Yeah that 1000 rolls is way more conclusive.

Those dice are biased with 99% confidence.

Now pls calculate the force with which dice are rolled and then think about the weight difference on one of the sides with 0.0001 gramm less.

When you roll a die, it doesn't just flip once and slam down on a result.

It rolls many times until the force dissipates to a level so low that it doesn't have enough to flip again.

At that stage, or where a die is on the brink of flipping; an air bubble does have a big effect.

You're not comparing the force of your arm to the air bubble. You're looking at the effect of major inclusions on a die that is on the brink of flipping on its final roll. If it's weighted away from blanks, you're not getting a blank.

Even if the force goes down through bumbing on the table and in the situation you are describing the orientation of the die towards the surface is very random because of those bumps/jumps of the die before the force dissipates. So while you are correct that the force gets less and less i still doubt that little off balance at the end of a roll will do much about the overall roll.

Of course i didnt calculate it. But the effect you describe is still very slim if you look at how many possible ways the die could have turned before that even coming into effect. Also i think it will effect only a tiny amount of rolls that if you take away everything else, like friction and all the other forces, exactly the amount of off weight will influence the die tipping over. Its still only one effect of many.

Take weighted dice to Vegas, and then use your explanation to let them allow you to use them.

If the weight only effects 1 dice per round or every 2 rounds or every 3 rounds that is huge, that could be as strong as palpatine.

They are making money of of it so yes every little bit off statistics count. We are playing a game. and i still highly doublt it would be even one die each game.

Excellent we have your highly doubt to counter the OP's experience of his dice rolling extremely hot every game for several months.

I just state my opinion about how physics work out here. Everyone can read it or ignore it.

His observation has nothing to do with the facts of if this minor difference in dice actually has any effect. You can believe it or and you can think about what i wrote or let it be. Saying that his observations even if its over several month is the better truth just isnt correct.

What have you done to verify your theory on his dice?

I believe taking his actual experience over your theory is indeed correct.

While i was posting this he showed actual data showing the dice are indeed hot and would indeed effect the game.

I would bet with you that if 100 random ppl would also note their dice results over the next 500 rolls that his actual results he posted above are indeed above avarage but a lot of ppl wuld have similar results while most would be closer to the average and several would be as much lower as he is above the average.

Sry its not a prove of his theory. of course i also dont prove my theory because i cant prove it. I neither have the equipment nor the kowledge of the exact physical calculation. Thats why i said my theory is about my knowledge of physics, i didnt even claim to prove it.

yes that is how variance works... thanks?

The point is his dice are rolling above average, they should not be used. why are you against this?

Your first part of the sentence explain why your second part still might be wrong. His stastistics could change back to avarage ober another 2000 rolls. Thats how variance works. Only because the first 500 oder 100 attemps are above avarage it doesnt mean that the next 500 will also be above average. Of course if he keeps writing it down its getting more and more clear if the dice roll consitently above average.

But what is the consequence of this? Its still no prove that all the dice of the same manufacturer roll above average. Its also not a prove that the original dice roll average.

Do you really want everyone taking part in a tournament roll their dice 500 times and all above average have to throw away their dice?

Yeah that 1000 rolls is way more conclusive.

Those dice are biased with 99% confidence.

But we have CaineHoA feeling that everyhign is going to be fine.

Yeah that 1000 rolls is way more conclusive.

Those dice are biased with 99% confidence.

But we have CaineHoA feeling that everyhign is going to be fine.

lol

Here are the results from the first 504 rolls (yes, I'm compulsive):

Observed: 204 evades, 137 focus, 163 blanks

Expected: 189 evades, 126 focus, 189 blanks

Chi-Squared Test: this page for reference ( http://www2.lv.psu.edu/jxm57/irp/chisquar.html )

(204-189)^2 / 189 + (137-126)^2 / 126 + (163-189) ^ / 2 = 1.190 + 0.960 + 3.577 = 5.727 so our chi^2 value = 5.727. Generally people use the p < 0.05 value for significance. Our degrees of freedom is 2 (3 possibilities - 1). So using that Table B.2 chart we are between p < 0.10 and p < 0.05. So we're close to it being significant, but we accept that it's in the 'range of acceptable deviation'. There's somewhere between a 10-5% chance of this happening 'by chance alone'. If you roll a few hundred more, it may shift a little.

