Cheating or luck?

[shaun hoversten]

Those talking about the odds are discounting the fact that OP had to roll a non-1 roll also. Now, having pips on 5/6 vs 2/3 is different, but assuming it was on 2/3 and all he had to do was not roll a 1, there are 11/36 rolls that have a 1 in its result, so in that moment, the odds of losing were 11/36 x 1/36 since both things had to happen. Thus, 11/1296 or just under 1%

I’ve stopped playing BG on BGA bc I’m a math major and the number of times the “not impossible but improbable” happens is statistically impossible. I’ve had games where I was smashing opponent and they got four doubles in a row to win. Odds? 1/1296. Odds when it must involve double 5s and 6s is even worse.

To each their own, but I have played many a games with a local in-person BG club and while things can happen, it is very rare, and def not in the line with this sites computer generated dice.

Though to be fair, in person “real games” involve a cube so a person would decline the cube before that miracle run of dice happens

[Remkar]

Those talking about the odds are discounting the fact that OP had to roll a non-1 roll also.

To be fair, they didn't say what pips they were on, so we don't know that they specifically needed a non-1 here.

I’ve stopped playing BG on BGA bc I’m a math major and the number of times the “not impossible but improbable” happens is statistically impossible.

Excellent, since you're a math major, you have access to math and stats professors. Maybe ask one of the stats professors about the differences between post hoc and a priori statistical calculations, then come back here and let us know how that changes all of your calculations and assumptions.

-signed, a retired professor who has taught statistics

[shaun hoversten]

Not sure why you mention priori other than there are a finite amount of outcomes with the dice (36), but each outcome most certainly does not have an equally likely chance to occur.

And post hoc I understand bc you’re talking about how after something happens, we perceive it to have occurred more often than it should, but the fact remains, if you play in person, and each person takes 1000 turns, statistically only once in those 2000 turns should someone have rolled four doubles in a row.

Play ten games on BGA and you’ll see it at least once. When you’re at an end game scenario and the only chance they win is they need X roll, followed by you have Y roll: and they need Z roll to finish , it happens far too often to be random chance.

A simple calculation can be done (generate any scenario that matches that) to see what the likelihood of it happening twice in 5 games is, it’s near 0, yet happens all the time on bga or you wouldn’t notice it.

[Remkar]

Not sure why you mention priori other than there are a finite amount of outcomes with the dice (36), but each outcome most certainly does not have an equally likely chance to occur.

Sorry, short on time right now, but that's not what I was referring to. Take a look at family-wise error rates in relation to post hoc vs a priori analyses and compare that to how you're approaching this.

[euklid314]

[...] if you play in person, and each person takes 1000 turns, statistically only once in those 2000 turns should someone have rolled four doubles in a row.

Play ten games on BGA and you’ll see it at least once. [...]

I have to agree with Remkar that for a math major you are arguing only with gut feelings - without any statistical content in your posts. Hopefully he will find time to elaborate on the general loopholes of your argumentation.

I will only focus on the only concrete example you gave - see the citation above.

You correctly note that one player getting four doubles in a row has a 1/1296 chance. That is, if a single player rolls 1300 times, he is expected to get this once (on average). If each person takes 1300 turns, i.e. both players take 2600 turns than this should happen twice.

Since an average game of Backgammon has 25 rolls per player, we can estimate the occurrences of "four-doubles-in-a-row-for-a-single-player". Note that we would not notice it if one player has 1 double at the end of one game and 3 consecutive doubles at the beginning of his next game. Thus we have to consider every game separately. In 25 rolls there are 22 sets of 4 consecutive rolls (rolls 1-4, 2-5, 3-6,..., 22-25). Same for the other player. We have 44 such sets of 4 consecutive rolls in an average game.

Therefore it should happen approximately every 30 games (30x44 = 1320) that one of the players has such a series. You claim that this happens every 10 games on BGA. It should be very easy for you to prove this claim.

Take the Elo No. 42 on 15.10.2023 (we do not know today who that will be, so it is somewhat arbitrary) and analyze his last 60 games played before 15.10.2023 (even better: his next 60 games after 15.10.2023). You claim that you will find 6 occurrences of such 4-doubles-in-a-row-series within these 60 games. Good luck!

P.S.: You should be very fast with this check - even by hand - because in almost half of the time a player will not even roll 4 doubles in the whole game - let alone in 4 consecutive turns.

[shaun hoversten]

Zzzzzz

Do you play BG in person? Or only online (bga/apps)?

A person with a general knowledge of probability can calculate the odds of X occurring. Do you want to leave that pip exposed? “They have to roll a 10 to hit it, so odds are 1/12 or 8.33%” and can make decisions from there.

In any small sample size, anything can happen, but theoretical probability will always trend closer and closer to experimental the larger the sample size.

So when you are playing; and calculate the odds are 1% you will lose (going back to original situation) and it happens, you think “well s***, that was bad luck”

My point is on BGA, that 1% (or more common maybe an 8% play) happens far too often.

A simple p-hat calculation can take a situation that is 8% likely to occur and tell you the chances it will happen X out of Y times. Bga gets to the near impossible in terms of number of times a 1-10% situation plays out.

I haven’t played on here in a long time hence why I have no raw data to provide you; but it really isn’t hard to understand.

Think those of you jumping down my throat must work for bga or sold them the RNG software

[RicardoRix]

so are you suggesting the RNG somehow takes into account previous rolls? And perhaps the state of the game, like it's near the end?
Knows the state of the pieces, and knows what you need to knock a piece off the board?
Does the RNG even know it's used to throw 2 dice?

Sounds like a lot of effort to program all that.

Here are your stats at backgammon to check
https://boardgamearena.com/playerstat?i ... 49&game=53

[Jellby]

Narrativium: one-in-a-million chances crop up nine times out of ten.

[euklid314]

Do you play BG in person? Or only online (bga/apps)?

I am a tournament player (real life tournaments with doubling cube, obviously using precision dice) and I am a mathematician.

I gave you a possibility to show that your claim was correct. You don't need to prove that with your own games (but if you prefer, you could - by analyzing your last 60+ played games). Analyzing one game takes approx 1 minute since you do not have to replay it but only search in the game log for doubles.

Instead of making a claim and showing it is true, you still argue with your gut feeling. I obviously accept that your feeling tells you something is wrong. But if you want people to take you seriously you have to show us that something with the RNG is wrong and not just with your feelings.

[shaun hoversten]

Apparently I’ve insulted the cult that is BGA BG players.

This is a thread about luck vs cheating so god forbid someone chimes in on how often “unlikely” rolls take place

I never said bga cheats but on a site where participation is critical, having an algorithm that gives favorable rolls to a lesser player and makes outcomes more 50/50 is not out of line.

And getting chastised bc how complex that would be, are you serious? In a world of AI and chap GPT where software allows us to clone voices and images and make nearly indeterminable replicas of people identities, you’re saying an algorithm that gives a losing player a couple extra doubles, or more 6/5 rolls at end game (which guarantees two pips come off vs a 4/2 roll) is unrealistic? Please

So a player you would beat 80% of the time in real life (take away cube) you’ll beat 80% if the time virtually?

I’m not saying I have proof it exists but the mere volume of criticism for even suggesting a computer dice simulator could favor one person is absurd