Feature request: true random dice

[djensen]

NO! I do not accept these hits. (This was the same result twice in a row; actually the previous roll, they got 7 hits).

[djensen]

@JuliusBorisovBeamdog

• Some monitoring needs to be put into place for dice rolls.
• The PRNG algorithm needs to be updated to a less problematic one.

I hypothesize that you’re losing players from repeated bad luck. Maybe the number is low but without monitoring, there’s no way to know. Heck, I run this site and I’m probably going to take a long break from playing after this.

[redrum]

@djensen Or you could play TripleA instead :)

[djensen]

@redrum Yes, I could do that. I don’t really want to play Global. I would like some Anniversary games though.

And when I do play TripleA, internal dice roller FTW.

[JuliusBorisovBeamdog]

@djensen We agree, it sucks when you get hit like that, but thats what chance does. We’re trying to emulate the same idea of random chance in the board game that also relies on random chance as well as strategy. Axis & Allies is not a game about pure strategy. Chance plays enough of a roll that its more about hedging your bets and strategizing to give yourself the best chance. But its still just that: chance.

It always evens itself in the end. We noticed a concern being raised, we came out and addressed the concern with the dice RNG with the exact methodology we’re using for the dice.

[djensen]

We all know this is a game of chance. We all know that the dice can “turn on you.” All of us here have actually played the board game IRL dozens to hundreds of times (the same is not true of a majority of the Beamdog team; Trent is one exception). We know what bad rolls look like. We’ve rolled 15 physical dice and got one hit.

We even know that we can have unreasonable opinions about PRNG and start to become overly sensitive to results.

As @redrum has pointed out, not all PRNG are created equal and if Beamdog is using the problematic algorithm, why continue? If users are complaining why not be more transparent. TripleA has features to address this showing how well or poorly you rolled compared to average. (Now why would they do that? :thinking_face:)

[djensen]

The other thing, which I think you might already know. People sometimes come here to vent as a form of therapy after being traumatized by dice. :wink:

[djensen]

This post is deleted!

[djensen]

Go here to complain about dice: https://www.axisandallies.org/forums/topic/34180/prng-dice-support-group-and-ranting

Please keep this thread and others for constructive, actionable feedback and ideas. That includes me.

[djensen]

I put in a support request to implement Mersenne Twister.

There also should be some dice testing in place. A&A players are super sensitive to dice rolls. You don’t want churn because people believe the PRNG to be bad; you need to prove it to them it’s not.

[djensen]

I’ve been winding down my games of 1942 Online. In all likelihood, I won’t be starting new games. It’s just not fun anymore. The crazy volatility in the dice is ruining the whole game for me especially knowing there is a better PRNG implementation out there.

[redrum]

@djensen Don’t worry, even though you were tempted by the shiny new thing, we’ll still welcome you back to TripleA :)

[taamvan]

@djensen I agree, the potential is there but the volatility and frustration of dealing with the interface has turned me off completely. I’m sure I’ll be back–and will hope to see you at gencon again in 2020.

[djensen]

I actually like a lot of the changes in the interface. It needs work, especially around transports but I like it. I just can’t stand these dice.

[taamvan]

@djensen You’re more an IT guy than me…I’ve seen plenty of anomalous results and like you admonished me a couple days ago–when were playing dozens of games we should expect to see all the possible outcomes…still there is just this “dice vs RNG” debate that we can never resolve. As a slight departure; Chessex has 2 stands at Gencon–the dice are made to look great but whether they are in any way random…I have no clue. One might expect that if no attention is paid to how the dice are physically made the plastic mix must make some sets biased…is that any different than our debate? My IT guru buddy says that random.org uses fluxuations in the radio transmission intensity as a proxy for true randomness…is this all voodoo?

[djensen]

@taamvan Yeah, the idea for this topic came from random.org. I’d be happy with the following:

• Switch to Mersenne Twister or other good PRNG
• Provide user facing metrics about dice results, kinda like what TripleA does

Sorry, I didn’t mean to admonish; hope my tone didn’t come off that way. It’s also why I created the dice support groups. As long as the rules are followed anything goes there. That’s where I moved my reactions to dice rolls. I’m hoping here we can come up with a solution. Whether Beamdog does anything is up to them.

[taamvan]

@djensen I overstated “admonish”, entreated would have been a better way to put it…we are all trying to be patient and passionate at the same time…as usual. As for the technical details…I’m clueless except as it pertains to actual face to face posturing and dice slamming which is why we’re all here.

[djensen]

Well, based on internal testing in the past two weeks, apparently the dice are “fair.” So I guess I’ll stop making suggestions here and chalk it up to bad luck. (I will continue to rant on the “dice support group” since that’s what it’s for, complaining about bad luck and making irrational statements).

[djensen]

Also, who am I kidding. I’m not going to stay away forever. I just need a break or I need to play more games. Ya know, to even out the rolls because that how it works (it doesn’t; it doesn’t work that way at all :wink: ).

[JuliusBorisovBeamdog]

We have updated the FAQ on Steam.

Updated the FAQ about the Dice.

A: A PRNG generates the value. We’ve examined its output versus real random data and have found it to be indistinguishable. This also applies to the Computer opponent.

Based on the community feedback, 3-4 weeks ago we did pretty significant amounts of analysis of dice rolls in our game. As it turned out, our dice are not unfair.

We rolled about 100 million dice against our PRNG as one of our tests, and these were the results:

https://us.v-cdn.net/5019558/uploads/editor/ct/nlpqwbm4etm3.png

The dice results you see are from patch #5.

We’ve also done smaller batches of rolls more in line with the quantity seen in a single game and the results are quite similar. We’ve done 100-size, 1000-size, 1 million size, 100 million, and 10 billion.

Thus far with any of our tests, we have not been able to identify any instances of the dice being “streaky” in smaller batches, nor have we found any evidence of the PRNG being “bad” based on larger and smaller sample sizes.