Conclusion: its close to being able to be rejected as non-fair, but isn't given our picked 5% chance basis for rejecting.

Edit: The 1002 rolls, I get a chi^2 value of 9.85 -- so that puts it p < 0.01 so we can reject that as a fair die.

Apparently Icelom has suffer multiple beat downs due to these grossly (sarcasm) unfair and biased dice.

Yeah that 1000 rolls is way more conclusive.

Those dice are biased with 99% confidence.

But we have CaineHoA feeling that everyhign is going to be fine.

lol

Thanks for the good natured laugh, it was actually an interesting debate with you.

Here are the results from the first 504 rolls (yes, I'm compulsive):

Observed: 204 evades, 137 focus, 163 blanks

Expected: 189 evades, 126 focus, 189 blanks

Chi-Squared Test: this page for reference ( http://www2.lv.psu.edu/jxm57/irp/chisquar.html )

(204-189)^2 / 189 + (137-126)^2 / 126 + (163-189) ^ / 2 = 1.190 + 0.960 + 3.577 = 5.727 so our chi^2 value = 5.727. Generally people use the p < 0.05 value for significance. Our degrees of freedom is 2 (3 possibilities - 1). So using that Table B.2 chart we are between p < 0.10 and p < 0.05. So we're close to it being significant, but we accept that it's in the 'range of acceptable deviation'. There's somewhere between a 10-5% chance of this happening 'by chance alone'. If you roll a few hundred more, it may shift a little.

Conclusion: its close to being able to be rejected as non-fair, but isn't given our picked 5% chance basis for rejecting.

Yep, but try with the 1004 results.

Here are the results after 1002 rolls (batches of 6):

Observed: 420 evades, 248 focus, 334 blanks

Expected: 376 evades, 250 focus, 376 evades

Fisher exact probability test (2x3):

p=0.085009

Chi-square = 4.93

If we're strictly looking at blanks:

Observed: 668 evades/focus, 334 blanks

Expected: 626 evades/focus, 376 blanks

p=0.0555

When you say batches of 6 -- are you rolling all of the dice and counting faces?

Here are the results after 1002 rolls (batches of 6):

Observed: 420 evades, 248 focus, 334 blanks

Expected: 376 evades, 250 focus, 376 evades

Fisher exact probability test (2x3):

p=0.085009

Chi-square = 4.93

If we're strictly looking at blanks:

Observed: 668 evades/focus, 334 blanks

Expected: 626 evades/focus, 376 blanks

p=0.0555

When you say batches of 6 -- are you rolling all of the dice and counting faces?

Correct

Here are the results from the first 504 rolls (yes, I'm compulsive):

Observed: 204 evades, 137 focus, 163 blanks

Expected: 189 evades, 126 focus, 189 blanks

Chi-Squared Test: this page for reference ( http://www2.lv.psu.edu/jxm57/irp/chisquar.html )

(204-189)^2 / 189 + (137-126)^2 / 126 + (163-189) ^ / 2 = 1.190 + 0.960 + 3.577 = 5.727 so our chi^2 value = 5.727. Generally people use the p < 0.05 value for significance. Our degrees of freedom is 2 (3 possibilities - 1). So using that Table B.2 chart we are between p < 0.10 and p < 0.05. So we're close to it being significant, but we accept that it's in the 'range of acceptable deviation'. There's somewhere between a 10-5% chance of this happening 'by chance alone'. If you roll a few hundred more, it may shift a little.

Conclusion: its close to being able to be rejected as non-fair, but isn't given our picked 5% chance basis for rejecting.

This is what I was thinking. The dice, at this sample size, appears to be biased - though not grossly enough to be called totally unfair and the deviation could very well even out. I think I'd have to see more (like 2-3 times more) data before I'd call them totally unfair.

Not that I have a horse in this race, nor particularly care about the outcome. But the same case could be made for official FFG dice, with similar/the same outcome, and be totally legal (in all events). And, while I have no reason to doubt the OP, we are all only taking him at his word on the anomaly and the outcome of the test rolls (though I can think of no good reason (other than trolling) why we would be deceived.

I guess I don't get all the hubbub and why we are on 2+ pages of this discussion LOL.

