is there anything up with the dice roller source code? doing random roles seem to get strings of good rolls and then poor, then repeat...
I to would like to express my frustration with the die roller... But maybe that's because I can't banish the offending dice back to the dice bag.
I haven't noticed anything like that. My last five d20 rolls were 16-15-1-15-6, which seems fairly random.
Interesting story. My roommate in grad school had a girlfriend who was studying statistics. She sat down one night when bored and rolled his dice for D&D over and over and over and over, enough to get a broad sample.
Her conclusion: "These dice do not behave the way they are supposed to."
While I've also seen the die roller do weird things (3 of us rolled 1s in a row, and I think at least one of us rolled 1 again after that), I am more likely to blame game dice irrationality than the thing not functioning.
ETA: I don't know what those die rolls above prove except I know is nearly impossible for me to roll in the middle range of a d20 (and never a plain old 10), which is consistent with my personal die roll luck, which can be verified by others.
Senior Software Developer
|
| 4 people marked this as a favorite. |
The dice roller is basically as random as it's possible to get a random number generator to behave without hooking up a lava lamp to the servers.
I do find the idea of providing personal digital dice bags that you can select from for rolling particular things amusing enough to think about how to implement next time I'm stuck in traffic or in the shower.
| 2 people marked this as a favorite. |
The dice roller is basically as random as it's possible to get a random number generator to behave without hooking up a lava lamp to the servers.
I do find the idea of providing personal digital dice bags that you can select from for rolling particular things amusing enough to think about how to implement next time I'm stuck in traffic or in the shower.
So what you're saying is we can have a lava lamp in the server room?
|
Interesting story. My roommate in grad school had a girlfriend who was studying statistics. She sat down one night when bored and rolled his dice for D&D over and over and over and over, enough to get a broad sample.
Her conclusion: "These dice do not behave the way they are supposed to."
That could be beacuse those dice aren't uniform. It's a pretty common problem with making dice. Ever notice some dice have far more rounded edges than others or have you tried stacking dice.
Take your all your d20s and make a stack with half being sure to stack the 20 face up. Take the other half of them and stack them with the 20 facing to the side (the 9 face up on each). If those two stacks aren't the same height you probably have egg shaped dice. That means you're likely to have a greater chance to roll a smaller range of numbers.
The best dice have nice sharp edges and corners. Those are good indicators that the dice are more uniform.
--Vrock and roll the bones...
She used a variety of well-cared for dice, many of which were precision cut. Even if some were worn or oddly weighted, that shouldn't have altered the total results so much, I don't think.
Gamer dice just run by their own rules, man. Can't convince me otherwise.
Next time I actually roll a 10, I might reconsider. I'm gaming tonight, we'll see.
Just for the sake of extending the examples:
Analysis:
1 - 55 (5.5%)
2 - 42 (4.2%)
3 - 49 (4.9%)
4 - 48 (4.8%)
5 - 60 (6.0%)
6 - 53 (5.3%)
7 - 46 (4.6%)
8 - 51 (5.1%)
9 - 49 (4.9%)
10 - 55 (5.5%)
11 - 56 (5.6%)
12 - 54 (5.4%)
13 - 45 (4.5%)
14 - 46 (4.6%)
15 - 38 (3.8%)
16 - 44 (4.4%)
17 - 53 (5.3%)
18 - 46 (4.6%)
19 - 56 (5.6%)
20 - 54 (5.4%)
Or, if you want to sort by frequency:
5 - 60 (6.0%)
11 - 56 (5.6%)
19 - 56 (5.6%)
1 - 55 (5.5%)
10 - 55 (5.5%)
12 - 54 (5.4%)
20 - 54 (5.4%)
6 - 53 (5.3%)
17 - 53 (5.3%)
8 - 51 (5.1%)
3 - 49 (4.9%)
9 - 49 (4.9%)
4 - 48 (4.8%)
7 - 46 (4.6%)
14 - 46 (4.6%)
18 - 46 (4.6%)
13 - 45 (4.5%)
16 - 44 (4.4%)
2 - 42 (4.2%)
15 - 38 (3.8%)
For what it's worth, Excel's CHITEST() function says that p = 0.889016307. I'm pretty sure that means that there's about an 11% chance that these numbers are not truly random, and an 89% chance that they are. I might have that backwards, though.
I can run a 10,000 roll analysis if anyone wants...
Actually, I can't. I tried it, but I got an error when I went to post it.
So I can repost my results, but no one can see the raw numbers.
Results:
1 - 511 (5.11%)
2 - 494 (4.94%)
3 - 502 (5.02%)
4 - 539 (5.39%)
5 - 512 (5.12%)
6 - 502 (5.02%)
7 - 491 (4.91%)
8 - 509 (5.09%)
9 - 528 (5.28%)
10 - 486 (4.86%)
11 - 490 (4.9%)
12 - 489 (4.89%)
13 - 475 (4.75%)
14 - 494 (4.94%)
15 - 467 (4.67%)
16 - 478 (4.78%)
17 - 526 (5.26%)
18 - 540 (5.4%)
19 - 464 (4.64%)
20 - 503 (5.03%)
Sorted Results:
18 - 540 (5.4%)
4 - 539 (5.39%)
9 - 528 (5.28%)
17 - 526 (5.26%)
5 - 512 (5.12%)
1 - 511 (5.11%)
8 - 509 (5.09%)
20 - 503 (5.03%)
3 - 502 (5.02%)
6 - 502 (5.02%)
2 - 494 (4.94%)
14 - 494 (4.94%)
7 - 491 (4.91%)
11 - 490 (4.9%)
12 - 489 (4.89%)
10 - 486 (4.86%)
16 - 478 (4.78%)
13 - 475 (4.75%)
15 - 467 (4.67%)
19 - 464 (4.64%)
CHITEST = 0.521369785
No Bobson, stop - you're using up all the random numbers! ;-)
When we run out, does that mean that only the non-random ones will be left?
It is almost two months that i don't get a good roll....ok maybe it depends on the fact that all the games i was in died and playing only in one now.
Think i'll try to roll some d20 here to check.
Some good rolls,but i don't need them here and now :(
Add this to the list, goes with what the OP said
|
RPG Superstar 2009 Top 32
|
While usually, I cry foul very quickly, unless Paizo decided to implement a RNG from scratch, the default RNGs provided with all common programming languages are strong enough for pretty much any non-security related matter - and certainly for making up a few random numbers between 1 and 100 ;) However, just like my "lucky dice", we always interpret data according to our own lens.
(Note to self: Do not talk about my "really lucky dice" which have a statistically significant bias towards rolling 5)
| 1 person marked this as a favorite. |
... I do find the idea of providing personal digital dice bags that you can select from for rolling particular things amusing enough to think about how to implement next time I'm stuck in traffic or in the shower.
I don't even want to know why or how you get stuck in the shower... ;)
The given Chi-Squared value needs to be compared against a Chi-Squared distribution with k-1 degrees of freedom. In this case k = 20 as you divided the data into 20 intervals. The Chi-squared value that you compare it to for 95% confidence is 30.144 (Pulled from a Chi-Squared table). Because your given Chi-Squared value is less that 30.144 we can be 95% confident that the data was generated using a random number generator that uniformally disperses integers between 1 and 20 i.e. a dice.
So the results appear to be valid.
Thanks!
