Does the AI know dice results ahead of time?

[Chieth Kieth]

It's the standard complaint in any game with an RNG element: The dice are rigged. Normally I wave them off as confirmation bias, but I'm getting a little suspicious. I don't think the dice are straight-up rigged, but I'm beginning to strongly suspect that the game lets the AI take future dice results into consideration when it makes decisions.

Exhibit A is AR1. The classic human move is to coup Iran so you cut off US access to western Asia. The AI only does it sometimes, and interestingly, it only seems to do it when it wins. I don't think I've ever seen the AI lose the AR1 Iran Coup. Since it happens on AR1 I'm gonna go ahead and play a few games in a row to see how it plays out, but I've got a feeling the AI only coups Iran when it knows ahead of time that the dice will go in its favor.

That, plus the fact that the dice only seem to go low for the AI when a bad roll doesn't hurt it, or when it's at the end of the turn and there's nothing left, makes me think the AI is predicting dice results and using them to decide what it's gonna do. I'll come back in a bit after I run through a few AR1s and see what the AI does.

[Chieth Kieth]

In ten AR1s the AI couped four times and won every time, only one of them with less than 5. That's not exactly statistically significant, but I can't be assed to do this a hundred times and see. I still get the sense that the AI only coups when it knows it'll win.

[Go_Rongor]

I witnessed the AI failing numerous coups and alignments, so I don't see a basis for assumptions like these.

[Twelvefield]

The AI was rolling terribly in my latest game. I felt so bad for it, I suicided myself by mistake. I could have picked random cards and won, the AI was rolling so poorly. I SHOULD have picked random cards, then I wouldn't have been so clever as to suicide myself.

[Darth Warez]

Dice rigged, no doubts. Too many 6 rolls in a row for AI in every game.
Looks like USSR AI working in cheat mode. US AI not doing that ♥♥♥♥.

[Narkulus]

Originally posted by Darth Warez:

Dice rigged, no doubts. Too many 6 rolls in a row for AI in every game.
Looks like USSR AI working in cheat mode. US AI not doing that ♥♥♥♥.

Ludicrous!! I don't believe for a minute that Playdek would rig the dice. What would be the point of that? This kind of post pops up in every game involving dice.

When it comes to dice there can be very long good streaks and bad streaks causing rage and suspicion in a player. Play a hundred games and log every dice roll, because that's about how many games you would have to play to come up with a valid analysis.

[Kane]

The AI is definately fishy in regards to dice rolls.
Besides the omnipresent every-AI-coup-is-a-success....erm..."feature" the AI decided to really mock me a few mins ago.
I played as USSR and the AI plaxed the Vietnam Revolts card and triggered the Event first, only to successfully Coup Vietnam with a rolled 6 immediately thereafter. Yeah....pretty likely...

[Drakken]

Confirmation bias.

[Narkulus]

Originally posted by Kane:

The AI is definately fishy in regards to dice rolls.
Besides the omnipresent every-AI-coup-is-a-success....erm..."feature" the AI decided to really mock me a few mins ago.
I played as USSR and the AI plaxed the Vietnam Revolts card and triggered the Event first, only to successfully Coup Vietnam with a rolled 6 immediately thereafter. Yeah....pretty likely...

Yes, because no man or machine can ever roll several sixes in a row. Inconceiveable!!!

[Flaske]

Confirmation bias, indeed.

I think a good way to prove your point would maybe be to create a spreadsheet of your rolls and his rolls. At least then we'd have something more than anecdotial conjecture

[BockBockChicken]

Originally posted by Flaske:

Confirmation bias, indeed.

I think a good way to prove your point would maybe be to create a spreadsheet of your rolls and his rolls. At least then we'd have something more than anecdotial conjecture

Yeah but if anyone did that they'll find it's not biased. Plus it's too much work for anyone to actually do.

General public, meet statistics. Statistics, meet general public.

[Chieth Kieth]

Originally posted by Flaske:

Confirmation bias, indeed.

I think a good way to prove your point would maybe be to create a spreadsheet of your rolls and his rolls. At least then we'd have something more than anecdotial conjecture

My rolls aren't the point. I don't think the AI is manipulating dice, but I suspect it knows what it'll roll ahead of time and takes that into account when it makes decisions. Bad rolls come up when I make it roll, or when a bad roll is still favorable, or when it's at the end of the turn and it can do nothing else, but otherwise it almost never misses a roll. Not because it can change dice, but because it tries its best to only make rolls when they'll be good. If it's forced to roll anyway, then it fails at a normal rate, but when it has the freedom to roll or not roll, it almost always wins.

Like I said in the OP, it only does the AR1 coup some of the time, but when it does it's always successful. I think it only does the AR1 coup when it knows it'll win.

[Flaske]

I mean, you say it "always wins", but what is "always"?

Let's do the math.


Let's say the Soviets put down a 2-ops card. Worst case.

They have to roll at least 5 to get the US out of Iran, so with a +2 modifier that's a success on a roll of 3, 4, 5 or 6.

It has a 4/6 chance of succeeding.


So to win this particular opening roll, let's say, 10 times games in a row, the ods are (4/6)^10

Or, about 1.73%.

Now let's say 10,000 people are playing TS.

That would mean that for 173 people, the Soviets succeed on this roll 10 games in a row!

For those 173 people, the die seem fixed! And maybe you're one of those guys, who knows.

.... but for the 9826 other people, everything seems normal.

[Darth Warez]

Originally posted by Narkulus:

Yes, because no man or machine can ever roll several sixes in a row. Inconceiveable!!!

USSR AI doing that in every game.

[Darth Warez]

Originally posted by Flaske:

So to win this particular opening roll, let's say, 10 times games in a row, the ods are (4/6)^10

Or, about 1.73%.

Now let's say 10,000 people are playing TS.

That would mean that for 173 people, the Soviets succeed on this roll 10 games in a row!

For those 173 people, the die seem fixed! And maybe you're one of those guys, who knows.

.... but for the 9826 other people, everything seems normal.

Sure that not seems normal for much more players.
1) USSR succeed not only first it always rolling six about 3-4 times in a row in every game.
2) US AI not cheating with rolls its easy to see.

[BockBockChicken]

Originally posted by Darth Warez:

Sure that not seems normal for much more players.
1) USSR succeed not only first it always rolling six about 3-4 times in a row in every game.
2) US AI not cheating with rolls its easy to see.

Step outside and record facts objectively will be the only advice anyone can give you.

It's good advice for life too. That way you'll learn to not overvalue anecdote and put trust in things like statistics and science, instead of being enslaved to confirmation bias.