doubles in Castles of burgundy

[FrankJones]

Also, please everyone remember, not all games log the stats correctly. I know I've mentioned this before, probably in this very thread.

Some gamed do not log stats correctly in abandoned or conceded games. And some games simply log the stats incorrectly regardless, because of the way the code reads unconfirmed actions and thus double counts some moves from the game log.

[lefalaf]

Also, please everyone remember, not all games log the stats correctly. I know I've mentioned this before, probably in this very thread.

Some gamed do not log stats correctly in abandoned or conceded games. And some games simply log the stats incorrectly regardless, because of the way the code reads unconfirmed actions and thus double counts some moves from the game log.

Not agreeing or disagreeing, but is there evidence of this somewhere? A developer with inside knowledge, BGA admin clarifying it for us? That sort of thing. Not speculation.

Seeing as we don't know how far back the stats are logged and if there is or is not a rolling lookback period that counts for stats, I have my doubts that this statement is anything but speculation.

I would expect abandoned to not count (as it is not a game), but conceded to count (as it is a completed game), but of course I have no idea what really happens as I didn't write the code, and BGA does do illogical things sometimes.

[FrankJones]

Also, please everyone remember, not all games log the stats correctly. I know I've mentioned this before, probably in this very thread.

Some gamed do not log stats correctly in abandoned or conceded games. And some games simply log the stats incorrectly regardless, because of the way the code reads unconfirmed actions and thus double counts some moves from the game log.

Not agreeing or disagreeing, but is there evidence of this somewhere? A developer with inside knowledge, BGA admin clarifying it for us? That sort of thing. Not speculation.

Seeing as we don't know how far back the stats are logged and if there is or is not a rolling lookback period that counts for stats, I have my doubts that this statement is anything but speculation.

I would expect abandoned to not count (as it is not a game), but conceded to count (as it is a completed game), but of course I have no idea what really happens as I didn't write the code, and BGA does do illogical things sometimes.

Yes. I am not just speculating here, I have analyzed this personally.

I am listed as a developer, though I don't actually write code for BGA games. But I know how coding works, and when is see a problem, I can usually figure out where the coding error is.

So, to the three points I referenced:

1) My best game, Earth: I noticed the end game stats were obviously wrong, so I took several games and went move by move and figured out exactly which stats were being double counted and how and why.

2) Abandoned games: You can just see the end-game stats. You can see by observation that the stats are partly or completely absent.

3) Same with Conceded games.

It seems each specific game's developer can choose how the stats are logged, which ones are shown, and how many of them, if any, are shown/logged for incomplete or conceded games.

[FrankJones]

Here's an example:

https://boardgamearena.com/table?table=889290077

The only stats logged are number of times each main action was chosen. This makes it seem as though I had 6 plant actions but no cards planted to my tableau.

And that was a conceded game, not an abandoned game.

Here is an example of the double counting:

https://boardgamearena.com/bug?id=181682

I have personally verified this myself to be exactly as described in the bug report.

[lefalaf]

I was just going to comment that upon further thought, conceding could be a mess as things were incomplete, so they really cannot be included without risking skewing some things.

Thanks for the examples. I'm surprised the dev couldn't track it down, but perhaps they didn't have time / are volunteer and pulled in other directions.

EDIT: So the conclusion for the original post that I take away from this is that close enough frequency of doubles to the expected % might be ok since we don't know how accurate the stats are to use as evidence that the doubles % is off.

[FrankJones]

I was just going to comment that upon further thought, conceding could be a mess as things were incomplete, so they really cannot be included without risking skewing some things.

Thanks for the examples. I'm surprised the dev couldn't track it down, but perhaps they didn't have time / are volunteer and pulled in other directions.

Yeah, It's a fairly minor thing, those stats are not all that important for that game. Number of times each action is chosen, number of germinate actions, those are some key stats.

The implementation of Earth is superb, and I'm glad the developer doesn't get caught up in majoring in the minors, as the saying goes.

And I'd much rather any efforts going forward are put into the newest expansion rather than fixing a minor end-game stats bug.

[Meamer]

As long as it's not 0, it's still possible. Therefore it will happen and is no proof for rigged dice.

Really. What if they were rolling doubles 90% of the time? The probability of that also isn't zero.

So it will happen at some point.

If you really want to proof anything, you have to check all rolls ever made in all games ever played on BGA. If you consider only one game and only the games one single player played, that's only a small excerpt and might proof nothing.

Even if every single roll in every game ever played was doubles... the probability of that is still not zero. It's stupendously, incomprehensibly small, but it's not zero. If your threshold for "proof" is a probability of zero, you cannot prove anything about a random system, ever.

[FrankJones]

Within reason.

If a million consecutive dice rolls came up doubles, we could reasonably conclude the function used to call a dice roll was coded improperly. If the probability of something is approaching 1 / Graham's number, or 1 / Tree(3), there's probably an issue.

[nik592]

and is costing me a lot of games

It really isn't.

Let's assume for a moment that the statistics are completely accurate and a coding error has somehow resulted in your true probability of doubles being 17.25% instead of 16.67%.

That's one extra double every seven games.

How much would that cost you? The impact of rolling doubles is highly variable; often you can easily pivot, occasionally it denies you a critical play, most of the time it's a nuisance you have to work around. But on average, it can't be worse than losing your entire turn.

This is the part that gets me. COB dice are essentially two numbers to work with. Higher or lower numbers (or same or different numbers) doesn't imply any advantage or disadvantage. I've frequently had rolls in COB where one number is pretty useless to me on a turn (that's not to say completely - at the very worst, you use it to get a couple of workers, which then lets you mitigate further rolls that aren't what you want), and sometimes where both are. As you say, you have to pivot to make the best of it. Even if half the rolls (or more) in a single game were doubles, I suspect you could still do quite well. I mean, the issue is more that if most of your numbers are the same across a game (which is not the issue OP is complaining about) that does make it a bit more difficult, as you run out of options pretty quickly. But again, at worst, you get some workers and you can mitigate some of the later rolls. You're probably still not going to do great on that game, but since the OP is complaining specifically about doubles and not (for example) "too many 3s", I don't think they're having a problem with the distribution of numbers (doubles or not), so I don't think that's the problem.

[Meamer]

Within reason.

If a million consecutive dice rolls came up doubles, we could reasonably conclude the function used to call a dice roll was coded improperly. If the probability of something is approaching 1 / Graham's number, or 1 / Tree(3), there's probably an issue.

Yes, you would reasonably conclude that, because unlike Ehronde, you aren't sitting there claiming "the probability isn't zero, what's the issue?"