they add mechanics that speed things up to the point that now, the first lucky cascade often dictates the outcome of the match. The big change in this patch is that a major tool at the player’s disposal – a predictable, baitable AI – is gone. We feel the extent of the pendulum swings more acutely now because they swing against us more often.
3 Likes
That score is not ideal. But, it will improve with more experience, this is only the first week after all. Talk to other players in your guild, you probably will need to alter your troop composition. I’m in Bracket 1 on console so yes I do understand.
Lyya, I was thinking about some way to prove/disprove certain hypotheses.
This would require rather complex approaches as developers probably don’t really know how exactly the engine works and rely on results of testing as much as the players do in their speculations.
Testing won’t be easy technically. For example, starting board with regard to gem distribution can be checked for normality with various statistical tests. These tests can compare set of let’s say 10000 random boards generated using known equations for normal distribution to boards actually generated in the game. If difference is statistically significant, that would mean that either pRNG is defective or that the starting board is being manipulated.
Automating this tasks won’t be easy and will probably require some screen capture linked to tabulating the board while at the same time generating random boards based on known normal distribution.
Comparing the sets of data would require some algorithm that cannot just simply be borrowed from somewhere and has to be developed. It is quite complex combinatorial math, outside of average advanced calculus college course. But I’m sure it can be done. One would also need to consider rules for board generation that are used by the engine (no matches of their own).
And I cannot even think of how to approach other gem spawning problems from statistical point of view. The combinatorics becomes really complex in this case.
So, no, there is no absolute need to trust the developers, unless you feel more comfortable doing so, but rather do the tests and get some results or stop unsubstantiated claims one way or another.
What developers actually claimed is that pRNG is manipulated and whether it is:
- Still present in current version (?)
- In favor of the player or not, does not matter, manipulation is manipulation especially if it is hidden/undisclosed
- Combo breakers are obviously there, as developers confirmed
- The more complex the code for board generation becomes, the harder it is for anyone to fish out errors and oddities in the code.
So, yes, we can trust the developers but since developers acknowledged themselves that the board is manipulated in some unknown manner, the issue becomes completely moot and should be ideally investigated. That is if somebody takes on this epic task. I’m up to it and might be able to help but I lack skills to create some automatic data recognition tools.
I’m going to keep this short as always.
I’ve been playing against the Unity AI for A LOT longer than 4 days. There is no AI vs player bias in gem spawning/drops. They drop like a fire hydrant that’s been turned on, for both sides.
4 Likes
Your earlier “game decides” sounded like there was intentionality, like a stacked deck. In reality, it is simply random, and sometimes you get the wrong side of random, so the deck is well shuffled and both sides have an equal chance for getting the better side of it. However, whoever does get the better side gets a significant and non-trivial advantage. Your “charge up first or die” comment is bang on, and while not weighted more for the AI vs the player, it’s not very strategic, and your decisions simply don’t impact the outcome that much anymore.
And if you want to “prove” things are truly random, all you need to do is look at the source code, and you can see how it really works. 
2 Likes
Im right there with santa leading this charge. Someone on these forums gets all but hurt when i bring up justice league. Because on console people figured out its the troop combo that manipulates the current ai the best. He says im a winer…yet here is another post (among the hundred) that sees it.
Winter is coming.
Beanie, I’m afraid it is not that simple. You can look and source code and depending on its complexity, still understand nothing. At certain level of complexity, things starting to get really murky. Does anybody in the world really understands how Windows works? No, not really.
For example, if you play some complex computer games, like games from Bioware, developers frequently fully acknowledged that they don’t understand how their final product functions. They do playtesting, that is all that can be done. While fans and community testers frequently discover “hidden” features and bugs. Sometimes, Bioware even fixed a few things.
I also remember considerable number of “hidden” features and exploits in Civilization games which were discovered only after extended periods of time by the community.
GoW is deceptively simple by the looks of it, yet at this stage, I’m sure the code is becoming more and more complex and hard to analyze to the letter. Generally, board games and 8x8 and larger board games are very complex and have sophisticated combinatorial math behind the scenes.
Chess uses same size board, has fewer pieces, has no random or other statistical factor yet it is one of the most complex games. Same can be said about go. While chess AI is exceptionally good these days and beats a grandmaster easily even running on the phone, it was a long way of getting there.
Do you want GoW to turn into some random Gorgotha casting and cascading ad nauseum again and again? And claiming this is all random? Well, somebody might like that, I’m not sure. Do you want it to be like chess? Somebody might like that as well. IMHO, it should strike some fine balance between the two. When balance shifts and dramatically, then things get out of hand.
If primary goal of developers is making money, then this might not have much impact. It will surely change part of player base. But whatever player base arrives might be more generous with their monetary contribution, so developers might consider cathering to that new player base. IDK how it turns out. But that surely leaves the current player base free to complain and ask questions. These questions are largely remaining unanswered by qualified developers. And I personally don’t like it.
