Equal chance for Mana dropping in battles?

I hardly think so!

I am a statistical analyst and numbers are my thing. Now it’s always been claimed that when Mana drops, there is an equal chance (1/7) for each Mana type, and Skull, to drop.

For the first two months I have played, that was spot on. The numbers vary only slightly but every Mana type were dropping at the approximate rate of 14.3%.

However, in the last week since I have leveled Khaziel to 10 and built a Team that revolves around brown Mana, the numbers have shifted… DRASTICALLY.

Brown Mana now is dropping at a rate of 7.8% when I use this Team, and ONLY this Team. I use a different Team like Shadow-Hunter/Behemoth/Valkyrie/Druid and the numbers will immediately return to normal.

To multiply the effect this is having, the AI is also targetting the brown Mana differently;

  1. Whether it requires it or not
  2. If there are 4 or 5-matches, or Skulls, they AI will ignore them to match brown

The drops are random, but not 1/7 gems and skulls.

It’s 15% chance each colour gem and 10% chance skulls.

Beyond that it’s all RNG unless playing on normal, in which case the ‘combo breaker’ can intervene and stop the AI getting ‘lucky’ drops.

EDIT: no idea what you’re seeing on the brown gems there… Sounds like recall bias… And don’t you have more fun things to do with the game than track EVERY single gem colour that drops… Must take you an hour to do each battle…

Quick question, what Banner are you using?
Second question, if it’s Khaziel, have you tested without the said banner?

I believe what banner you’re using will largely affect what gems the opponent may target in an attempt to thwart you. I can’t speak for the mana dropping, but you might consider testing this theory a bit more before jumping to any rash unseeable conclusions. (Though I’m unsure how exactly the AI works on consoles.)

Best of luck, maybe if it keeps up you can ask 505GamesSupport!

I believe this is true as well, and honestly that’s not an issue. Makes sense that the AI will target what colour(s) you have bonuses for.

Nope, I run a program that “counts” the Gems every time the screen changes. My bias has no input in the process,

Using my program it takes no extra time to play other than starting it up before I start playing.

Can you post a video example of your initial point 2?

Console may be different, but I have never seen the AI select any type of 3-match before a 5 match, or 4 match if available.

So earlier today I pulled my third Gorgotha and built a Team that is 3x Gorgotha and a Behemoth (for that Task that needs 3 of the same Troop) and put the Whitehelm (Yellow x2) Banner on it.

The exact same scenario I explained in my original post is occuring, however this time it is Yellow, not Brown Mana.

I am OPENLY DECLARING that your Mana Bonuses directly affect Mana dropping. IF you happen to use a Bonus that is +2, you WILL see that colour drop about half than normal.

Direct in this meaning;

Bonus +1 : 1 drop rate
Bonus +2 : 0.5 drop rate

Never? So your AI plays the game perfectly?

I believe the AI is programmed NOT to play a perfect game and “make mistakes”.

if i remember correctly after lvl 100 on the pc it stops “making mistakes”

if i remember correctly the AI will always do a match 5 skull, match 5 color (top to bottom and/or right to left), match 4 skull, match 4 color, match 3 skull, spells/other match 3.

I can’t find the post now, but a while back it was determined that consoles have a different AI to Steam/Mobile.

What @kzintiwife says is correct for Steam/mobile, but may not be for consoles.

AI is far from perfect - many times it is better not to select the 4 or 5 match, and opt to cast a spell instead.

For example,

  1. AI goblin team will opt to match 4 or 5 green gems when it already has full mana on the troops using green - it is normally far better to cast a spell, then match on the extra turn.
  2. AI will opt to match 4 or 5 red gems when Sheggra is ready to cast - again, an intelligent player would normally opt not to make this match, as there is more chance of a 4 or 5 skull match when more red gems are on the board.

It may be that it can opt for 3 gem match if you are lower than 100, or if the AI on console is different, but out of interest & verification, please confirm your level and do post your video showing this occurrence.

I do not believe this declaration to be true (it’d be a fundamental deceit in the game if it lowered the drop rate of the colour that your team is clearly designed to use).

Perhaps @Sirrian could comment?

This sounds rather likely. Unless another console users turns up to disagree, I’d say it’s a safe assumption.

In a way this could be used to your advantage. Make your banner a color you rarely use for the team and receive more of the colors you actually need. Sure, you lose some bonus, but is +2 worth 50% less spawn?