Here are the results after 1002 rolls (batches of 6):

Observed: 420 evades, 248 focus, 334 blanks

Expected: 376 evades, 250 focus, 376 evades

Fisher exact probability test (2x3):

p=0.085009

Chi-square = 4.93

If we're strictly looking at blanks:

Observed: 668 evades/focus, 334 blanks

Expected: 626 evades/focus, 376 blanks

p=0.0555

When you say batches of 6 -- are you rolling all of the dice and counting faces?

Correct

Unfortunately this confounds the problem a bit. It's possible that you have a single (or two) bad dice in there and it's not the whole set that's bad. All we can say is that there is at least one unfair die in there.

Here are the results after 1002 rolls (batches of 6):

Observed: 420 evades, 248 focus, 334 blanks

Expected: 376 evades, 250 focus, 376 evades

Fisher exact probability test (2x3):

p=0.085009

Chi-square = 4.93

If we're strictly looking at blanks:

Observed: 668 evades/focus, 334 blanks

Expected: 626 evades/focus, 376 blanks

p=0.0555

When you say batches of 6 -- are you rolling all of the dice and counting faces?

Correct

Unfortunately this confounds the problem a bit. It's possible that you have a single (or two) bad dice in there and it's not the whole set that's bad. All we can say is that there is at least one unfair die in there.

You could split them in half and see if one pile starts rolling better?

I think you would need to roll and account for individual dice rolls independently from one another to get an accurate sample set.

Also, are you rolling each die in the same fashion, on the same surface, with the same amount of force, etc.....? To truly get a correct and unbiased sample set of data you would need to automate the process so that each roll is replicated as exactly as is possible every time. Just taking a handful and tossing them onto your play mat is not an accurate and scientific (which seems to be the contended backing for accuracy of the original hypothesis) way of collecting your data.

I can only agree to thatdave. Its what i wouldve said next as well.

Its totally possible that his way of rolling the dice is influencing the roll more than a bubble in any of them.

I think you would need to roll and account for individual dice rolls independently from one another to get an accurate sample set.

Also, are you rolling each die in the same fashion, on the same surface, with the same amount of force, etc.....? To truly get a correct and unbiased sample set of data you would need to automate the process so that each roll is replicated as exactly as is possible every time. Just taking a handful and tossing them onto your play mat is not an accurate and scientific (which seems to be the contended backing for accuracy of the original hypothesis) way of collecting your data.

Though it does say the way he rolls, and the dice pool he throws isn't producing a fair result. But yeah, technically with a chi-squared test you'd want to test one die at a time. Imagine if you had 3 bad dice, each of which rolls more evades / focuses and blanks, and place them all in the same pool. They could potentially produce a distribution that falls within the accepted deviation by canceling each other out. Pick one of those out individually however, and it would produce skewed results.

I think you would need to roll and account for individual dice rolls independently from one another to get an accurate sample set.

Also, are you rolling each die in the same fashion, on the same surface, with the same amount of force, etc.....? To truly get a correct and unbiased sample set of data you would need to automate the process so that each roll is replicated as exactly as is possible every time. Just taking a handful and tossing them onto your play mat is not an accurate and scientific (which seems to be the contended backing for accuracy of the original hypothesis) way of collecting your data.

Though it does say the way he rolls, and the dice pool he throws isn't producing a fair result. But yeah, technically with a chi-squared test you'd want to test one die at a time. Imagine if you had 3 bad dice, each of which rolls more evades / focuses and blanks, and place them all in the same pool. They could potentially produce a distribution that falls within the accepted deviation by canceling each other out. Pick one of those out individually however, and it would produce skewed results.

I used the same surface each roll. I tried to use the same amount of force when rolling, but it's definitely not 100% consistent.

I would contend that the batching of 6 dice is a problem, though. The lack of inclusions and mass production should imply that they are identical. If they did have inclusions, then yes I completely agree that the dice can't be batched.

There is also no way I'm rolling a single die 1000 times. Especially since the skeptics will merely point out that yes, one of my six 3rd party dice might be weighted and that I should repeat for the rest of them....

