Biased AI cascades, and matchmaking

blacklotus

Omega Red wrote:

I don't think the AI gets more or better cascades than me but I always found interesting how after a big board shake new stones seem to fall always in sets of the same colour. Like, you use Koth's first ability and then four blue stones fall straight, a prefabricated match basically; then three green, then three yellow, etc. This always irks me as it seems coded deliberately that way to make the game faster and more arcade-like.

Think Puzzle Fighters, the Street Fighter inspired match3 arcade game.

Omega Red is one of my favourite characters to use in the fighting game X-men:Children of the Atom. His strong tentacles strikes can fake the AI easily.

Hibernum_JC

I know I haven't been on here in a while (we're super busy with stuff!) but I can reiterate, once more -

The AI absolutely does NOT know what gems are going to fall, and does not even take into account the layout of the board when making a match. There are a few rules as to how it picks it's matches, but none of that ever analyses the board, knows what's coming, knows that making a match will make a cascade, nothing.

Confirmation bias is a huge thing.

As for matchmaking, it's a question I am asking our engineers right now, so I'll get back to you guys on that.

yunnnn

Hibernum_JC wrote:

The AI absolutely does NOT know what gems are going to fall, and does not even take into account the layout of the board when making a match. There are a few rules as to how it picks it's matches, but none of that ever analyses the board, knows what's coming, knows that making a match will make a cascade, nothing.

Confirmation bias is a huge thing.

I love how this thread has necessitated a mod/dev to comment, even though it can all be explained using basic logic:

Basic Statistics
Forum members who complain about rigged AI cascades are simply lacking a basic knowledge in statistics. This game only has 5 colors + loyalty orbs (ignoring the rare void orb). There are not a large amount of unique colors. So, if 2 orbs line up, a third orb falling into place has a ~17% chance of matching. That is a pretty high chance! (Often times, I'll make low mana matches that can possibly trigger cascades using the same logic; it's good strategy!)

Ever so often the AI is going to hit the 17%. And even still, it could hit a 17% on top of that 17%, which occurs ~3% of the time. Think about how many games you play in QB. The AI is eventually going to do some nasty stuff even if it plays poorly!

Confirmation Bias
Consider QB win rates. Mine is 95%+, so let's consider what happens when a huge cascade happens:

1. I get a huge cascade. I was gonna win anyway. Who cares. I forget about it.

2. The AI gets a huge cascade. The AI had a very small chance of winning (< 5%). Now that it got the cascade, it has suddenly say... 30% chance of winning. Because the AI went from "auto-lose" to "maybe-win", this registers in the brain as a significant event.

Events 1 and 2 happen just as often. It's just that when (2) happens, it's much more significant, and leads your brain into bias. Further recurrences of these events drive confirmation bias and super long thread topics.

TL:DR; You lost a single **** match (out of 80 games) and lost top rank because of an AI's YOLO cascade. This is simple statistics, not rigging.

Match Making:
Yea, this is totally broken.

Corn_Noodles

Hibernum_JC wrote:

The AI absolutely does NOT know what gems are going to fall, and does not even take into account the layout of the board when making a match. There are a few rules as to how it picks it's matches, but none of that ever analyses the board, knows what's coming, knows that making a match will make a cascade, nothing.

Great, so the AI doesn't know. But you didn't address the possibility if cascades happen more often in certain circumstances like a rubber-band mechanic.

glggwp

cascade has been noticeably more common since zendikar. many mythics are costing around 20 mana. guess they need it

EDHdad

Great, so the AI doesn't know. But you didn't address the possibility if cascades happen more often in certain circumstances like a rubber-band mechanic.

Statistically, cascades are more likely to happen when more than 3 gems on the board are destroyed and replaced with random gems.

In other words, a cascade is statistically more likely to follow a cascade. This is why effects like Sword of the Animist and Chandra's first ability can be so strong.

If you want to quantify how much the AI cascades vs you, keep track of how much loyalty you generate throughout a game vs the AI. The last time I measured this, I was getting and average of more than twice as much loyalty as the AI over the course of a game. I have seen zero evidence that the AI has any idea how to even make a cascade with the gems which are already on the field.

Corn_Noodles

EDHdad wrote:

The last time I measured this, I was getting and average of more than twice as much loyalty as the AI over the course of a game.

In my experience, loyalty gems are at the bottom of what the AI will try to match.

Foznertep

I don't think I have ever seen the AI make a deliberate loyalty match, actually.

EDHdad

Right. But I also get more loyalty from colored gems than the AI gets from colored gems. The AI doesn't see cascades which are blatantly sitting on the field. Example, if it could match 3 gems and cause a cascade, or match 4 gems, it will match 4 gems. If it can make a match which screws up your Evolving Wilds or Animist's Awakening, it will make that match.

Hibernum_JC

Corn Noodles wrote:

Hibernum_JC wrote:

The AI absolutely does NOT know what gems are going to fall, and does not even take into account the layout of the board when making a match. There are a few rules as to how it picks it's matches, but none of that ever analyses the board, knows what's coming, knows that making a match will make a cascade, nothing.

Great, so the AI doesn't know. But you didn't address the possibility if cascades happen more often in certain circumstances like a rubber-band mechanic.

There is no such thing. We do not use rubberbanding or anything of the sort. If cascades happen more often in certain circumstances, it's pure RNG. We do not, under any circumstance, trick the RNG so that any side in a match gets more cascades than the other.

Like I said, all of this is confirmation bias - when something detrimental to you happens, you tend to notice it a lot more than when it is beneficial.

Corn_Noodles

Hibernum_JC wrote:

There is no such thing. We do not use rubberbanding or anything of the sort. If cascades happen more often in certain circumstances, it's pure RNG. We do not, under any circumstance, trick the RNG so that any side in a match gets more cascades than the other.

Thank you.