Dice vs. certain players...

[Polar Bear Cavalry]

I'm normally the last guy to suspect BS with the dice, but certain circumstances with this game (specifically in ranked play) have compelled me to track the results of rolls in search of trends. I'm obviously not able to track defensive rolls, since I'm typically not on to see them, so I'm just tracking stuff I do on the attack. I went in with a couple of floating theories:

1.) Strategic bombers get shot down more frequently than regular aircraft.
2.) My rolls on attack are generally fine, but the rolls on defense are absurd.

I haven't noticed these trends in custom games with my friends, or even in games vs. the AI. But I have noticed them in ranked play, and more specifically against certain opponents that I keep getting matched against. And it is getting really frustrating.

I tried to take an extended sample. The numbers are below.

MY ATTACK ROLLS
1 (1951 rolls, 16.41% hits)
2 (or less) (2093 rolls, 35.41% hits)
3 (or less) (2857 rolls, 47.02% hits)
4 (or less) (271 rolls, 62.95% hits)

Nothing really out of the ordinary here. Doing well on 2's, getting screwed a bit on 3's and 4's, but whatever. But check out their defense.

THEIR DEFENSE ROLLS
1 (252 rolls, 24.00% hits)
2 (or less) (1917 rolls, 34.03% hits)
3 (or less) (606 rolls, 72.50% hits)
4 (or less) (513 rolls, 70.59% hits)

This is ludicrous. My opponents are hitting on 3's 53% more frequently than I am. And this isn't a small sample size. WTF? Seriously.

What's worse: though I can't track the rolls, I almost never rout on defense, so I'm guessing the trends aren't very favourable there, either.

AA defense is something I tracked separately. The results are:

STRATEGIC BOMBING DEFENSE
140 attempts, 50 AA hits (35.71% hits... which is more than double what they should be expected to hit at)

GENERIC AA GUN DEFENSE IN BATTLE
89 rolls, 18 AA hits (20.22%, which is still high, but much more reasonable and admittedly a small enough sample size to not do this justice)

Further to all of this: though I haven't been tracking it, the vast majority of the "bad" rolls almost always occur against the same players in ranked play. I have taken to launching games hours (sometimes days) after playing against certain people, hoping that I won't draw them again when a match finishes. I'd love to be able to just block them, but I'm assuming that will never be a feature, and I'd rather get to the bottom of this anyway.

So seriously... what gives? I don't want to be *that guy* ♥♥♥♥♥♥♥♥ about dice (get in line, I suppose), but tracking this took some serious effort, and I'm generally miffed at what I'm seeing.

I'd love to see some "overall" results posted by the devs, maybe tracking how well infantry do on defense, or how well AA guns hit. Not sure if that's possible, but I'd love to know how I stack up to game-wide trends. Because from where I'm sitting, it sure feels like I've been marked to get screwed over.

How some of these guys have 90% win rates in ranked play, I'll never know.

[WebbsNYC]

Pepper...you, like I, write a lot. I've read a great many of your posts, and while I don't necessarily disagree with you, I think you take an overly Socratic approach to your responses. Sometimes.

The bottom line is, that yes, I can invent all sorts of scenarios where the "statistics" spell out one thing, and yet, my point above is that, for anyone playing this game, there is a given "optimal" set of moves for any given situation. From turn one, to two, to three. No matter what scenario you invent, there is a statistical probabilistic "best" move sequence. Now, even if you make that best move sequence at each individual decision point, you're still going to lose 20-30% of the time. Why? Dice.

No two ways about it. Perfect...and I mean PERFECT play, will still result in 2-3 losses per 10 games. So, I don't care about optimal retreats. I don't care about first moves or best moves...even doing the absolute statistically "best" moves, you will still lose.

I guess to some extent, that is what describes a "good" game (as it relates to a dice game). Lots of skill, lots of luck. Chess is more, no luck, lots of skill. I guess I am just one of those guys that hates watching a well planned and well thought out set of moves completely disintegrate because my 8 tanks miss for three rolls in a row while the 3 infantry I'm fighting can't miss. It's part of the game. It is what it is. I, for one, am learning I don't really like that dynamic. It's not so much planning or strategy as much as what you roll.

[aardvarkpepper]

Publicado originalmente por WebbsNYC:

I guess I am just one of those guys that hates watching a well planned and well thought out set of moves completely disintegrate because my 8 tanks miss for three rolls in a row while the 3 infantry I'm fighting can't miss. It's part of the game. It is what it is. I, for one, am learning I don't really like that dynamic. It's not so much planning or strategy as much as what you roll.

