Goblins are overpowered


#1

OK, OK. This is done to death. But I’m serious.

If I play vs. Divines, the game’s over fast. Never more than a dozen turns. I like that. Even when they get good luck, I lose and I move on.

Goblins… aren’t so kind. When they get on a roll…

WIZZ-BANG
GO CRAZY
WIZZ-BANG
GO CRAZY
WIZZ-BANG
GO CRAZY
4-MATCH
4-MATCH
4-MATCH
4-MATCH
WIZZ-BANG
4-MATCH
4-MATCH
retreat

It’s stupid how long they take to get anything done. They’ll go more than a dozen turns without doing damage at all. It’s even more a pain in the butt that they’re part of Pet Rescue. I’d rather see them nerfed than anything done about Divines.

Isn’t there supposed to be a combo breaker? I’ve had 4 different Goblin Teams make 8 4-matches in a row without cascades today. What gives?


#2

When it comes to goblins, freeze is your best friend. Second is Gobchomper.


#3

Streaky RNG is what gives. That’s a ‘feature’ (I’d say flaw) of the Unity code these days. Theoretically those Goblins are balanced enough. Fizzy and Nobends shouldn’t loop for very long given the % chances of consecutive repeated explodes. But they do - the RNG (anecdotally here, I’ve a maths background and know how silly this might sound academically) seems to ‘click’ into a streak, like a dice deciding to only roll 6s for some time… yesterday I lost to goblins when Fizzy explodes 12 out of 14 times, including eight in a row… I know enough probability and combinatorics maths to know that shouldn’t happen - certainly not as often as it does given the number of times I play.


#4

But the devs will say all that intelligence is “just in your head.” :thinking::wink:
Feels like almost a year I’ve been arguing that the AI dictating luck and not seeing how that is tied to difficulty scale is just simply idiotic.


#5

I’m not saying that the RNG is biased. And it’s nothing to do with the AI. The enemy team decision-making is odd sometimes but not my point here - the drops and cascades and strange percentage chances are a different issue entirely.

The RNG is streaky. There are more cascades and skull drops than there used to be. For sure. It feels a far faster, spikier, more dramatic and swingy game than it was with the old code. I get plenty of luck with cascades and skull drops too. But I hardly notice them as I usually win anyway. It’s when a sudden cascade fills the AI team from nowhere or four skull hits land from above in a row - that’s when I notice.


#6

RNG is dictated by artificial intelligence. Then a separate AI is battling you. It’s the “luck factor” that causes the streaks your seeing. Because it over rides what simple math tells us about odds.
People complain about chests. But it’s the only true RNG in the game. Everything else has a dice roll, but if too favorable/unfavorable of a roll… Then gets the dice bumped to help the player or the AI.


#7

I’m not joining this tired old debate again… the game isn’t rigged…


#8

Not against the player, at least.


#9

Don’t forget nobends dodge skull damage %…what is 20% and he will often dodge 8 or 9 times in a row…


#10

To hell with these goblins we got a bigger feline problem!


#11

Goblins should not have all six colors, if color variety was bumped down to three or four it wouldn’t loop as well. Change purple and blue in all goblins to a different color.


#12

Even the chests have some streaks and repetitions, we tend to notice it less because very few people in very few situations will open chests one by one.


#13

The mention of “streaking” is relevant.

Ha ha, no, stop giggling. I’m a pony. It’s only lewd when we wear clothes.

What I mean is longer ago than I care to dig up, someone did some real math on the RNG and found two interesting things:

  • Over many trials, it does represent a plausibly random distribution with expected probabilities.
  • If you focus on smaller sets, it has more “runs” (which we’re calling “streaks” here) than is expected via probability.
  • In particular, it is biased to “runs” of 2 or 3.

The context was the TDS resurrection trait. The gist is both the devs and the players were right about TDS: over the entire set of games played, TDS resurrected as much as the probability said. However, focusing on individual games, after seeing 1 TDS resurrection, it was more likely it would resurrect again than probability would indicate.

So I don’t find it questionable that other game events are subject to streaks.

What I do find questionable is: I’ve played a few hundred games as goblins. I only very rarely get to go 10 or 15 turns in a row. Fizzbang or Nobend always end up breaking the pattern and at the end of the day I have to do damage to win the match anyway. I never seem to have a “streaky” day.

But yesterday, every pet rescue I had was vs. Goblins. (That might be intentional, I suspect pet days pick from a small, daily-rotating pool rather than “all pets”, and it’s not unlikely the day’s pool can be all vs. Gobllins.) I had to retreat in the middle of Fizzbang/Nobend’s 14th free turn four times.

