POLL: Random Gem Spawners

So, I’ve never seen polls actually bring about any kind of change (and if so, not quickly), but I figured it couldn’t hurt to get everyone’s opinion on the recent gem spawner changes.

  • I feel the game is best served with random gem spawners in their current state on PC/Unity, where they randomly fail regardless of how many are on the board or how many you are spawning.
  • I feel the game is best served where only random gem spawners that spawn a high number of gems are even semi-reliable at getting extra turns and only on a favorable board like Mobile/Adobe.
  • I feel the game is best served if all random gem spawners are semi viable so long as you get enough of a color on the board.
  • I feel the game is best served with random gem spawners are not random but clustering toward gems of like color like the first week of PC/Unity, generating reliable extra turns regardless of how many of that color were on the board.

0 voters

With that in mind, how do you feel about the spawn streak breaker code?

  • The spawn streak breaker code is fine the way it is now on PC/Unity.
  • The spawn streak breaker code is too heavy handed on PC/Unity. Mobile/Adobe was better.
  • The spawn streak breaker code was always too heavy handed on Mobile/Adobe. We need an even lighter touch.
  • Why do we even have a spawn streak breaker code? Let random be random. If this causes imbalance with outliers, balance the outliers.

0 voters

Lets discuss here, rather than have everything spread out among multiple threads.


I don’t know if you can change the wording now, but I’m not 100% sure by what you mean about the streak breaker code being “heavy handed”. Regardless of what you do mean, I’m solidly in the “let random be random” camp. Having the game deliberately manipulate processes the way we now know that it does is really upsetting to me.


some games are fine (and actually benefit from it) with pseudo-RNG. Dota comes to mind: it’s well implemented and they can’t have games potentially worth millions of dollars decided on a lucky crit streak or stunlock.

but for a match 3, I just dont see the point. people should play on average enough games that the RNG will balance itself out without needing to manipulate it.

I’ll be trying to get examples of this later. I don’t think Console has this at all. Basically, from what I understand, the spawn streak breaker code will occasionally move otherwise “randomly” created gems away from lining up in a 4 or 5 match. On both PC/Unity and Mobile/Adobe, gems seem to have a strong aversion to spawning adjacent to existing gems, and having a bunch of gems clustered together seems to make it even worse (eg., gems ready to be swiped into the “T” or “L” position will be exceedingly rare to have a gem spawn in the empty spot, even if there are several sections of the board like that). Initial information I dug up was that it was a post from Sirrian stating that it was only active after five consecutive extra turns, but new information indicates it is active from the first extra turn. This is very hard to really prove though given you need a massive sample size to prove anything other than it is already “random is random”. I’d welcome people posting any examples they can in this thread, though.

1 Like

I think I totally understand the way that the streaker breaker code was intended to work, was working when it was backwards (up to yesterday), and is working now. But even with that knowledge, I have a hard time interpreting “lighter touch” and “heavy handed”. Might just be me, but it might also be quite confusing to someone who didn’t read that whole thread and watch your demonstration videos.

While it may never have been operating on console, I can’t help but worry that something like it is at the heart of the skydrop behaviour that constantly plagues us. Knowing that they deliberately introduced code to intervene in random processes really undermines my belief that gem spawning ever acts randomly.

And having it “work backward” was creating a near 100% incidence of extra turns, regardless of number of spawns or how many of that color were on the board to start. Thats what I mean by heavy handed - the overall influence the code had on where gems spawn in, period. A lighter touch would mean it was only active after an actual long streak, and/or only move away gems sometimes when it activates.

1 Like

Also I think troops can still be viable why still failing to get an extra with a loaded board.

There are matters of concern about manipulation of RNG that are continually brought up and not addressed such as devour rates and respawns by the ai. Does the silence speak for its self? There appears to be a handicap modifier of around 1.5x in favor of the ai whether intended or not.

1 Like

I read the poll answers, and they didn’t make a lot of sense to me what it was you were trying to imply or say, so I’ll say it simply- there should be no resulting code which modifies the potential result based upon whether there would be a match or not.

