The site's dice are accurately pseudorandom.

D12 does not use mt_rand according to the bbcode snippet Matty linked me to. Nor is it what Cireon talks about on his page. They both are talking about the same thing but it is not mt_rand.

I said that is what we were using based on this post:

https://dominating12.com/forums/2/general-discussion/879/the-sites-dice-are-accurately-pseudorandom/post/14120#post-14120

I will say, however, that I am not a programmer nor do I pretend to know what is involved. I was only making comments based on my own research and what I have read about it. I am in no way an expert and I apologize if I have misread or misinterpreted anything the wrong way.

We used mt_rand in the past, but not any more.

Dough_boy, thanks for sharing the experiment. Personally I think 10k is still a pretty small sample size and the numbers we're seeing in the results really look like they could be caused by random chance. Have you tried repeating the experiment multiple times to make sure these results are representative of the actual distribution?

It was 40k. What would you consider to be a valid sample size? I think I am going to add one more test too.

I will send you the code and excel file so you can vet too.

We run with 100k samples here. I haven't done the math, but if your sample size is 40k, then all three results look like their error is of a similar magnitude as what we have now, but feel free to tell me if I get that wrong. I think it's a good litmus test to run the experiment again with the same sample size, just to check whether the conclusions would be the same.

I play with actual dice several times a week.....  I rarely see the disparaging results on the attacker. I might be crazy  but it really seams the dice are weighted against the attacker...…  at least that's my assessment after playing with actual dice 3 games a week for almost 20 years...… 

Do we keep track of each roll? Seems like that would be good stats to have (could report on anomalies, stats around how many of a combination are rolled, etc).

I swear I just had a turn where I had 13 on 3, lost 10. But the last 4 rolls my first two dice never seemed to change. Almost like it was stuck at 4 & 2.

Keeping track of rolls is on our todo list (or at least the number of times you get a certain roll), but it's a lot of data that we'd need to store (especially in balanced dice where there are hundreds of different numbers you can roll!) so it needs some thought and consideration to make sure we can handle that data.

Yeah. Wonder if at least in the short term it should just be logged? I think there are only 294 possible attacking combinations and 42 possible defensive combinations. Each one could be assigned an id, so 6 | 6 | 6 = 294.

Maybe after 30 days the tie back to a specific game/user would be gone and just the overall count would be left?

I think it's fine to store some metadata/statistics for each game a user played in, as long as we limit how granular it is.

We might just say that for balanced dice we don't keep track of rolls (because I don't really want to track rolls for a 130-sided die).

We also would just track the actual die rolls 1-6 I think. Maybe we can also store the different outcomes (lose 2, draw, kill 2 for 3:2), but then we already start adding quite a few cases.

I would agree on the balanced dice.

The reason that I like the ID is it is easy to group by for counts. But if you wanted individual that would be fine too.

I was thinking roll_id, "log_id" (assuming ties to game, and maybe user), attack_dice_id,defender_dice_id,outcome_id,stamp

With something like that it would be interesting to be able to run numbers round most common rolls, outcome, etc.

Would storing only 3v2, 3v1, 2v2, 2v1, 1v2 and 1v1 battles won/lost per game be sufficient?
We could store that without wasting too much space.

I would personally like to be able to see the difference between balanced dice and normal dice

Edit: If it's not sufficient, maybe store that per round?

To me the complaint is around "dice not being realistic". If we showed that the defender had an equal distribution of defense, so roughly "even" on 42 possibilities, and the attacker had an equal distribution of attack, then it should put anything to rest.

So say in the case of a 3v1...on 60 rolls was it roughly 10 of each dice for the defense? Or did the defense manage to roll more 5/6's, etc. Theoretically over thousands, hundreds of thousands of real battles they should start to equal out.

Also, not sure how many premiums we have, but what if it was only stored for premium members? Still gets data, but likely less.

yes i am posting again....the dice programming simply sucks any fun out of this site....I just had a turn where I attacked 4 times with big advantages...
1. killed 4 lost 17
2. Killed 2 lost 8
3. killed 2 lost 6
4. killed 1 lost 7

Yes, that is a total of killed 9 lost 38.......this would never happen with real dice, I dont care if i rolled 1,000,000 times...
I should add that during the course of this turn, the site kept having problems functioning, I had to refresh page several times. Perhaps this was causing constant resets into "defense rolls hi mode" or whatever fixed program this site has..