I think your claim that both sides have an equal chance is just that - a claim. As you’ve not seen the source, nor have I, you are simply making an unsupported claim, and I admit that I am as well. I simply feel that the developers have likely tweaked things so the AI gets more advantages than the player in terms of convenient gem drops and such. I get that you don’t feel that way, but we’ll just have to agree to disagree. Until there’s some sort of actual proof or legitimate testing done to actually prove it either way, it’s all just speculation.
Given the admitted manipulation that they do publicly talk about, it’s not very hard for me to imagine some behind-the-scenes tweaking as well. Something as simple as “give the AI side a 10% higher chance of having a 4+ drop from the sky” is something that would be subtle, easy to do, and very difficult to prove, requiring hundreds or thousands of data points to show a statistically valid bias. Assuming they have an interest in adjusting the game’s level of challenge, I can’t think of a good reason why they’d not experiment with that sort of thing.
But again, the larger point is simply that all this manipulation, regardless of bias, just makes the game feel very contrived to me when these battles more and more get ended in just a few massive cascades that appear extremely unlikely to occur (and in the past DID occur far less frequently, essentially proving that Unity messes with it - at least it convinces me), yet happen amazingly often, and just make the whole thing feel like a coin flip, as already discussed.
I guess at the end of the day, regardless of whether it’s due to bias, or rules changes that effect everyone, etc… point is, it’s feeling less fun and more frustrating all the time, and that’s the real shame of it all, at least for me.
2 Likes
That’s what I find ridiculous. I don’t need free chain after chain drops. I’ve set my team to win by playing it and not by watching gem chains happening.
Although from the last three days the opponent had far more unbelievable chains than I got. Before it was more… um, balanced and didn’t happen so often.
But after all, the player wants to win and enjoy the game. The AI doesn’t feel frustrated if it loses.
I wonder how newbie and common player unfriendly this new experience is, when you just sit there and watch.
I would actually enjoy playing a mode in which there were no cascades at all, and 4+ matches simply got you another gem. The whole dynamic of chaining together “lucky” matches just drives the game further towards RNG-driven matches the more often the devs manipulate the engine into producing those matches.
Please do not compare GoW with chess…GoW is lightyears away from chess, since it bases on rules and strategy and not on RNG!
I would rather compare it with Backgammon.
There I can handle lucky dice of the opponent or my side and I can can deal with dumb moves on my or opponents side and I do have a liitle bit of strategy in here. I have a little influence on the progress, but everytime the factor of rolling the dice and of being lucky or not.
I know the comparison is bit lame, but you all know what I mean.
Can’t it be that easy as Backgammon?
Very good points, but GoW simply isn’t that complex. Also, it would be hard to analyze something which was a small variation on correct behaviour, such as the combo-breaker bug. But the type of thing being described, basically if AI then generate different colour would need to be at a different and more visible level of the code. Something like that would be hard to hide even if you did want to spend the time to do it, let alone by accident.
You’re only half right about neither of us seeing the source code, and I stand by my previous statements. There are some slight differences in the behaviour between player and AI, but it’s generally not the type of manipulations you are imagining, and they are almost always in favour of the player. Quite simply, there is no AI bias or favour, and the conspiracy theories about them are incorrect.
Totally agree with this. I’d like my choices to matter a bit more again.
3 Likes
Well, it’s nice for you if you’ve seen the source… assuming you’re seeing the latest source code, and the version that’s actually in production, etc, etc. You can rest more solidly on your beliefs. As I am not in that position, I have to retain my skeptical attitude. Naturally, as I said, I only have anecdotes and feelings about it, but I guess it is what it is. I can’t strongly assert or prove there’s a bias, I can only say it feels that way to me. If I am ever in the position to examine the production source code myself, then perhaps I could convince myself otherwise!
But again, the point isn’t really about bias towards the AI - the point is more about the battle being balanced on a knife-edge. Before this, it felt like one side or the other had to keep pushing pretty hard, for a few turns before the thing finally tipped over in their favor, like trying to push over a weight on a solid tripod. You could drive to a win, but it took several turns, good strategy and gem work, and you basically had to “work” a bit to get it to tip. Also, if you weren’t careful, it could easily tip back the other way, and wobble like that for a while before finally falling over one way or the other.
Now, it’s like tipping over a pencil standing on its point. If you’re able to nudge it the right way, it falls over fast and hard and you win in a devastating series of massive cascades, often all in a single turn. On the other hand, if things tip slightly the other way, suddenly there’s another blur of gems, and half or more of your team is just… gone (seriously, at 4x speed it’s sometimes just ridiculous how your team can simply melt away in seconds and you’re not even sure how it happened). I’m not really sure where “skill” factors into those kinds of dynamics, generally speaking.
Obviously, this doesn’t happen every single battle, it’s just pretty common. More common that I’d have thought possible via straight RNG. On top of that, there’s hardly any “recoverability” at least at my level. Usually one side or the other gets pounded so hard, and loses so many troops so fast, that coming back from that for an ultimate win, while possible, is very rare, at least for me. The game used to feel more like a boxing match to me - hit, get hit, hit again, recover, combo, defend, etc… but now it’s really more of a “one shot, one kill” pistol duel experience for me now; it’s all just about who gets the first shot.
3 Likes