Board Game Arena

MoiMagnus wrote: ↑22 March 2021, 16:01

Ze Monstah wrote: ↑22 March 2021, 15:35 To Romain, again: Anyway, regarding the info you provided, I do not comprehend it 100%. Exactly like Faker above, I consider some of the wordings people use here, to be a bit... hard to get, mostly because of language barrier, but not only.
It is like a medical doctor talking to patients with specific medical terms, without explaining the words, forgetting that not everyone has the same level of scientific knowledge as he/she does.
The situations are not the same, but both seem rather snobbish to me.
Still, a solution for me would be (unless you use easier words or elaborate the heavier ones you use), to improve my English, a thing I am actually going to start doing.

What Roman is saying is that if winning/losing was pure randomness (flipping a coin), then during 84 games you should have in average:
+ 1 sequence of 7 successive wins (or successive losses)
+ 1 sequence of 6 successive wins (or successive losses)
+ 3 sequences of 5 successive wins (or successive losses)
+ 5 sequences of 4 successive wins (or successive losses)

Your games have more sequences than that, but
(1) 10% of players have similar sequences as you do. It is not weird.
(2) Games are not coin flips. You win more often than you lose. And some days you are better at playing than others.

In conclusion, you have the feeling that you are having long periods where you lose and long periods where you win. But that is normal, that's how true randomness works, sometimes. And other times you will quickly alternate wins and losses, because that's also how true randomness works.

Thanks for the explanation. If that is what he was aiming for, I now understood. I think my maximum number of consecutive losses was 15 or 18, i will check. It was like 5 or 6 months ago.
For me this is not ok.
And i want to contradict you on nr. (2): games in general might not be coin flips... but backgammon is a game which should be a coin flip for decent players (by decent i mean, players who don't take unnecessary risks, and make the best decisions, increasing their chances and decreasing their enemies')! Unless a player changes tactic in a given time period, or if he suddenly goes totally tactic-free in her/his play, for some reason.

Also... the number of consecutive losses or wins, was only a small part of the details i came up with. And the number of consecutive losses, for me is not ok, while for you, it is ok ("Your games have more sequences than that, but (1) 10% of players have similar sequences as you do. It is not weird."), because even if 10 losses in a row, might seem bad or fishy for me, for you it is acceptable because... more players (10%) get those, and because... it is simply acceptable.
I guess that even if i came up with a series of 25 consecutive losses (not on purpose), you would also consider it PURELY RANDOM.
My logic: you would consider ANYTHING random and acceptable, despite any number, statistics, and so on, so I will not debate this any further.

I respect and value your opinion (even if i disagree with it), but I think I will remain with my opinion, especially about a possible hidden algorithm bug, and stick to games which tend to 0 luck needed.

Thank you for the time you took to write the replies about this subject.

N_Faker wrote: ↑22 March 2021, 15:47

Ze Monstah wrote: ↑22 March 2021, 15:35 And the "cheater" and "ELO boosted account" terms used... I cannot say anything, because I don't consider myself related to any of those; if anyone does, feel free to bring arguments.
But of course... perhaps I misunderstood what he said.

I'm not talking about you.

Cheating on BGA largely consists of multiaccounters playing full games against themselves, some have gone the extra step of avoiding explicitly losing plays and other plainly obvious indicators. Which in the case of backgammon has caused them to end up in losing positions with the account they intended to boost due to the unpredictability of the dice.

Of course you don't...

So in a more flexible English, what you are trying to say is: if i have 2 BGA accounts (A and B) and play backgammon games with both accounts at the same time (with the intention of making B lose to A but without making B's moves look too obvious, in order to not become suspect of ELO BOOSTING), things might not go my way, because the dice might unluckily make B winner, instead of A?

This seems a bit hard to happen, but i think it would depend on how fard B's "fair" acting would go, and the dice numbers they get.

Also, i heard that he/she could turn the game from real-time to turn-based during game, and no one will be able to replay his/her game (because of a BGA bug), so no "fair" acting needed for B, it can play as bad as it likes, because no replay will be possible if the game is turned to turn-based (this is a tip for the cheaters).