I’d imagine the +1/+1 work on 75% each color range, so that could be of better use. Maybe stick to that format for your teams unless something changes. I can say this though, this AI is definitely not the same as the PC/Mobile AI.

Doubtful Sirrian would know about this particular aspect of the game as he was not actively involved in the alternative AI’s creation to my understanding.

Also, I wouldn’t quite call it deceit, more like a secretly enforced balance.
Which sounds sort of like a government… wait a minute… …huh…

I definitely do this. I will line up Browns to force the AI to react while I charge up my other Troops.

I use Gloom Leaf/Gorgotha/Keeper of Souls/Behemoth, so all the AI is doing is stalling me. Once Gorgotha gets charged up, it’s an endless chain of Doomstones/Lay Waste with the occasional Corruption and Soul Strike.

All this has taught me is to be patient and play a game of attrition with the AI. I haven’t started losing because of it, so I’m not complaining. Just simply an observation I happened upon.

I wouldn’t call it that. You get a +2 Bonus, so there has to be a balance.

It’s a fundamental problem of gaming over the last 20 years. Developers think all you have to do to make difficulty harder is by cheating or “stacking the deck” against the player.

So, to put any further conspiracy theories to rest, here is the code that generates our gems:

if (bInitialSetup)
	var nRoll:int = m_pPuzzleBoard.m_GemRNG.percentile();
	m_nType = Math.floor(nRoll / 15);
	m_nType = m_pPuzzleBoard.m_GemRNG.range(0, PuzzleGem.GEM_TYPES - 1);

Where m_GemRNG is standard Mersenne Twister RNG that knows nothing about the board or your masteries or troops or anything.
Gem types 0 through 5 are Blue, Green, Red, Yellow, Purple & Brown, respectively
Gem type 6 is a Skull.

As you can see, during board setup we reduce the chance of a skull to 10%, while other gem types are 15% (so we don’t overly advantage the first player). After that, it’s all even chances.

That’s it! Simplest code ever. Copied straight from our codebase.
No dirty tricks happening at all. Anything else you might imagine is simply recall bias or a small sample size and a run of bad luck (which any RNG, even a relatively good one like the Mersenne Twister, can supply).

The ONLY reasons these values will be changed after generation are:

  1. The combo breaker is on & we want to stop the AI player getting a lucky drop
  2. During board setup, we might change a gem so the board doesn’t start with any matches on it.

No worries Sirri my man, I wasn’t suspecting foul play on your part.

I and TaliaParks are throwing suspicions towards the console releases which clearly operate on a different AI coding. Even if it was originally ripped from your data base straight, alterations seem to have taken place, as the original coding would never opt for 3-of-a-kind of 4+ if it was normal.

Console users have mentioned time and again the AI does take 3’s over 4+ on occasion. Leading me to believe there’s differentiation in coding, and likely in places you wouldn’t suspect.

If, however, you can manually check said console coding yourself and you or 505GamesSupport can quarrel said concerns, then I’ll happily bite my tongue. :yum:

Aha! Thanks for the clarification, @Zelfore!

I don’t have any of the console code here to look at, but I’ll ask the 505/Pipeworks guys and find out what’s up.
They’re a good bunch of guys, and I can’t imagine they would have done anything nefarious on purpose without us knowing about it…
Their port of the game is based on our code, but it is in a different language, so there are bound to be numerous code differences… I’d be a little surprised if they changed the gem generation though… we programmers are notoriously lazy when it comes to stuff like that!

Play XB1 here, and can confirm that the AI will at times forego 4&5xmatches, and take a simple 3match. (sometimes after it has taken a couple 4/5’s already, other times before it has taken any at all) And yes, sometimes for a color it may not even need. But at times, I’ve seen where it doesn’t miss a single one, and have seen the AI take so many darn turns I could probably walk the dogs before I get another one!

But all in all, there still seems to be a lot of randomness involved overall. I never know if the AI will be on it’s game and be ruthless, or if it feels like it’s being wielded by some drunkard who found the controller! :grinning:

I’m on ps4 and yea sometimes it’ll match 3 instead of a 4 or 5… sometimes it’ll take a 4 over a 5, sometimes takes colors need of us need… and yes taking a bajillion turns pretty often I’d say at least 1 in 10 games if not more… also seems to get a lot of luck with any ability that might get an extra turn but might not… gem spammers and exploders and such… celestasia and prismatic orb are annoyin to fight because sometimes it will use the ability 10 or more times in a row for no real reason… sometimes even when it has a loaded guy that could finish me off…