Can we have a grown up discussion about the dice?

[HeyZeus]

To be clear:
I DO NOT suspect a willed and guided hand behind the veil
I DO NOT think anyone is capable of cheating
I DO NOT consider a few battles gone awry proof of anything
I AM NOT griping about losing games to unfair dice (and I never have)
I DO think the rng has a weird "bug" or whatever you want to call it
I KNOW that the dice are equally skewed for everyone (we're not exclusively just attackers or defenders).
Finally I HAVE NEVER trolled (or even posted) in any forum in my life

So...

Around 7-8 months ago i opened the wildly boring app Numbers with the intention of logging my dice rolls. The reason for that strange decision was because I felt something was rotten in Denmark and I wanted to make sure I wasn't "seeing things". Also because the first time I found this message board (after googling AAA online bad dice) I was immediately cut down by a super offensive/combative kanuk (who might or might not have a vested interest in the game) after suggesting not all being right. So, to make sure i wasn't delusional, I logged around 8500 dice like attacking fighters (or cruisers), bombarding battleships, defending fighters, 1 or 2 defending infantry/artillery (or destroyers, you get it, I don't mean to suggest there is differentiation between a fighter or cruiser. So when I say some unit hitting at something, that goes for any other unit hitting at same capacity).

Now, I know people here do not believe me (or anyone daring to suggest something is really off) and there is nothing I can do about that. But seriously, who would spend even a minute in Excel/Numbers if they didn't have to.

Some of the results from my first dice logging effort were like this (unless specified these are only 1st round of dice/combat):

2 defending inf/art/des 0hits: 26.06% 1hit: 55.22% 2hits: 18,71 %% (dice rolled ˙~2750)
1 defending inf/art Misses: 1759 Hits: 1797 (Dice rolles ~3550)
1 defending sub Misses: 293 Hits: 130 (dice rolled ~425)
1 BS bombard Misses: 562 Hits: 567 (dice rolled ~1130)
2 BS bombars 0hits: 272 1hit: 453 2hits: 26 (dice rolled ~750)
Defending fighters (any round of dice) 257/309 hits (83% hits)
Attacking fighters (any round of dice) 695/1998 hits (~35% hits)

In other words, everything's "wrong"

I've mentioned these result a few times on this board and the response has been varied. The defenders think I'm lying and some have experienced the same weirdness. Most people are indifferent I guess.

I duplicated and reset my spreadsheet and went for it at again. Seeing how tedious this is I have so far only done attacking fighters. If this post is not shut down I will update the (and other) results, but so far I have this

Attacking fighters (1st round only):
After 100 dice rolled - 34%hits
After 200 dice rolled ~32% hits
Attacking fighters total (any round):
385 dice rolled 143 hits (37,14 %).

"Exactly" like the first time. You would at some point expect the number to approach 50%, but so far it's the same thing all over again. And I just have to wonder why am I getting these results if everything is hunky dory? Why are attacking fighters hitting at ~35% and not 50%? It annoys me a whole lot. And I know the numbers are real.

Please do not opine on this if you haven't bothered to collect a minimum of data to back up your claims. Please refrain from trying to teach me (or anyone else) about probability (I know enough). Don't accuse me of being a sore loser. And for zeus's sake, please read what is written before making an utterance.

Good weekend everyone!

[bob88lablah]

100% percent. Great Post. These kinda posts can make something happen. Thank You HeyZeus

[vtweatherman]

Nothing is going to happen

[rocky1242]

Well, assuming your numbers are correct. I would say having 2 or less ht approximately half the time after over 3500 rolls is significant enough to say something is wrong. It doesn't make sense how units that hit with 3 or less get hits less than 50% because they hit with a 1 or 2 too and there can't be a negative number of 3s. I've experienced winning and losing >95% chance battles that people usually complain about, but I haven't noticed a higher than normal number of 2 or less units getting hits. I might get curious enough to start tracking them too.

