Ask for 3d6

And I refer you back to my "without a proper understanding of statistics"
You don't need total certainty. You can gather evidence and make an informed decision and be confident that you are correct, or at least correct enough to challenge the GM.

I was taking issue with the disregard and ignorance of statistical processes. Actually tracking rolls is a pointless exercise in the context of a beer and pretzels game.

I'm going to assume that they're trolling, but to the user who insists that it's impossible to be sure that the die is just a d20 being rerolled on 1, 2, 19, or 20...

You can calculate the exact probability that you will get more than n 3s and 18s in m rolls. On a d20 with retools, you would expect those 12.5 times out of 100 rolls. On a 3d6 you would expect a little under 1 time out of 100 rolls.

Keep track of your GM's rolls for 100 cases, noting whenever he gets a 3 or an 18. If it is actually a 3d6 then he has about a 1% chance of rolling more than 3 of those. This means that you can be 99% sure that he is not rolling 3d6.

But if he is rolling a 1d20 (with rerolls) then he has about a 1% chance of rolling 4 or less of those.

Of course, this is not fool proof, but it will give you 99% confidence.

>I'm talking about total certainty -- which is impossible.

You can't even have total certainty that you're not a brain in a jar, so fucking what.

9

I actually took statuses classes in college.

I never really "got" any of the higher level basic stuff and despite my best efforts and perusing multiple different ways of explanation didn't retain much of it after the class was over and I'd gotten a high enough grade to pass. Like on some level I knew "this works and it is logical" but could never really grasp that logic, only a vague outline of it that was. It sufficiently detailed to constitute a "working knowledge".

Any recommendations on how to cope with this?

I had a 1% chance to be born with autism and it still happened so I'm not accepting 99% certainty on anything pal

Don't worry, you can get whatever chance you like, just keep going

Odds for an 18 on an 'honest' 3d6 is just under 1 in 200. Seeing them very often from an RNG means something is likely wrong.

The real giveaway would be consistently generating outliter results. On honest 3d6 you get 9, 10, 11 or 12 as a result 48.2% of the time, so even a relatively small sample should skew heavily to the middle of the bell curve. If you see a bunch of 3, 4, 5, 16, 17, 18 then it's going to be a pretty good hint.

you can not understand statistic without a understanding of higher mathematics. You can know how to use them, what the end result mean, but you will not understand the proceses and why they happen

Just ask him flat out. He'll probably say that it doesn't matter / is easier than rolling 3d6. Then tell him about the probability curve.

Yoou Dm is like a player i booted from my for refusing to roll the dice i asked him to (Ex; 2d10 for d20 because he lacked one and REFUSED to use one of mine, he did shit like this all the fucking time) I basically kicked him out after he rolled 2 ones on the 2 d10s and i said 'guess you got a nat 1, sucks to be your wizard' and he argued why he didn't get a nat one, because it was a 2, and that because he can't roll a one he can't crit fail.

I told him unless he starts to use a d fucking 20, he can leave and i'm keeping the pizza.

He took offence to that.

I said that the odds of a 3 OR an 18 is just under 1%

Then you can track a larger number of rolls. Use Poisson distributions
en.wikipedia.org/wiki/Poisson_distribution
If your GM rolled n 3s and 18s then calculate the odds of rolling 0, the odds of rolling 1, 2, and so on until you reach n.

>without a understanding of higher mathematics
Multiplication is higher mathematics now?

assuming d&d, you don't really roll 3d6 often. 6 times at creation and then that's it. 6 times is not really meaningful for anything
of course, you could be playing something else, in which case never mind me.

not, and that's using it, there is a big diference between being able to tell what you got from a t-student test, and being able to tell how you got it, and how the whole thing works

Dude, you're literally multiplying the number of possibilities to the power of the number of times its invoked. That's primary school shit.

For someone who thinks total certainty is impossible you sure like talking in absolutes.

>I had a 1% chance to be born with autism
No, maybe humans in general do, but you did not have a 1% random chance to mutate into an autist upon birth.

Those six rolls influence your character for the rest of the game

You know what he fucking meant so get your ass back to your triggerbitch subreddit where your pedantry gets you into the secret clubhouse.

You're retarded, but I'll bite anyways.
Chances of rolling a value assuming re-rolls on 1,2, 19, 20:
>3: 6.25%
>4: 6.25%
>5: 6.25%
>6: 6.25%
>7: 6.25%
>8: 6.25%
>9: 6.25%
>10: 6.25%
>11: 6.25%
>12: 6.25%
>13: 6.25%
>14: 6.25%
>15: 6.25%
>16: 6.25%
>17: 6.25%
>18: 6.25%

Chances of rolling on 3d6:
>3: 0.46%
>4: 1.39%
>5: 2.78%
>6: 4.63%
>7: 6.94%
>8: 9.72%
>9: 11.57%
>10: 12.50%
>11: 12.50%
>12: 11.57%
>13: 9.72%
>14: 6.94%
>15: 4.63%
>16: 2.78%
>17: 1.39%
>18: 0.46%

Notice anything?
You should pointbuy or standard array anyways.

kinda missed the point there
you're never going to get a meaningful sample size from 6 rolls, is the point m8

Bullshit, the proof of t-test working uses integrals and even those integrals use substitution that comes from numerical solution. I mean it is some time since I took statistics, but this i remember,

how about instead of calling people retarded so you can have some smug sense of superiority, you try to look at where he might be misunderstanding and try to be constructive? Find the place that they are making the mistake at- from what I see so far it seems that they're misunderstanding fundamental about statistics- start there, try helping instead of just being mean.

