Start a new topic
Implemented

New Dice Algorithm / True dice rolls

Game does not use true dice roll probabilities. Many, many times I've had 97-99% chance to win and dont.


HOW TO VOTE FOR THIS FEATURE? Tap the 'Do you like this idea?' below


87 people like this idea

There is DEFINITELY something bugged with the new dice algorithm. I installed the game on a different device, created a new game account, and here was the dice stats after one single-player game with 5 AI: 1 - 27% 2 - 14% 3 - 14% 4 - 14% 5 - 14% 6 - 14% There surely is something bugged here.

1 person likes this
A lot of ansgt in this thread, partially I think because of the mixed messages ("dice rolls are completely fair" according to FAQ vs. "dice rolls need reworking" per this thread). Lee, perhaps if you could clarify, the noise would reduce...

1 person likes this

Well I had a game today, where 9 lost vs 1 twice. I mean the chances of this happening is somewhere in the ball park of 1:1000000. I don't think I'm that lucky. I don't get how it can be this hard to approximate true dice rolls. Any decent programmer can write an algorithm that can do it in less than an hour ;/


1 person likes this

Good morning Reginald,


I have also played this game for 40 years starting with the board game; my brother and I had a rivalry and played hours upon hours.  Our generation grew up playing many many  games that include dice such as monopoly... guess what role is statistically the most probable 7 right....  We grew up playing cards which is also a game of statistics.  The point is our generation is uniquely positioned to make observations about the probability of dice rolls without being PHDs in math; we have a pretty good sense for when things don't seem right.  Further you don't need to have a PHD to look up the probability of winning a 3 attacking dice on 2 or 1 defending dice where the defender wins a tie; the probablility information for ALL POSSIBLE outcomes is readily available on the internet.


While there are many responses and opinions on this board about this specific topic, please be aware that some of those posting also understand that when we played the board game we got frustrated by the results that went counterintuitive to odds.  AND we blamed our brother for not holding the dice high enough off the table or dropping the dice on the floor, surely that doesn't count.....  My point Reginald is that those of us that have played this game also understand that our emotional response to upside down odds is not warranted if we play enough games and the results SMOOTH out over time.  Do I get frustrated when I get autosetup with horrible board position, on top of multiple 5 card only turn ins in a row, coupled by the only option I get on fixed turn in is three infantry.  Do I get frustrated in the same game when my start position is last of 6 players, YEP because it is a BIG disadvantage.  Should I conclude that everything in the game is stacked against me because I've had upside down odds for one game, NO!  One game isn't a large enough sample set to make that conclusion.  The key to this whole topic is are the people posting on this topic just being emotional or are they going a step deeper and looking at the hard number results.


So I embarked on that quest.  I asked myself if I was gettting emotional or if that patterns I was seeing were correct; that I was getting CONSISTENT upside down odds on autoattack where I had the clear advantage.  I tracked 2 games then 4 then 6 then 10 and the results I got were not consitent with probability.  When I attacked 3 dice on 2 I was averaging around a 50% win ratio.  When I attacked 3 dice on 1 I was around 60%.  Those results are not consistent with probable outcomes; I could give you the probability for each of those outcomes but I will let you look it up yourself.  I did the same thing with board positioning, turn positioning etc and etc.  There were times I felt I was getting a raw deal but when I tracked a large enough sample set I realized that I was wrong and I was being emotional; everything is against me.....


I reached out to SMG support and asked what probability they expected for each scenario and while I could share these emails with you TRUST me when I tell you SMG's support staff have no clue what dice probability should be.  I was told that if I attacked 3 dice on one that I have a 90% chance of winning, REALLY Reginald, really!  I could only wish that were the case but I glad it isn't because that would mean when my opponent attacked me he would get those same odds and that isn't just statistically wrong it is galactically wrong and you don't need to a PHD to intuitively know it's wrong.  OK, OK that is just one guy who is 20 something that understands programming and game gliching but really isn't up on dice probability right?  NO!  I went back and forth for over a month with different analysts and they all quoted me different odds for the same scenearios in the same email thread and not once but over and over.  And none of them ever quoted me the correct odds even after I provided them links to Risk statistical odds by people that actually have PHDs in math.


So when I provided my spreadsheet of a sample set of 10 games and my results to a higher level analyst that seemed to understand he said he would have to look into it because those results shouldn't happen that way.  While you got a bad 10 game snapshot you say.  No I did this over and over painstakenly with basically the same results.  Can this happen, well of course.  But is it probable, NO.  When you start going down this road you have to be careful.  Can you attack 100 on 1 and lose, YES it can happen but is it probable NO!  Dont' tell me if that happened you wouldn't be in a state of shock even though you no it's possible....  That event isn't wrong but if happens with frequency IT IS!


