Board Game Arena

Jellby wrote: ↑23 August 2026, 15:08 Actually, with RNG algothims you can prove it's 1/6 (or whatever). You don't need evidence, you have mathematical proof (as long as you trust/assume the computer behaves according to the specifications). With a die, especially with a die you just got in your newly purchased game, there's indeed no evidence it will be 1/6, or even that the rolls are independent.

Are you sure that's mathematical proof the algorithm is perfectly random, or proof that its reliability is above a certain threshold?

If there existed an algorithm that creates perfect randomness, why would we have different algorithms and graphs that show the differences between them? There would be only one algorithm and everyone would be using it.

I'm not sure how we would even go about creating true randomness. I mean, you could always create an algorithm that performs a countdown, or a system that looks at the current time and derive a number for that. But then that wouldn't be randomness, it would be predictable for the user. How do we create an algorithm that produces a series of numbers that are truly random seems like a question that remains a mystery. We can do something like go through the decimals of Pi, but there is no way to prove the sequence is random. Maybe we can prove every digit will show up an infinite number of times, but that's not the same as proving the sequence is random.

Jellby wrote: ↑23 August 2026, 15:08 By the way, I'd leave "fairness" out of the equation. Ramdomness is not fair.

Randomness is a pretty fundamental aspect of the fairness of many games.

Fairness means the game is proceeding as defined by the rules. If the rules of the game involve randomness, then following that without bias is what constitutes fairness.

People often will complain when they get unlucky and say that's not fair. But that's not true. The randomness is part of the game, so whatever happens is the fair result. To say otherwise would be anti-sportsmanship and not respecting the rules of the game as it was created.

Not fair would be if people tried to control the luck (in a way not allowed by the rules) to try to get the results they think they "deserve".

Now, admittedly, whether a source of randomness follows a perfectly exact distribution is something you may argue is not explicitly stated in the rulebook of most games. When you play a game that uses dice, it's probably not stated that each side must have a 1/6 chance of coming up. The rulebook just says "use the dice that is provided in the box". So it could be argued that the distribution being uniform is a human interpretation that we invent when we play those types of games, not necessarily something that is officially part of the rules.

After all, whether the actual probability for each side is 1/6 or some number very close to it doesn't fundamentally change the game. The outcome is still unpredictable for the players, and no one is intentionally being disadvantaged. You could even argue that if internet functions are not perfectly random, that's less a flaw and more something that authentically recreates the game experience, because in real life the dice are not perfect either.

Shivaware wrote: ↑23 August 2026, 19:25

BarnardsStar wrote: ↑23 August 2026, 18:47
Like I said, it's probably in the statistics gathering; it might be something about how the game handles the die rolls. It's not random_int, I know that. It could even be randomness, but when something this unlikely happens, I start to look for possible causes other than sheer randomness. So what is it?

I'm not premium, so I can't see these statistics.

I seem to recall someone going on and on about statistics not being recorded if someone concedes. Is it possible this is what's happening here?

And if that does happen, maybe someone is choosing to concede games when there have been more doubles than usual just to skew the results

I think this is the most likely explanation. Given that it is virtually impossible that double rolls and conceding would be perfectly independent events (especially since like you say players may see it as disadvantageous), then we have to conclude that not counting the games where someone conceded would have a slight impact on the statistics.

tbhp wrote: ↑25 August 2026, 15:31 Are you sure that's mathematical proof the algorithm is perfectly random, or proof that its reliability is above a certain threshold?

I didn't say there's proof it's random. In fact, a deterministic algorithm is by definition not random. What you can prove is that the probability of getting some result over a large number of samples converges to something, or that the results of the algorithm have some other property (but not all, that's impossible) that random samples should have. But don't ask me about the specifics of the proof. I'm just pretty confident some of these facts have been proven, but I don't even know the details of the algorithms.

I'd just like to observe that the global "all players" proportion of doubles now reports 16.7%. I think the "all players" statistic may just be reporting recent values? Otherwise ... when I made my post two days ago, there had been 4369880 games with 16.61% doubles, and now it claims 4376948 games with 16.7% doubles, which means the last 7068 games would have had to have been something like 72% doubles.

I think it's far more likely that these percentages are based on some recent window of games, in which case fluctuations like this are far more normal and expected, and this whole thread has been for nought.

That's bizarre. Seems like an asinine way to display statistics. Especially without indicating as such.

BarnardsStar wrote: ↑25 August 2026, 21:02 I think it's far more likely that these percentages are based on some recent window of games, in which case fluctuations like this are far more normal and expected, and this whole thread has been for nought.

I wonder if it's the same window they use to adjust the max time and increments for real-time games.

Given that what they call "all time" ranking ignores those who haven't played in the last ~2 months, I wouldn't be surprised if the "all players" statistics had similar restrictions.

Maybe they want the stats to be reflective of the current culture or "meta" of the game? I mean, in the example of double rolls here it doesn't make a difference since that's not a player's choice, but for all the things that are dependent on what players do, the stats can differ over time (though in that case, you'd think it would be cool to have timelines that show you how they evolved).

This makes me realize, this is an actual advantage that premium players get, because by viewing the stats they can be more aware of how likely players are to pick certain things, make certain choices, which can be important if this is a game where knowing the "meta" can give you an advantage (though a seasoned player may have an intuitive understanding of those things that may be even more accurate than looking at the stats).

Since I posted about a month ago, my doubles % in COB has actually gotten WORSE! I now have a 17.25% double rate over almost 90000 rolls in 3900 games. Statistically the chance of this is beyond ridiculous and is costing me a lot of games. I realize there are streaks of bad luck but COME ON. The actual statistical chance is 1 in 628259, or .000157%. Come on Dave, gimme a break...

Lol. What a thing to be completely stressed out about in life. And to think someone is actually tracking this. And worried about it. Especially since that stats for nearly eveything at bga is skewered a bit. .Sore losers are obsessed with excuses