Lorem ipsum dolor sit amet, elit eget consectetuer adipiscing aenean dolor

Bad gem matches

I’ve contacted Gems of War Tech support on numerous occasions about the obvious problems inherent in the design of the game. In order to make the game “more competitive” there appears to be one match in GW each day that favored the gem matches to the opponent. I had the opponent the down to one troop and he had two 5 gem matches materialize AFTER a 4 gem match thus filling his mana and allowing him to do significant damage (Famine). This happens day in and day out. I never seem to get that kind of luck. The DeVos will tell you it’s “observer bias”. NO it most definitely is not. It hasn’t made the game any more enjoyable and has taken something away from GW. Why shouldn’t someone at level 1069 easily destroy all others below them. One last point WHY DO THE KEEP NERFING ALL THE GOOD TROOPS. PLEASE FIX WULFGAROK IT SAYS HE IS IMMUNE TO DEVOUR YET ON THE ios pltform HE ISNT

1 Like

What you are seeing is another AIBoost that I have coined “AISoleSurvivorBoost”. When you get the AI down to one surviving troop, there appears to be a 3x or more chance for falling skulls, match 4/5s and match 3s of needed color for that surviving troop. For example when Dragon Soul is the last troop, it will fill and cast and loop itself, with a few cascading skull hits. I have lost an entire team of 4 to a single AI troop with this boost. (Celestasia matches 4 every turn so has infinite barrier and health, Dragonette drains your troops faster than you can fill them, etc.).

1 Like

It is not observer bias? Care to show some proof of that?

And stop SHOUTING, it doesn’t make your point more valid.


Care to show some proof that it is indeed observer bias? This is quite a double-edged sword. According to Central limit theorem, https://en.wikipedia.org/wiki/Central_limit_theorem , Ok, never mind, that is probably way outside of what is easy to explain. Just keep in mind that any claim of presence or absence of observer bias has to be substantiated. :wink:

Just jumping around and accusing people is not going to help with anything IMO.

It’s definitely observer bias.

And they’ve already said they’re fixing Impervious in an upcoming build.


Ye, here we go, you certainly know. :wink: It is a clear case of observer bias by the way. You are the observer here. :wink:

It’s clearly nonsense, conspiracy theory-level stuff. They definitely aren’t programming this way. It’s absurd to argue that they’re deliberately trying to give the enemy extra skull matches when there are fewer troops left to try to throw matches. It just doesn’t work that way.

Prove it isn’t.


I’ve seen it happen both ways. I’ve lost full teams to a single remaining troop after some lucky cascades for the opponent. However, today in GW, I had a match where the opponent had a single Fizzbang left (with 200+ magic). If there was some sort of “last surviving troop buff” then I should have been in serious trouble. However, the Fizzbang didn’t get any cascades and I won easily.

It is not that easy to prove. I was thinking about a few things that are easier to do. In either case, it might require:

  1. Collection of data. We are talking about considerable number of observations. It is not that easy to approximate but it should be in the range of a few thousands.
  2. Comparison of collected data to simulated data. Simulated data should fit some known random mutidimensional model. According to Central limit theorem, the best model will be regular normal distribution.
  3. If difference between simulated and actual data exceeds certain predetermined significance value, it is possible to make a conclusion that there is some sort of interference. It can be software glitch, performance glitch, error or deliberate manipulation. There is no way to clearly distinguish between various causes of interference. Comparison can be done by multidimensional Kolmogorov-Smirnov test or Anderson-Darling test.

The easiest way to collect data is probably just taking the starting board and simply counting number of various colors present on the board. This can be batch-automated from either screenshots or actual screen picture or just counted manually.

Significance can be preset probably to 0.05 (95% chance that hypothesis is true or false). Higher levels of significance (99%) will require considerably more data.

In either case, it is an epic endeavor only possible by joint effort of a few people, nobody can do it alone. :crazy_face:

However, I have my doubt as to whether it is actually needed. First, developers acknowledged a few times that pRNG (“luck” or whatever you can call it) is manipulated apparently by a simple slider or whatever parameter is there, so we already know that. And the tests won’t tell us what is used to interfere with pRNG. And developers won’t tells us what is manipulated and how specifically. That makes the point a bit moot.

1 Like

He is presenting a theory - game is rigged. So I ask him for a proof. And you go behave like a child saying - show me a proof that it’s not right.

I am playing many hundreds of battles every week, most times I win easily, sometimes I get a bad patch when I lose 2-3 in a row and there is little I can do. So either I am extremely lucky or the game is only rigged against some people (which sounds even more ridiculous than it being rigged against everyone).