Ignoring the RNG and randomness: that should never, EVER happen. I don’t even want to play a game where I get 14 free turns. I want like, 3 followed by a big hit.

I want the game to cap the number of free turns you can get. I think somewhere after 5-6 free turns, the board should be shuffled such that there are no 4 or 5 matches, and 7 mana from each of your surviving troops should be drained and converted to bonus gold at the end of the match. I don’t give a flip if this hurts the Goblin meta. Nobody plays Goblins at the high-end except as a GW defense. The only reason it’s a good GW defense is this dumb probability it’ll get so many free turns. It’s not fun to play, it’s not fun to fight, it’s not fun.


#14

I feel the pain of goblins and have felt “rage” almost to rage quit but to start blaming the RNG based on 100, 1000 or even 10000 samples is craziness.

Randomness applies to an infinite set. Sure there are tests for randomness but losing/winning 99/100 times, although annoying as hell still doesn’t implicate the RNG.

Balancing, sure: Team build, sure: but I don’t think the “the games out to screw you”

I’m sure there’s a build to counter. I mean front load fast and kill em. Don’t let them match. I know you can’t do all the time but more proves randomness.


#15

Here’s how the “runs test” I alluded to but didn’t want to write a page about works.

Let’s say I flip a coin 100,000 times. I find that I have about 50,206 Heads. That’s within a pretty acceptable 50% margin and I know the coin is fair.

Now let’s say I flip the coin 4 times 25,000 times, and record the results of each 10. I expect something like:

  • 50% evenly split heads/tails
  • 45% with 3H or 3T
  • 5% with all H or T.

But instead I find:

  • 35% evenly split.
  • 64% with 3T or 3H.
  • 1% with all H or T.

This data set can still add up to “within probability for H or T over 100k tosses”, but looking at the smalller sets reveals a pattern. This coin has a bias towards “streaking”.

With a fair coin, these three betting strategies should both “win” roughly the same:

  • Always bet on the result of the last toss.
  • Alternate H or T.
  • Always bet H.

With a “streaky” coin:

  • Betting on the result of the last toss wins >50% of the time.
  • Alternating H or T is irregular and might win more or lose more.
  • Always betting H is the same.

So what the thread I’m referencing found with respect to TDS is:

  • Over many trials (10k), TDS resurrected the “right” number of times.
  • TDS “liked” streaks of 2 or 3.
  • So if TDS did resurrect, it was more likely to resurrect again the next time, and vice versa, up to 3 times.

So let’s discuss 2 TDS resurrection betting strategies.

“Probability”:

  1. Every event, bet with the stated % that TDS will resurrect.

“Streaks”:

  1. If you have no information, bet TDS will resurrect with the stated %.
  2. If you have a ‘last result’, bet TDS will repeat that result with a higher % so long as the same result has happened less than 3 times.
  3. If you have a ‘streak’, and it breaks, bet on the new result repeating.

The “probability” version would be right the same amount of time vs. a fair RNG or the “streaky” RNG, because in the end “resurrect 3 times” happens as often as “don’t resurrect 3 times” so it balances out. But the “streaks” version would be right more often than the “probability” version, because a busted RNG with an exploitable pattern is in use.

Imagine if I had a coin that guaranteed 50% results over 10,000 throws. That means if I know about 9500 of the throws, I can start winning ‘more than I should’, even though someone who just looks at the 10,000 throws thinks it’s “fair”.

That’s the kind of bias we’re talking about. For that kind of bias, small samples are important because the large samples hide them.


#16

I have no problems with Goblins.
Annoying at times yes, but not hard to beat.
Gob Chomper
Gob Chomper
Apothecary
King Highforge
Don’t want to guess a win rate but can’t remember last time I lost to a goblin.


#17

I think you are completely missing the point. No one in the thread has a hard time defeating Goblins. This thread is about the AI creating/allowing unreasonable turn chains.

(Unless you just stopped by to tell us how good you are at Gems of War, and in that case I say, “Hey cool, right on and you go, girl!”)


#18

That was not my point and I am not missing the point.
I just don’t find that that AI is a problem with Goblins anymore than any other looping team.
I posted my anti goblin team in an attempt to be helpful.
Your last comment seems to have been said just to be condescending?


#19

Most other looping teams do damage as part of their loop. They can’t go 50 turns in a row without killing you, math proves it true.

Goblins, somehow, decide to just self-buff, apply debuffs, or explode gems forever. That doesn’t damage you, so eventually your choices are “retreat” or “come back tomorrow”.


#20

Goblins are fine. If you get that you are not supposed to win every single match you will understand this. My goblins were 0/4 last Guild War so some are not having problems with them.