EDIT: But anyway, you knew about this "real-time to turn-based" glitch which lets people cheat without being caught (the impossibility of replays), you know and admit that in the absence of the "turn-based trick", this thing is difficult to happen (even if B tries really hard not to be suspicious), and this thread is not about cheating, but about some suspicious luck; so I don't understand why you brought this thing in the first place, knowing that it has no connection to the OP's initial post, and most of all, that it could be interpreted (by the way you phrased it) as a provoking reply regarding the ARENA game you accepted to abandon. Sorry if I am wrong.

Ze Monstah wrote: ↑22 March 2021, 16:34 ...because the dice might unluckily make B winner, instead of A?

Yes.

Ze Monstah wrote: ↑22 March 2021, 16:34 This seems a bit hard to happen, but i think it would depend on how hard B's "fair" acting would go, and the dice numbers they get.

It is rare, especially as they would play poorly on the 'losing account'. Yet, when the scale is grand enough, even rare things will become common.

Ze Monstah wrote: ↑22 March 2021, 16:14 I think my maximum number of consecutive losses was 15 or 18, i will check. It was like 5 or 6 months ago.

With 2k8 games, you should have a max series (in average) of 11 to 12 (assume to simplify you have like 40% chance to have 12).
That means 13 is about 20%. 14 is 10%. 15 is 5%. 16 is 2.5%. 17 is 1.3%. 18 is 0.6%.

If you got 50% chance to win each game.
Again, we are pretty much at the limit. You could argue for either side with it.

Ze Monstah wrote: ↑22 March 2021, 16:14 Thank you for the time you took to write the replies about this subject.

I've quickly made a google sheet:
https://docs.google.com/spreadsheets/d/ ... edit#gid=0

It randomly flips a coin (with a probability of 55% of Win/Loss, which is your winrate but you can change that)
And for me, in 50 coins, I already had a sequence of 9 consecutive LOSS, though it didn't happened again after trying two more times.

You can check by yourself, and see what true randomness look like, and how sequences of win/loss actually happen.
(If you want to reroll all the randomness, just write anything in any of the white cells and press enter)

My logic: you would consider ANYTHING random and acceptable, despite any number, statistics, and so on, so I will not debate this any further.

With roughly 3000 games, it would be unlikely for you to never have had 10 consecutive wins or losses at some points.

But you are correct that any single number would hardly convince me, as what you're describing is still in the order of magnitude as what my statistics course told me was "normal". A 25 consecutive win or loss, or anything more than 20 consecutive win or loss, would puzzle me enough to check how likely it is.
(In fact, I might do it latter this week anyway)

EDIT: I fact, I did it right now.
1 every 1000 players would obtain a sequence of 25 consecutive wins or consecutive losses, assuming 55% winrate and 3000 games..
=> Very possible. But definitely fishy.
1 every 50 players would obtain a sequence 20 consecutive wins or consecutive losses, assuming 55% winrate and 3000 games.
=> Totally normal.

If I got 15 fair losses (by fair i mean that i tried to win and played my usual normal way) in a row (5% chance from what you say), you would say: 5% is rare, but things like this happen in a luck-based game, you are just unlucky.

If i lose 25 or 100, you would say the same: really rare (like a legendary pokémon) but still... bad luck.

Not to mention other details and the probability of a problem with the algorithm...

I see no reason to debate this any further, I have no desire to convince you of anything more, and it is just a choice to engage (or not) in luck-based games, in the future.

I got in this thread to answer the OP's question, and not to upset any of you or the moderators; but wrote more details and explanations.

I might come up with the the maximum nr. of losses in a row, in my games (at least 12), when i have time to look in all the ~2000 games
.
If I managed to upset any of the ones I mentioned, I am sorry, nothing I can do now.

Good week, stay safe and play chess!

EDIT:
To meeple(something): My limit to "exaggerated losing row" is 10, so a lot lower than the co-repliers'.
I guess the reason you insisted on the loss row is that it is the only thing that can be verified without the limited replaying.
I want to mention that i brought it up, in corroboration with the other details, and that my tolerance for fishy is lower than most of yours', the defenders.
There are too many games to look at... 2900 I know that the maximum is 12, 15 or 18 (i remember it was a 3x kind of number), but I doubt i ever reached 20, in order to satisfy the replier from the above...
Is there a program or a way to check all 2900 games for successive losses or wins, without doing this manually?
And Magnus, if you found the time to do the math presented above, and expressed your opinion that everything is acceptable and random, even if it might be fishy, could you please do the counting of my games, as well?
Thanks in advance.