I don't have the results handy at the moment, but I ran a similar test with the sagaborn dice. I used to jokingly call them my "cheater dice", until I found out after 100 rolls that the blank sides rolled 10% less, and that 10% was distributed between hits/crits. I labeled each side, rolled it 1000 times through a dice tower with an evade, FFG attack and FFG evade. The other 3 were pretty close to expected values per face over 1000 rolls.

I quit using the sagaborn dice. I liked them because they were 19mm, not 16mm, and perfectly clear! I haven't run the Nerd-X dice through this yet, but I think I may now after seeing this.

With 3 people, it takes about 3 hours to roll 1000 times. Its not too bad if you want a definitive answer! Or find the guy that built the dice roller ages ago...

Somebody do this test with the core set dice!! Batch of 6.

Although I'd say that there is considerable variance seemingly in core set dice. Some peoples seem to be more hot.

I think you would need to roll and account for individual dice rolls independently from one another to get an accurate sample set.

Also, are you rolling each die in the same fashion, on the same surface, with the same amount of force, etc.....? To truly get a correct and unbiased sample set of data you would need to automate the process so that each roll is replicated as exactly as is possible every time. Just taking a handful and tossing them onto your play mat is not an accurate and scientific (which seems to be the contended backing for accuracy of the original hypothesis) way of collecting your data.

Though it does say the way he rolls, and the dice pool he throws isn't producing a fair result. But yeah, technically with a chi-squared test you'd want to test one die at a time. Imagine if you had 3 bad dice, each of which rolls more evades / focuses and blanks, and place them all in the same pool. They could potentially produce a distribution that falls within the accepted deviation by canceling each other out. Pick one of those out individually however, and it would produce skewed results.

I used the same surface each roll. I tried to use the same amount of force when rolling, but it's definitely not 100% consistent.

I would contend that the batching of 6 dice is a problem, though. The lack of inclusions and mass production should imply that they are identical. If they did have inclusions, then yes I completely agree that the dice can't be batched.

There is also no way I'm rolling a single die 1000 times. Especially since the skeptics will merely point out that yes, one of my six 3rd party dice might be weighted and that I should repeat for the rest of them....

It's not just an inclusion issue, mis-shapes can affect the dice as well. I did a chi-test on one of my d20s after doing a salt water test. It showed the 20 coming up more often than not in the salt water. However when I did the chi-test it ended up being the 14 or 17 that was being rolled less often -- the 20 was 'normal'. So it's hard to say what they'd do.

Either way, your 6-dice pool seems unfair

I'd like to see the data on the sagaborn dice compared to FFG dice (and others apparently)

Once, I isolated 'hot' dice (4 defense dice that roll way too many evades ) from the core sets- and it really helped my game.

Then I mixed 'em back up into my dice-bag so I'm not a d-bag... and I only try to isolate them during a match trying to guess which ones are hot. If a match goes on and on for many rounds of defending I can usually find the hot ones.

Yeah that 1000 rolls is way more conclusive.

Those dice are biased with 99% confidence.

I brute-forced the probability distribution at 504 rolls and 1002 rolls, looking at only the evades (counting blanks and focus as a single result).

With 504 rolls, the probability of fair dice rolling at least 204 evades is 9.15%.

With 1002 rolls, the probability of fair dice rolling at least 420 evades is 0.2274%.

This says essentially the same thing as the Chi-Squared test, but stated this way may be slightly more intuitive to understand.

I knew a real math guy would chime in at some point!

Again, I'll point out the caveat that these results are not truly accurate due to the inherent inconsistencies in the data generation/gathering. That's not to say that the dice aren't fair/unfair, only that flawed data is a surefire way to get flawed results (my primary expertise is chemistry, not math).

Either way, good on the OP for ditching dice that were at the very least perceived as 'unfair'. I'd also like to add (again) that similar results can be achieved with the FFG dice (as well as most other dice which are not engineered and/or tested for precision). My third party dice are no better or worse than my FFG dice.

Here are the results after 1002 rolls (batches of 6):

Observed: 420 evades, 248 focus, 334 blanks

Expected: 376 evades, 250 focus, 376 evades

Chi-statistic = 9.87

p-value = 0.00719

If we're strictly looking at blanks:

Observed: 668 evades/focus, 334 blanks

Expected: 626 evades/focus, 376 blanks

Chi-statistic = 7.42

p-value = 0.00645

edit: corrected Chi-square results