>100 rolls
Actually, I think you need considerably less than 100 rolls to still have significant discrimination between the two distributions.

But I'm not entirely sure how I should use the maths to properly show this. At the very least finding which of two known distribution a sample belongs to should be easier than figuring out an unknown distribution from the sample.

>not just using 6 numeroted piece of paper that you put in a box and draw
God damn kids and all their electronic gizmo.

Wow. Why not 10d2?

>Why not 10d2?
Not that user, but pic related

fuck off autist

This is the correct answer. We can stop fighting now.

>This is the correct answer.
Of course it is.

>We can stop fighting now.
Of course we can.

>Will we?
Of course not.

>Veeky Forums should turn into math class
They don't have any obligation to offer tutoring sessions to idiots on Veeky Forums

I'd assume he probably isn't given enough result where he knows the GM is actually rolling the "die" to actually check. If you only get a die check on it like 8 times, you really don't have enough information.

A new group I'm in use what I consider an odd method of rolling 1d100
So they use a percentile dice (00, 10 -> 90) and a 1d10. Now in my experience, if you roll 00 and 0, that's 100. But they count that as 10. To get 100, you'd have to roll 90 and 0, which I'd usually count as 90.

Does this effect dice probability any, or is it just a weird way of counting dice?

Doesn't affect probability. The numbers are coming up all the same, they're just reading it differently, the read 0 on the d10 as 10, and add it to the % die accordingly.

This is what I would do. Can someone tell me if it works?

On 1d16+2 (which is equivalent to what your DM is rolling), the odds of 9-12 three times in a row is 1.6%. It just doesn't happen all that much.

Meanwhile, on 3d6, the likelihood of rolling a 9-12 thrice in a row is ~11.16%.

Keep track of every roll made with this digital dice. At the end, circle any runs of 9-12. Count the entire set of rolls and subtract two to get the number of possible sets of 3. Then compare to the number of circles. If the number of circles is substantially below 10% of your other number, you can be pretty damn sure your theory is correct. If not, you can be equally certain you're imagining things.

This is assuming you have a pretty big number of rolls to work with, though.

No, in either case there's only one way (out of 100) to get the specific result, so the probability remains 1/100.

Their way may seem odd, but they're just counting 0 on the single number die as 10, and then straight up adding the two together, instead of counting it as zero and adding for all cases but 00-0.

>GM asks me to pick between three face-down cards
>two are black, if I pick the red one I win a boon
>after I pick, he turns one of the other cards up to reveal it's black
>asks if I want to switch to the other face down card

Do not fight it. Think how you can utilize it for your own good. If system uses randomness with different distribution than originally intended balance of power shifts considerably.

I literally used this one yesterday
>Friday at the office
>asked to personally review some randomly chosen entries of automated weekly audit
>randomly choose first 5
>manager goes "hey, that's not random choice"
>Dilbert.jpg
>manager accepts my point
>but I still had to do couple more reviews

lern 2 posterior probability

Does anyone know how to generate a curve on anydice for these stat generation rules "roll 4d6, drop the lowest roll, reroll 1's"
I've tried playing with the "highest 3 of 4d6" function, but that started breaking when I tried to go from d6 to d5+1

Switch. You have a 66.66% chance of being correct opposed to 33.33% if you keep your card.

depends

did he know the card he overturned would be black?

Can someone please explain this shit to me.

The card doesn't change. The fact that one of the choices was wrong doesn't change the actual position of the other cards.

So why would switching be better? Nothing has changed but the elimination of one wrong option. Is this theoretical bullshit?

It's called the Monty Hall problem, and it's one that even trips up mathematicians sometimes. The "trick" is that the guy who opens the door (flips the card in this instance) knows where the right one is, and is altering your chances through his actions.

montyhallproblem.com/

It only works if the DM knew the answer in advance - I know that sounds like even more bullshit, but it means that the DM is giving you a clue you didn't have when you made your first choice.

I roll 20d1s because those are the only dice I have.

Yes, of course.

It might be easier to understand if the example is more extreme.
I lay down ten cards, you pick one
I flip up eight black cards
The position of the two remaining cards has not changed
How confident are you in your initial pick?

The Monty Hall problem comes down to "Either you guessed right or the person in charge told you the right answer." If you had a less than 50% chance of the first one, you should switch.