Don’t get me wrong, there are many issues with the game right now but without substantial proof I doubt the game is rigged. And if it is then Devs are pretty crappy at their job because most people get over 80% win ratio.

And I hate you because you made me agree with @Delinquent :stuck_out_tongue:


Thanks for your gallant remark. I really appreciate constructive attitude and exquisite manners.

It is a fair sentiment. Now, the questions is why this is happening? And what can be done to avoid it? Is it programmed by the developers? Is it the natural way of how pRNG is supposed to work? Do you like it that way? Do other people have similar experience? How many? The list of questions is essentially endless. Issues like this are a matter of simple curiosity, they obviously might have little impact on personal enjoyment, just a little insubstantial setback to proliferate the collection easily compensated by spending more time or money. So, whether this pace is set deliberately or as an oversight and whether it promotes popularity and increases monetary output for the publishers/developers? Or it is a rational consequence of changing the engine/code and is neither deliberate nor oversight?

Problem is some of our stat freaks and have oodles of spreadsheets to show it’s not observer bias.

Like earlier in the week 25% resurrect Infernal King — #5 times in a GW match…(defense team)

Or today’s Kraken team of mine not being able to devour…and impervious is broken… but defense 3 kraken team chomp chomp chomp all team gone.

Of course things even out, AI Fizz gets 90% explodes and you get 10%. So with a 50% average between 2 players we’re all good right?
AI Justice choosing to create skulls is probably observer bias too… well apparently that’s a bug as sometimes I can’t choose them but I should be able to. Nice.

1 Like

Both of these are unlikely but not outrageously so. 25% resurrect happening five times is 1 in 1000 odds. Given the number of games played each day against Infernal Kings and Dragon Souls, it’s expected that someone gets unlucky and has to fight the same troop six or more times.

1 Like

How many more times? Should there be some limit? Why you think that 6 times is reasonable and 9 times unreasonable? What might be some acceptable reasonable borderline? Do people expect some rounding like 1:100,000 or maybe 1:1,000,000?

Right now GoW does not seem to have any rounding at all. Largest “bonestorms” I’ve noticed had 10 skulls and only skulls without anything else dropping onto the board. It was 5+5 twice in a row in the same move by me, not by AI. I have no way of knowing what the AI drops playing at 4x speed. It has an independent chance of about 3.5E-9. Is this reasonable?

I’ll try to answer myself. Let’s presume there are about 1200 player on Steam on average per hour. Let’s also assume that average match takes 3 minutes giving about 20 matches per hour. Let’s also assume that during the match, the board is cleared something around 3-4 times, so maybe 200 gems spawn. That gives about 4000 spawns per player per hour. Multiplying this by 24 and by 1200 gives about 115 million spawns per day. So, getting a single instance of 3.5E-9 should take about 30 days for all players to play to get single instance of this event.

1 Like

I’m very sorry as a mighty level 1069 you can’t auto magically beat everyone. But grinding levels is not the same as skill…

1 Like

The plural of anecdote is not data.

1 Like

If we are looking at pure RNG, then anything should be possible. I have heard that the developers placed a hard limit of 14 (or 15, or 16) kingdoms that may give tribute at any given collection. I can’t confirm any of that, but if true it does suggest that the devs have a number they consider so rare that it should have probability zero instead. So that is a starting point.

On the happy side of things, I have heard stories about people getting mythic troops from a single gem key – which should be roughly the same chance as IK or Dragon Soul resurrecting five times.

Tangentially, Ive got a level 600 account (in a very low level guild). That account has got 4 mythic troops: A Plague (got at level 80-ish, and 3 Elemaugrim (at around level 500). I` forget how I got the first Elemaugrim, but the second came from 3 gem keys, and the 3rd a month or so later, from another 3 gem keys.

just thought id share that.

I’m not asking to win every single time. What I am asking for is simple - that what the AI has happen happens also with the player. If I ever had an Infernal King resurrect 5 times in one play (or TDS which is my favorite troop) I would be flipping out going OHMG what an amazing thing to watch…WOW…

But you see I see more of that happening with AI than with me the player. I see battles that I was winning clearly by strategy (not by cascades) - and then Poof — AI gets a set of cascades that just changes the game.

I played three battles yesterday against FIZZBANG - who I so loved… I can’t use her, she never ever fires off and destroys gems now. But yesterday against the AI - over and over and over again, where I thought OK maybe they tweaked her again, bring her back, and go play again.

Nope… not a single bang.

I want to loose because it was a good game. I don’t want to loose because of random behavior that is completely and utterly uncontrolled