[HeyZeus]

They are correct. What could I possibly stand to gain from cooking dice roll outcomes? What a depressing thought.

https://imgur.com/a/u89BDmO

[HeyZeus]

I've now surpassed 300 dice rolled for attacking fighters (R1)
--------
303 dice rolled: 106hits (34.98%).
-------
And again, the more dice you roll the more that hit rate approaches 35%. Go figure.
Fighters total (any round of dice) is slowly approaching 40% (38.97%)

[dmr]

If you want calculate the probability of certain outcomes of a set of dice rolls in Excel you can use the binomial.dist function:

So the probability of rolling 106 or fewer hits out of 303 fighter attacks is BINOMIAL.DIST(106, 303, 0.5, TRUE)

https://support.microsoft.com/en-us/office/binom-dist-function-c5ae37b6-f39c-4be2-94c2-509a1480770c

[HeyZeus]

----
401 fighter dice rolled: 144hits (35.91%)
----
499 dice rolled: 180 hits (36.07%)
----
600 dice rolled: 218 hits (36.33%)
----
709 dice rolled: 256 hits (36.16%).
---

...pffff

[HeyZeus]

I've switched to tracking some other units because there is glaringly obvious freakishness going on and attacking fighters seem to have stabilized at around 36% hits. Now, the following is a very small sample, and not thrown in as any kind of proof, but how about my first 38 bomber dice rolls:

10 hits. TEN! (I'm only logging R1 dice roll and only battles with a single bomber involved in the attack, so this is 38 battles in other words). That's slightly above 26%

The first 65 lone defending infantry?

45 hits. Yes, a 4 followed by a 5. So 69% hits. Here's the probability of defending inf/art scoring at least 45/65 hits (if my skills using a dice probability calculator are up to snuff):

0.00000000366732

So defending inf/art/dest are outgunning attacking bombers by about x2.7. Doesn't seem right, does it?

As oft happens while writing something here, weirdness occurs (same round of germ combat): first my strat bomber is shot down in Moscow. Then I attack trans jordan with 1 bomber and 2infantry vs 1 infantry. My bomber rolls two straight 5s (my infantry misses of course) and the 1 defending infantry rolls two straight 1s. Illustrates my numbers above pretty much to a T, I'd say. And it's oh so insanely frustrating to witness this happening again and again and again and angina.

Last 2 games yesterday, one right after the other, I had 3 and 4 battles respectively against a lone defending inf/art. All hit R1. That's 7 straight hits for a lone defender. And then you'll say "Uhh but it's not THAT unlikely if you know probability blah-di-blah". But this was not an isolated incident, it happens on a regular basis.. And if you've ever played any kind of dice game before you KNOW that this just does not really happen, whatever slight chance probability gives it. Perhaps once a leap year. Perhaps. I can actually remember the one and only time anything near as improbable happened to me on the board and that was when my 1 japanese infantry holding the fort in Manchuria took out 6 attacking russian infantry before himself succumbing. That was around the turn of the millennium.

I'll keep tracking more units and post them when I've reached a decent sample size. And then I guess that will be it. Because screaming into the void is starting to feel kind of futile.

[rocky1242]

My gut feeling is that there must be some form of unconscious bias that makes you more likely to record the result of a defending infantry when it hits than when it misses, and the opposite for attacking fighters. If the dice engine rolls 1 and 2 more often than the other numbers, that doesn't explain why units that hit with 3 or 4 or less are hitting less than expected. If 1 or 2 is rolled 50% of the time and 3 is rolled 0% of the time, attacking fighters would still hit 50% of the time. The only way your data could be accurate would be if the dice engine acts differently depending on what unit is rolling. Based on how a RNG works, I don't see how that is possible.

[HeyZeus]

Publicado originalmente por rocky1242:

My gut feeling is that there must be some form of unconscious bias that makes you more likely to record the result of a defending infantry when it hits than when it misses, and the opposite for attacking fighters. .

Oh zeus. Seriously????? This is like talking about the fossil record to creationists.
Your explanation for my experience is that bias makes me record only dice rolls that are favorable to my belief? Do you know how insane you sound? And how offensive it is (and I'm hard to offend)?

TRY. IT. YOUR.SELF. And, no, it does not entail from my results that the rng must be able to differentiate between units.

Your gut feeling is worth sh*t if you haven't done the exercise. And do only R1 rolls, because by R3/R4 things are such that you can be led to believe all is OK. I'm pretty sure there's a developer flaw somewhere.
(Btw, what is usually the case by round 3 and 4? The units you are rolling for is a much more homogenous group, is it not?)