I don't get freaked out when I attack 5 on 1 and lose all my men.  I get frustrated yes because it is statistically improbable.  But if this scenario happens a couple times per game over multiple games isn't it reasonalbe to say well that is an improbable pattern if it continues and if it continues game over game over game isn't fair for me to ask Support what's going on.


Reginald I have played over a 1,000 games on the SMG site and I can tell you that smoothed out over time my odds are roughly 50% attacking 3 dice on 2 consistently.  Have I had one or two games or dice rolls where I obliterate the other guy, YEP.  But it is just as emotional to EXPECT those statistically improbable odds to continue.  


When I asked an analyst what their dice algorithm odds were he said that he couldn't provide it because that was proprietary.  Well that's odd dice probability is NOT proprietary and that answer is problematic.


I too have had to play my same turn over several times becausee I kept losing connection, I concluded I was getting a weak wifi signal in that room and simply stopped playing the game there.  I've also uncovered a bug where I lost wifi and when I reconnected my opponent was granted a significant amount of men and this was confirmed by support.  So there are problems with the game to be sure.  But pound for pound I'd rather play it than the board game any day of the week because I don't have 2 hours to play one game.


Regards,


Rich




1 person likes this
I want to bring up that blitz and manual dice rolls likely use the same algorithm. The difference being that blitz doesn't wait for user prompts. I also saw some talk about 6 vs 3 battles behaving differently then other cases. With all due respect to those who said that, based on my programming experience it doesn't seem very likely. I seriously doubt SMG would write a different method or block of code for every battle size. What seems more probable is that there is one block for each possible dice combination (3 vs 2, 2vs1 etc.). When battles happen the program probably loops through the appropriate block until the number of dice involved changes. So I think 6vs3 works the same as 9vs5. The crucial part of the entire algorithm is the PRNG it uses. I think that is better place for this discussion to be centered on.

1 person likes this
For anyone who actually wants to check whether the dice algorithm is off, there is a fair amount of data out there in the form of recorded games on youtube.  For example https://www.youtube.com/channel/UCBsz1fgsud-bWjuf4PRr1KA.

I play risk just about every day and I don't think the algorithm is off.  I also record and review my games (but don't typically post to youtube).  I haven't noticed anything strange about the dice.  Sure, I lose battles I'm statistically favored to win, but that's how probability works.  What would truly be strange and questionable is if we didn't ever lose battles we were favored to win.  Humans are very good at noticing out of the ordinary things but not so good at noticing things that are as we would expect.  It's like how everyone has this idea that most drivers are idiots because they see a few dumb drivers each day.  But they don't remember all of the the good drivers, who are actually the vast majority.  The bad drivers stick out in our memory, but the good ones don't even though they're far more common.

My guess is all this complaining about "bad dice" is unintentional selective memory.  I bet if everyone complaining about bad dice would actually do some real number crunching and calculate the probabilities of their rolls using data from their games, they would discover that what happens in the game is very close to what would be expected from fair dice.

In future games try to pay specific attention to how often the dice behave more or less as you would expect.  In my experience it's the vast majority of the time.

1 person likes this
Well said, Mark. It is true that the mind only remembers significant things. And of course the people coming to the forums are ones who were pissed that something they didn't expect happened. It doesn't mean the algorithm is off. However, there does seem to be some issue with 1's being rolled less often, as the recorded dice rolls on all 6 of my accounts show less % in 1's than others. This wasn't the case prior to the implementation of the new algorithm, which was over a year ago now. But I do remember my dice rolls way back in the day were 16% across the board. Now it's like 14% for 1's and 17% for all others evenly. Perhaps SMG did something to give defenders a little bit more boost. As slightly less 1's would give more favor ability to the tie winner (defender) but clearly not enough to offset the actual advantage to the attacker, which can be seen in every game you play. It is also possible that coincidence has happened to me but there have been others making the claim and I've made other test accounts to check it. It's still such a slight variation that it isn't enough to notice.

Now, when they did their last update there was a flaw in the blitz. Every blitz move made against a territory with 1 was always a loss of 1 to the attacker before taking the territory. It didn't take me long to figure that out. I complained and it was fixed within a day.That flaw was obvious. No clue how the dice could possibly be flawed like that but I'm sure this game is a mess of code anyway.