So: Dark Troll (Double the number of Purple Gems on the Board. Then create 4 Purple Gems.) – If there are 9 Purple Gems on the board, the final result should be 18 Purple Gems, +4 (which may overwrite the existing Purple), which means from 18 to 22 Purple Gems should result from the cast of Dark Troll.

Jarl Firemantle (Create 9 Red and 9 Yellow Gems, then deal x damage to an enemy). During the creation stage, Jarl is modifying 18 gems total on the board (including existing Red and Yellow). This is not an X, then Y - this is an X and Y. Jarl will be modifying 18 of 64 gems (28.125%). There should not be code involved which prevents or modifies any matches based upon existing board layout.

Elemaugrim (Deal X damage to all enemies. Create 8 Purple Gems, boosted by Burning enemies.) During the creation stage, this will result in 8 to 12 Purple gems total being created (potentially overwriting existing Purple gems.) There should not be code involved which prevents or modifies matches based upon existing board layout.

Gar’Nok (Deal 1 True Damage to all allies, then create 7 Red and 7 Brown Gems both boosted by Orc Allies.) This will create 16, 18, 20, or 22 Gems, half of which will be Red, half of which will be Brown. This is generated at the same time- as Jarl above, this is not an X then Y, this is an X and Y. I’ve seen casts of Gar’Nok which treats it as THEN, when the text reads as if it should be simultaneous- from 16 to 22 Gems, 25% to 34.375% of the board. There should not be code involved which prevents or modifies matches based upon existing board layout.

– Tranimo


For those, I’m actually pretty satisfied that it is recall bias that triggers complaints. I have tracked reasonable samples for both and always found that things come out relatively close to where they should over a reasonable amount of time. Of course, I still have GW battles where my troops never devour or resurrect and the AI’s always do. Those piss me off more than the ones where everything goes my way make me feel good.

If you’re not convinced, try tracking it yourself for a while. Over 100+ attempts, I doubt you’ll find that your 1.5x modifier is actually there.


I have a few guildmates that have tracked the matches into the thousands and sent support tickets about their findings.

All I know is AI is outta hand, getting all these extra turns, all these extra skulls, cascades of 10 matches already and garnok does not stop looping for me… still. The fact you can NOT even BARELY get an extra turn or a 3 match on a gem spawn is kinda dumb, what is the point if it’s just gonna miss and give all the mana to the AI? Yes, maybe was a little much before, but this is waaaay to much of a decrease. I am struggling to connect any loop with any team, matches are taking longer, It’s almost like whats the point of x4 if the AI is gonna spend the whole time looping and getting skull cascades to drop… I don’t know about anyone else… But this is my experience within the last 24 hours…

1 Like

Oh and lately I’ve been matching 4 times of gem color and I WILL NOT GET MY EXTRA TURN. NOT frozen or anything…


If they’ve done the work and their methods are sound and consistent, then why are they not posting this publicly? I know we joke about the 50% devour being 0% us and 100% AI, but if the deck is really stacked and they have done the work to show it, then bring it forward.


Poll misses the “Make Jarl usefull again, i don’t give a damn about the rest” option, or simply “9+ gems or gtfo!”


Yep. It’s slightly higher than 50% (pre-nerf kerberos) but not as high as 1.5x.

Speaking of recall bias, I remember Lyya once said: “When luck goes the player’s way, s/he sees it as skill, or unremarkable. When luck goes the AI’s way, s/he sees it as cheating.”


I just wanted to say …

Thank you @Mithran . All your hard work, dedication and research are extremely valuable and precisely what the community and developers need at this moment.

So… Thank you!

I am one of those who like the random being random. I honestly believe that the only out-of-hand and unfriendly defense team that creates gems is justice one. So balancing troops specifically is by far the best solution, I believe.
Having looping teams on invade was (and I hope it will be again) fun - and should be at least partly reliable, while still possessing that “Snap!” moment from time to time if the board is not favorable. The game shouldn’t force an algorithm to move gems.

1 Like

I’ve commented on this already, but in the off chance that the devs look this far down the thread:


How much is 'slightly higher, and how do the player vs. AI results compare? Enquiring minds want to know.

55.96%. I have to ask my friend who tracked and made the spreadsheet for the comparison player vs AI. I forgot if he just tracked it for the AI or for both player and AI.