I'm a bit curious as to what would happen if the game discarded first attacker result and sent a second request for R1 attack?
But then again I have no idea how the dice rolling is implemented. Does the game make an API call to the rng for each set of dice to be rolled (first the 1s then the 4s) or does it send one request "roll me 10 dice" and then the game itself divides the result into groups of 1s, 2s, 3s etc? If the former then discarding first result would not do much.

Does the game request dice for BOTH attacker and defender for each round of dice and then divvy that up? How does it work? There are several ways this could be fumbled.

If the game sends an API request for each set of units rolled or at least includes instructions for the server to do separate rolls and return them as such (Which is I think the only way it should be done)) then what I'm experiencing makes no sense and I could understand your nonplussedness. But if it's "eyy, gimme 20 dice" and then the game divvies that up into whatever then I can definitely see there could be a problem because there is no doubt that the rng is streaky. The ratio of 2 units rolling face value pairs are close to 40% for what is actually a 1/6 probability.

The majority of the weird results I've recorded are a result of very similar scenarios; ground and air vs 1 or 2 ground units (mostly an overwhelming attacking force).
So for defenders it's mostly dice for the same unit value (hit value or whatyoucallit), homogenous , while for attacker it's at least 2, often 3 and 4 different unit values, so heterogenous.

While the 2-3 paragraphs above will probably be torn apart by the professors in this forum, I have a strong feeling that there is something to it. Again, because I AM RECORDING EXACTLY WHAT I AM GETTING FROM THE GAME, and those results are ludicrous.

I'll keep my own gut feeling to myself btw.

[Lax]

This type of conversation happens on the Baldur's Gate 3 forums more often than not. People thinking the D20 rolls are absurd and looking for confirmation biases. What I think tends to happen is people look at dice roll percentages and see a high chance to hit but disregard the low chance to miss. And then, they are taken aback when the roll misses. Same thing happens on the XCOM - You don't only have a 75% chance to hit something, you have a 25% chance to miss.

I know that sounds obvious, but somehow that gets lost in translation when it comes to thinking about roll outcomes.

[HeyZeus]

Publicado originalmente por Lax:

This type of conversation happens on the Baldur's Gate 3 forums more often than not. People thinking the D20 rolls are absurd and looking for confirmation biases. What I think tends to happen is people look at dice roll percentages and see a high chance to hit but disregard the low chance to miss. And then, they are taken aback when the roll misses. Same thing happens on the XCOM - You don't only have a 75% chance to hit something, you have a 25% chance to miss.

I know that sounds obvious, but somehow that gets lost in translation when it comes to thinking about roll outcomes.

When you jot down every dice roll in a spreadsheet, how is that biased? Like I stated above, I'm aware of both confirmation bias and negativity bias and that is exactly why I started recording the dice rolls. Would be advisable to read what is written.

[rocky1242]

So Zeus what's your explanation for why defending infantry hit 50% and attacking fighters hit 36%? It can't be that 1 and 2 are rolled too much because then the fighters would still hit at least 50% too. You said it doesn't require the RNG to roll differently for different units, so I'm all ears. I said recording bias was my gut feeling, I didn't claim it was happening for a fact. You recorded results for 1000s of dice rolls, are you claiming that you never rolled the dice and didn't record it while obtaining that much data?

I think I will try logging dice rolls, but I want to make sure I'm logging exactly what you think I should. I should only log the first round of defending infantry by themself right?

[montrealsteak]

Maybe it’s more appropriate to keep track of the overall occurrence of each number regardless of hit or miss. For example, there will be a certain number of wasted 4s, 3s, and 2s when infantry are attacking at 1. All those wasted 4s, 3s and 2s may ultimately bring down the hits for bombers, battleships, fighters, cruisers and tanks. This assumes that in the entire game of rolls, there is a statistical limit to the total of each numeral occurring.

[rocky1242]

Well I set up some 1 fighter vs 1 infantry battles and recorded the first round of combat. Out of 244 battles the fighter hit 118 times and missed 126, the infantry hit 87 times and missed 157. It doesn't prove the dice roll exactly as expected, but I feel confident that the attacking fighter hits more often than a defending infantry. I don't see defending infantry hitting 50% of the time after 3500 rolls as mathematically possible, there has to be some form of error.