Axis and Allies is not a game about strategy and tactics. It's a game about risk management, packaged as a game about strategy and tactics. In a long letter I wrote to the developers a while go, I wrote something like "What we don't want is for the players to play five hours and feel it was just a bunch of glorified dice rolling" - and I remember editing out the end of that sentence, which originally read "*even though that's exactly what Axis and Allies is*".

Well, suffice it to say I feel the same way most of the time, and that I consider Axis and Allies a sort of light therapy for me to remember I can't always control things - and that's okay.

Yes, as I stare at the dice that came up all fives and sixes, I tell myself it's okay, as I rock back and forth as I stare at the dice that came up all fives and sixes as I rock back and . . .

==

Now, even if you make that best move sequence at each individual decision point, you're still going to lose 20-30% of the time.

When the ranked games feature came out, I spammed a few games to get a rank - just to test the feature. Two of my Axis games were slow to finish, so I spammed two additional games to get my rank faster.

I didn't take any of those games seriously, but in ranked I'm currently platinum 6-1 with Axis, and gold 5-0 with Allies.

I consider myself the low side of middling ability in Axis and Allies, in terms of experience, intuition, correct calculation, position evaluation, and ability to read and respond to opponents.

Imagine if I were actually skilled.

Then too consider a lot of games I imagine are won by forfeit because of timeout - either because an opponent just didn't log in on time that day, or because over a few weeks they ended up having to travel or attend to work commitments so couldn't make their move.

So if you had a player that could consistently get on their computer, and that had some reasonable skill, that took the games seriously - you know?

[WebbsNYC]

The rank isn't really the important thing. It's just a measure. The important thing is that you enjoy playing, and I am discovering that I do not. I used to play the board game all the time as a teenager in the 90's, but I gotta say, the dice here do indeed seem a bit off. The developers say they are fine...so - *shrug* - I guess. But then again, it's not like people have never lied before for one reason or another. I just don't see the motive here.

They need to look at the strategic bombing runs. Something is wrong. No one should be getting hit 35% of the time over 140 trials (see above). With only 6 potential outcomes, 140 trials is enough that there should be basically no way for him to have that experience. It would be easy for me to dismiss that as internet nonsense, but I see it myself...my bombers are shot down basically non-stop. And again, no joke, when they do make it through, my average roll is absolutely no higher than a 2.5. Again, this is across 50 games, and perhaps hundreds of individual bombing runs. I'm getting shot down 30% of the time. Easily. And then, again, when I do make it through, it's almost certainly a 1. Not kidding, 40% chance of a 1. The remaining numbers are roughly evenly distributed. Except 6's. Those come up every 20-25 rolls or so. The bombing dice are broken. Maybe some interns ♥♥♥♥♥♥ with them while working over the summer. Maybe Beamdog has some sick fetish of watching people squirm, but I am pretty sure these are not random outcomes. If they are, then I am experiencing something like 6 standard deviations from normal.

Anyway, they need a low luck option. This game, generally speaking, needs a low luck option. And even without it, it will take some real convincing to prove that the AA dice aren't a mess. Everything else, I'll let everyone else argue about, but the AA dice are not random.

Like the original poster, I have also noticed unusual defensive rolls. If one or two troops are defending against a larger force, those two will almost ALWAYS hit. 5 tanks against two infantry...maybe 2-3 tank hits for the attacker, but the two defenders will ALWAYS hit. It's become a running joke to me..."and now the 5 defenders will inevitably hit at least 5 of my units during the 3 rounds of combat this will take because my 15 attackers simply will miss everything on round 1." And lo and behold, that's exactly what happens. I'm basically expecting to get screwed, and it's starting to make the game not fun.

Developers, if you're reading this...you're going to start losing players without a Low Luck option or something to moderate the extreme outcomes in certain key battles. Perhaps that is "just the game" but I might as well go flip coins and try to call it heads or tails. What was fun as a teenager is rapidly losing it's appeal now that I've realized I'm basically just rolling for luck and hoping for the best against anyone in a roughly similar skill bracket.