Here's one, Veeky Forums

Your DM lays down 5 cards and asks you to choose one. You know 3 are red and 2 are black. A red card is a boon, a black card a curse.

After you choose, the DM flips up a different, red card and asks if you'd like to switch to another of the remaining cards. Do you?

No.

No. That's stupid. Unless face-up cards count as technically remaining, in which case, fuck yeah, I pick the one I KNOW is a boon.

>mfw user puts his finger on the face-up red card and says "you didn't remove it so it's remaining."

>rolling d20 for statline instead of 3d6

Why have I never though of this.

Rolled 4, 18, 8, 13, 14, 4 = 61 (6d20)

I remember it used in rolling threads here on Veeky Forums.

The red is still on the table. Switch to it.

>he doesn't roll 1d16+2

>He doesn't roll 9d2/2+3d3

>He doesn't discard fixed dice sizes altogether and roll (skill)d(ability score) and rescale all the numbers in the game to fit

Why wouldn't you assume he was talking in practical manners, i.e. taking away plausible deniability from his DM rather than theoretical ones?
It's a bit absurd to respond to "I have no way to prove that" with "don't you know ANYTHING about sample sizes and certainty?"

In that scenario randomness is needed because you want a distribution representative of the group as a whole, not just a subset. If there are any trends across the set you'll miss tonnes of data because your distribution is, for all mathematical purposes, non-random.

You may just be the greatest mathamatician of our time.

>rocks fall

>3d6-based system
>GM buy nontransitive dice and insists it's okay because they have the same average result

>no way to prove it due to the nature of numbers being random
Why do you say things like that. Don't you realize how much it hurts me?

In other words, . Just generate an arbitrary large amount (~100 is probably enough, I don't feel like running the numbers) and you'll see that the ends are overrepresented, do yourself a nice proportion-based Z-test and you have statistical proof it's not how it should be.

That's not an average, that's long-run outcomes. Given an arbitrarily large number of repetitions, you'll see the distribution getting very close to one or the other.

>"Highly unlikely" happens more often than you'd think
Considering I can quantify the probability of such things, I think I have a very good idea of how often it would occur

Go home, Monty, you're drunk

desu I'd hit someone for that.

Found the sith lord

That's just stupid

>You should pointbuy or standard array anyways.
DM rolls *dice*
That is the total amount of stat points you have available to distribute with no lower than 8 and no higher than 18.

>no lower than 8
Why?

Isn't this just the Monty Hall problem?
Because if it is then switching is optimal.

Reread more carefully and pay attention to how many cards there are and which one the GM flips.

Right, I misread.
Guess it would have fooled me.

Look at it this way: You have a 2/3 chance of guessing wrong initially. If you guess wrong and switch you win. Thus, if you switch, you have a 2/3 chance of winning.

What if the DM doesn't know which card is which and he randomly turns over one and it turns out to be the black one?

The logic still applies. If he randomly turned over the red card, then the "If you guess wrong and switch you win" part wouldn't hold.

honestly, i just prefer 0 to 99 for percentiles. i hate the idea that rolling the ten and getting a 0 becomes 90% great, but 10% fucking devastating.

So wait. If you switch, you theoretically improve your odds, but the DM knows where the card is. So couldn't he provoke you to switch away from a winning card using this information?

Not really. I mean he could try, but he can't change the fact that as long as he's flipping a bad card, your odds go up if you switch. If you know he only offers the switch if you've got the card, then you could do something with that.

Staying gives you 40% boon switching gives you 20% boon.

Wait that's 3 black 2 red i'm dumb.

So staying gives you 60%, switching (not to the exposed red card) gives you
>(3/5)(1/3) + (2/5)(2/3) = 3/15 + 4/15 = 7/15 = 46.666...%
Right?

t. not mathematician

You want the real reason, or the fluffed reason?

The fluffed reason is because they feel like being forced to play a character with a modifier for any stat of less than -1 is unnecessarily damaging to their enjoyment of the game.

The real reason is they don't want to have to deal with having to figure out how to roleplay someone that's actually deficient in some way and want to feel powerful, because they see playing RPGs as nothing more than a power fantasy.

But you don't know if he is doing it Monty hall, and always flipping a card and offering a switch, or he's fucking with you because he knows.
This time you have to play the man, not the odds.

In that case if he's flipped a bad card, then switch. If he flips the good card, uhh, take it, I guess? If he flips a bad card he's increased your odds of getting a good one at least slightly by eliminating one wrong choice. The only reason to ignore that is if you have reason to believe he wouldn't flip a card unless you'd already picked the right one.

>control-F "hypothesis test", "variance", "standard deviation"
>no results

1. There's more to it than the average. You can still run a test for variance/standard deviation. Or even the probabilities of each integer in the array [3,18].

1a. The graphs for each one are going to look very different. In this case it should be pretty clear if you just plot out the probability distribution.

2. People like you guys are the reason why casinos can stay in business.

this is the best explanation for the Monty Hall problem I have ever read?