 


1 person likes this
Yes I was playing with a friend and it was something like 15v4 and she lost all but 4 of her troops of this was an actual battle she would have like 12 left

1 person likes this
I have not been able to win ONE game since I started. put 8 hours into i think. Maybe that's just cuz I'm not good. but i reall just don't get it lol

1 person likes this
Ok so I got out my risk board and I played with my seven year old he won ! With all my strategy and skills I lost I am doomed! Because I suck at rolling dice I roll ones he Rolls sixes Hum....... but seriously while letting him win I kept a close eye on my army size and his plus the outcome and how real dice work .... I had 28 he had 20. At the end I had 7 Another one I had 9 he had 2 at the end he had 2 I had one that can't be right if I have 9 I should win... My point..... it is hard to understand when u turn in cards lay down 23 troops gonna do some damage . you are attacking ones twos fours maybe a five or a nine 23 should be more than enough wrong size does not matter dice do..... I have watched my fellow players get destroyed 4 vs 11 4 attacker he won legendary! The dice add that element of better not waste these armies . roll it out if you have no back up well better not risk it!,, If you can understand me try getting your risk board out setting up some battles get your kids to do it watch the epic battles it's a real uncertain thing ups and downs un less your my little sister she's so good at rolling dice I quit playing against her !!!!!

1 person likes this
Kevin, Im not one to advocate SMGs methods for dice, however what you explained pretty much invalidates what I have learned 9 years of university study in mathematics, probability, statistics, and a topic called "stochastic processes. First off, dice rolls are independent events. That being said, the likelihood of rolling any specific number on one die is equally likely. The issue isnt the RNG, which if Im correct, is the Merseinne Twister, the same one I use in Monte Carlo type simuations in my work.There may be some cognitive belief that the rolls give apparent short term patterns, however its a sure bet that the more rolls are made, the more the likelihood is a Uniform Distribution i.e. evenly distributed in equal likelihood in what is known as the Law of Large Numbers. Ive played over 3k games and have noticed anomalies in Blitz Mode. This is because SMG doesn't use a truly realistic Markov Transition agorithm for Blitz. Now that being said SMG likes to conflate issues when people bring up dice rolls. Its not the RNG for each roll but the values of the matrix elements in the Transition matrix that they, human beings have assigned or "biased". For example lets say you are attacking with 10 armies on a country that has 3 armies. Several calculations involving RNGs take place. First the number of dice you choose as an attacker up to 3, is a human input. Then a "roll outcome is simulated. A random number is generated, and the outcome is determined depending if that number, evenly distributed between 1 and 0, is either greater or equal to the likelihood of the specific outcome. Cheaters and hackers that design their APIs have learned to adjust outcomes in their favor by altering the likelihoods of outcomes in their favor. This is done by biasing the specific values for the Transition Matrix elements that govern the specific situation. In a true Markov process, or chain, the next outcome only depends on the present state, and not history. SMG in my opinion hasn't done their due diligence in preventing hacking APIs that adjust Transition matrix element values, or have designed a purposly flawed Transition Matrix apart from realistic board condition. The moral of the story is players must be cognizant of Blitz Mode anomalies not born from the RNG, but the algorithm that depends on it. If the TM is truely realistic then 10 v 3 should win more than what Blitz outcomes show in the long run. They need to document their TM element values are validated say with 100 billion dice rolls for each outcome possibility. A parallel processing computer can do that nicely. The other option would be for them to put yhe algorithm up on the cloud, but that would cost them, and eventually the player, more money.since compute time isnt free. Keep pressing these people to produce a more realistic Transition Matrix. If you see me out there, my handle is SMG Blows. Im tired of the Blitz Mode failures to reality.

1 person likes this
I've noticed my roll stats always staying at a consistent 16%. This by itself seems sketch to me. There should be some fluctuations here and there. If I know that even with a small sample of rolls such as a single game, that all players including myself will roll 16% equally, I have a slight inclination on when streaks are likely to occur. For instance, if I just busted something absurd like 3 armies defending and defeating 15 armies, I know that I am less likely to have that happen again. I'll take more chances than I would normally, like attacking a 4 with only 5 armies. This seems to work more than it doesn't. Anyone else have a similar experience?

1 person likes this
Its not the dice rolls thats the problem. The problem is tbe Blitz Transition Matrix probabilities that utilize tge spe ific random number to test against the predefined probability for each transition. SMG has defined the probabilitu for each transition.For example a 10 vs 4 transition probability has been defined as x. If the uniform ra dom number is less tgan x, then its a fail. The issue is their choice of the value x. They need to have a more accurate Transition Matrix.

1 person likes this
SMG, your loyal playerbase accept luck as a core mechanic of Risk and that streaks are inevitably part of that. Please don't do anything funky with the algorithm to placate those blaming "unfair" rolls...

1 person likes this
Here is my evidence
IMG0684.PNG
(735 KB)
IMG0683.PNG
(864 KB)
IMG0685.PNG
(736 KB)
IMG0681.PNG
(843 KB)
IMG0680.PNG
(739 KB)
IMG0679.PNG
(866 KB)

1 person likes this
Login or Signup to post a comment