[WebbsNYC]

PS...if the developers would like to set up 10 games of me vs the highest rated Allies player and the highest rated Axis player, then set up 10 games of me vs the exact median player for each, my guess is that we'll all be quite close...namely because of the dice. Sure, the higher rated guys should do better, and the lower rated guys should do worse, but the spread between us all is going to be very narrow. I propose I will easily beat the best players 3-4x out of 10. And I propose I will be beaten by the median players 2-3x out of 10. All due to dice. The better players are still better, and the worse players still worse. But given the dice, basically anyone is beatable, and your winning a certain game means virtually nothing as that same player could easily trounce you the very next game with just a light breath from the fates.

I guess this is all a long way of saying if one of you suckers happens to randomly take me down by rolling an unusual series of 1's and 2's, you can go suck a lemon...and you'll eventually find me somewhere in the Platinum leaderboard!

On that note...Developers...we need a challenge mechanism (or rematch request) to initiate ranked games against those that want a piece of what I'm puttin down. Gauntlet thrown.

[WebbsNYC]

Duuuuuddddeeee! Strategic bombing is broken. Literally just sent two. One was shot down. The other...yup...rolled a 1. This is all day every day.

[aardvarkpepper]

Publicado originalmente por "Alright, hit it!":

I
STRATEGIC BOMBING DEFENSE
140 attempts, 50 AA hits (35.71% hits... which is more than double what they should be expected to hit at).

https://www.statisticshowto.datasciencecentral.com/calculators/binomial-distribution-calculator/

0.000004223% 50 or more hits

https://www.mathsisfun.com/data/confidence-interval-calculator.html

could be useful

https://www.youtube.com/watch?v=zMRrNY0pxfM

people say "more than double" whatever, the image that comes up is "oh, so instead of rolling a 1, you rolled a 2, what's the big difference"

need to think of things in proper binomial distribution calculator terms

0.000004223% = maybe there really is something weird

note - I think with binomial events if you're calculating confidence interval, use 0s and 1s to signify binary outcomes; you don't need the specific 1-6 dice roll outputs. or something. I forget. going to sleep. night.

[Polar Bear Cavalry]

Publicado originalmente por WebbsNYC:

Duuuuuddddeeee! Strategic bombing is broken. Literally just sent two. One was shot down. The other...yup...rolled a 1. This is all day every day.

I haven't been tracking it, but sending multiple bombers often seems to result in at least one getting shot down. I half joked with friends that I'm stopping strategic bombing altogether, because it just isn't worth it. You'll almost always lose more IPC's than you'll inflict.

Another theory I've thought of: what if the dice are rolled out the same way casinos award rig slot machines to pay out X percent of what they take in? And more to the point: what if certain dice only "pay out" after a bunch of other crap dice have been awarded? The motive behind such a system would be to ensure that a relatively even number of hits vs. non-hits are awarded, so on the surface the "law of averages" would basically be served and everything looks fine... but such systems are able to be compromised, and this is part of the reason why I mentioned that I've noticed the dice almost always go certain ways against certain players. At what point does it stop being "luck" and start being "trend", honestly?

There are other oddities as well. If I throw 6+ dice, the number of times I've rolled nothing but 5's and 6's dwarfs the number of times I've rolled 1's and 2's. It's astounding. Realistically, both should be extremely rare. And as much as I don't want to be *that guy* who complains about the dice, something is seriously up here. I've played variants of this game for close to 30 years, and I've never seen such consistent wackiness... in particular with the strategic AA (normal AA for regular battles is far less effective).

[Polar Bear Cavalry]

Publicado originalmente por WebbsNYC:

#1 Axis player and #1 Allies player...I don't care how good or smart you are or how many games you've played. If you're not losing 20-30% of your games on just ♥♥♥♥♥♥♥♥ dice, you're cheating, or should be spending your money on lottery tickets instead of video games.

And most of the time I look at the ladder, it's the same person on both sides. From a mathematical perspective alone, that is genuinely astounding.

[nicewolf64]

Publicado originalmente por cody.ouimet:

The dice are all rolled one at a time, not in batches with a single output. There is no systemic difference in our system between rolling a single dice in a roll or rolling 100 dice.

i just started a new game and am going to track every time i roll 5+ dice at the same time. the fact that each dice is rolled individually makes no difference. whether as a batch or individual, they all use the same RNG algorithm. I am assuming that time is used in your algorithm since its the only thing that changes constantly. how many dice can be rolled in 1/1000 of a second these days? or a 1/10000 of a second? does each roll occur as the same amount of time after the last roll?

