Statistical anomaly on obtaining /!\ ?

Forum rules
Please DO NOT POST BUGS on this forum. Please report (and vote) bugs on : https://boardgamearena.com/bugs
User avatar
Murrur
Posts: 28
Joined: 28 March 2020, 15:06

Re: Statistical anomaly on obtaining /!\ ?

Post by Murrur »

if you get warning on every 1/6 or 2/6 turns, then it's not random at all, it's very predictable. Truly random numbers may repeat same number for longer times and get lots of warnings on another times. To be random, it needs to be random also in Time Scale, sometimes randomness concentrates sometimes it expands.
User avatar
agodd
Posts: 173
Joined: 20 March 2020, 23:42

Re: Statistical anomaly on obtaining /!\ ?

Post by agodd »

frafa wrote: 12 June 2020, 19:00 Hello

As of today, on game statistics, for all players, obtaining /!\ on dices with 1 /!\ are evaluated at 15.49%. For dices with 2 /!\, it's 29.23%. As it's 6-faced dices, I expect results to be close to 16.67% and 33.33% respectively.

Considering the huge amount of rolls done by all the players, isn't it strange that there is such a discrepancy between theoretical and actual probabilities?

(It's good for the players though, they should lose control less often ;) )

It makes me wonder. If someone has an explanation... :?:
Maybe it was due to the Grren Die bug discovered few days ago
User avatar
Phoxtrot
Posts: 287
Joined: 03 January 2012, 20:55

Re: Statistical anomaly on obtaining /!\ ?

Post by Phoxtrot »

agodd wrote: 17 July 2020, 15:34
Maybe it was due to the Grren Die bug discovered few days ago
Lol. Wrong way around.
The stat is too low and the green die bug (if not purely cosmetic) was actually increasing the risks.

Nah, it's as many people said, insured dice are counted.
(Or possibly something with dice rolled but not re-placed after a loss of control, but I doubt it as it can't be all that statistically significant)
User avatar
frafa
Posts: 8
Joined: 31 March 2020, 22:17

Re: Statistical anomaly on obtaining /!\ ?

Post by frafa »

ObiQuiet wrote: 30 June 2020, 02:07 Where do these stats come from?
As of today, on game statistics, for all players, obtaining /!\ on dices with 1 /!\ are evaluated at 15.49%. For dices with 2 /!\, it's 29.23%.
It was from the game statistics on my profile for the game. You can view your stats, the ones for all players and the ones for winners. These were the figures for all players at the time. It hasn't evolved much since then. Still below 16% and 30% respectively.

But my concern was for the overall statistics, not my own, or ones from a specific game. The sample would be too small to judge. I think the first reply in the thread got it right. The secured dices must be included in the computation of this percentage to explain these results.
User avatar
frafa
Posts: 8
Joined: 31 March 2020, 22:17

Re: Statistical anomaly on obtaining /!\ ?

Post by frafa »

diamant wrote: 29 June 2020, 19:01
frafa wrote: 12 June 2020, 19:00Considering the huge amount of rolls done by all the players, isn't it strange that there is such a discrepancy between theoretical and actual probabilities?
No statistical result can seem strange by disconnecting it from the data set from which it is derived.

I’m willing to bet that what you call "huge amount" is a ridiculously small number for a sample size, leading to a wide fluctuation interval to conclude that the probabilities observed in considering your tiny sample match the theoretical probabilities.
"My ridiculously small number" is just the overall number of rolls made by all players in all games of RallymanGT on BGA, showed in the "statistics for Rallyman GT" screen in your player profile. It probably amounts to millions of rolls. It should be plenty enough to achieve statistical convergence, on such an simple event (rolling a 6-faced die).
User avatar
diamant
Posts: 576
Joined: 18 April 2016, 16:39

Re: Statistical anomaly on obtaining /!\ ?

Post by diamant »

frafa wrote: 28 July 2020, 18:58"My ridiculously small number" is just the overall number of rolls made by all players in all games of RallymanGT on BGA, showed in the "statistics for Rallyman GT" screen in your player profile. It probably amounts to millions of rolls. It should be plenty enough to achieve statistical convergence, on such an simple event (rolling a 6-faced die).
You would be right if all the rolls were counted in those statistics.

The bug that allowed a player to replace only some of his dice after a Flat Out and Loss of Control has only recently been fixed. There is no evidence that the dice which were not set back were included in these statistics, which can therefore be biased.
User avatar
mumpsimus
Posts: 8
Joined: 13 July 2014, 02:04

Re: Statistical anomaly on obtaining /!\ ?

Post by mumpsimus »

I've manually gone over a short 1 lap game and kept count of all the dice and warnings, and I've found that secured dice are counted as rolled, which will account for the numbers being totally off in each individual game, and thus also in the aggregate statistics.

(If you want an example game that is short enough to do it with pen and paper, or a spreadsheet: https://boardgamearena.com/table?table=251856162
RH secures 7 dice in this game and rolls 62, and jokersson secures 2 and rolls 54)
Post Reply

Return to “Rallyman: GT”