[Sheol]

The dice are jacked up. I don't have supporting stats but if 4 Infantry attack 6 tanks and win the battle ... that is statistically impossible ... yet, it happened.

Many battles reports I read have me with total losses of relatively strong defenses and zero to 1 loss for the attacking force no matter how large. I want to see stat reports telling me if I rolled average, rare-strong, or rare-weak rates.

Some anomalies are expected. But they happen far too often for chance luck to explain them away.

I too have observed insane loses in strategic bombings. 35% loss feels right.

[WebbsNYC]

Here's a great example of getting screwed by dice and basically losing a game in Turn 1. I'm not saying this is statistically impossible, but wow...it sure does seem to happen frequently. This is upon opening a game today where my opponent took his Japan turn 1 overnight...

Sea Zone 47...Japan won. Destroyed 1 aircraft carrier, 1 cruiser, 1 fighter, and 1 transport, WITH NO UNITS LOST IN BATTLE.

Borneo...Japan won. Destroyed 2 infantry, WITH NO UNITS LOST IN BATTLE.

Sea Zone 49....Japan won. Destroyed 1 submarine, 1 cruiser, and 1 transport, WITH NO UNITS LOST IN BATTLE.

Yunnan...Japan won. Destroyed 2 infantry, WITH NO UNITS LOST IN BATTLE.

So basically, Japan destroyed 80 IPC worth of units and lost absolutely nothing. My units had a combined defensive strength of 20...and not a single dead enemy unit.

Now, yes, I am fully aware this can happen. This one occurrence means nothing from a statistical standpoint. I understand this. And yet, I see nonsense like this happen over and over and over again. Something is wrong.

Or, maybe I really am just several standard deviations away from the mean, and out there somewhere is some guy who never has his bombers shot down, always rolls 6's for strategic bombing damages, and has magic paint on all of his equipment that prevents any sort of damage. I don't know.

[gbelle]

Webbs your above example is not exceptionally bad results. If this is round 1 you didn't attack his BB/CV off Indonsia. Then on Japan 1 they used that fleet and possibly another BB from japan to soak damage on sea zones 47/49. Yunnan is probably outnumbered good and overrun in one round. Sucks that they didn't hit but happens often. What attacked borneo? I can think of anywhere from 2 inf and 1 figthter (low end) to 1 inf, 1 art, 2 fighers or maybe just 1 inf,1 art, 1 fighter most likely. Any 3 of those scenario's for Borneo could easily end up with that result. Or most likely they just lose 1 inf.

Again I don't know how hard they went into those fleet battles but of all the attacks above the expected result may have been they lose 2 inf (1 in each battle), maybe a destroyer also if they couldn't clear your fleets in one round AND your fighter hit in both rounds.

[O Lucky Day O]

Yeah without the data on what was attacking, those results don't tell us anything. Every one of those territories can have similar results in Round 1 (and often do).

[WebbsNYC]

SZ47 was a disaster and they wisely brought the battleships to absorb some damage. The other combats were much more balanced and realistically should have resulted in some dead attackers. And yes, I agree...these results aren't absolutely crazy or unthinkable. But this type of thing should happen VERY infrequently, and yet I see these types of rounds quite often.

Just very weird outcomes. Like the events happening are 3 standard deviations outside the norm, and yet they are happening 20-30% of the time. Sure, on average they balance, because there is good volatility and bad volatility, so in aggregate, everything looks fine, but there are far too many very unlikely outcomes happening. At least in my view. That same game, the very next Russia turn, I had a similar outcome where I lost (badly) a battle that really should have been a fairly routine territory grab.

Just odd stuff with the dice.

[WebbsNYC]

Just took two fighters and a bomber against 3 German infantry (don't ask...it was necessary).

I get one hit from a fighter, the other and the bomber miss. He hits two out of three, destroying my fighters. I press because I NEED these guys gone. My bomber hits, his two infantry hit for 1 more, destroying my bomber and leaving him with a single infantry.

Possible? Obviously. Likely? No. No, no, no. In 5 dice, his infantry hit for 3. In 4 dice, my air hit for 2. Again...impossible? Hardly. but these are low-percentage outcomes that just happen over and over and over. I could sit here and write a neverending stream of these all day long. Am I getting some favorable upside as well? Sure. But these wild outcomes make the game way more unpredictable than I have ever experienced before, and more, in my opinion, than stats and probability would indicate. It's like we're playing with the opposite of low-luck...like "high-volatility dice" or something.