Dice Hate

[CubeB]

So, I have a friend (A Swordsage), who is getting really frustrated, because he can't hit. Literally, he cannot.

It's an e6 Eberron Game. The party consists of a Human Transmutation Specialist, a Shifter Ranger, a Human Warlock, a Warforged Incarnate/Totemist (Meeeee), some... weird homebrew Wordcaster thing no one understands, and the aforementioned Human Swordsage.

The issue is, for whatever reason, he can't seem to roll higher than a 3.

Seriously, he misses with touch attacks. He is practically dependent on our Hero Point mechanic. According to our dice bot, his lifetime roll average is 8.

He is becoming frustrated, depressed, and tempted to just let his character die because the dice just aren't falling in his favor. No one else has this problem.

Is there anything I can do to help him improve his chances of at least hitting?

[Arbane]

That is impressively terrible luck. As someone who has trouble getting double-digit numbers on a d20, I can sympathize.

Let him play a different character for a few sessions? If the Curse is on the character, he'll be fine. If not, he's screwed. Get him to play some system where low is good on rolls.

[eggynack]

8 doesn't seem too far off normal, as roll averages go. Anyway, assuming luck as some sort of actual factor, which this player is somehow bad at, the best solution is probably removing said luck as a factor. Best way to do that is have him reroll as a caster of some kind. Casters have tons of ways of interacting with the game without ever having to touch a die, and when they do rely on luck, it's often spread across enough foes that it averages out, and reliant on the opponent's rolls rather than your own. Could be worth doing, even if only to just get him away from rolling for awhile. The whole thing seems to be hurting his heart a bit.

[Grizzled Gryphon]

Either go with a diceless game, or a caster. Focused on Magic Missile, or something? there is a magic Missile specialist in the "Best of Dragon" book. Or "Best of Dungeon", perhaps? Don't recall the books name atm, but its in there. Might be just the thing for this poor soul.

[PolymeraseJones]

Sounds like maybe his dice roller is borked? Has he tried using a different roller, or maybe getting ahold of some physical d20s?

I remember I used to have a problem where my roller would only ever roll extremes - long stretches of rolling awfully followed by periods of rolling fantastically. Switching to physical dice fixed that pretty well.

[eggynack]

Either go with a diceless game, or a caster. Focused on Magic Missile, or something? there is a magic Missile specialist in the "Best of Dragon" book. Or "Best of Dungeon", perhaps? Don't recall the books name atm, but its in there. Might be just the thing for this poor soul.
You don't really need to use magic missile. All kindsa magic doesn't touch dice at all. Buff spells never really care about luck, at least not your own if the buffs are targeted elsewhere, and BFC's often just sit on the battlefield and screw with folks without a single roll. Utility and out of combat effects, from certain divinations to rope trick, also tend away from that sort of thing, and even if you do need to roll, that downside can be mitigated with repeated attempts quite often. Most importantly, those spell types I listed tend to be some of the best out there, much better than something like magic missile, so little is being sacrificed to bypass luck.

[Duke of Urrel]

I consider luck to be an after-the-fact abstraction. Attributing any motive or personality whatsoever to "Lady Luck," or whatever you call "her," is superstitious. Ditto considering bad luck to be a real "force" that clings magically to a die, or worse, to a person.

Here's the good news. Since "Lady Luck" doesn't actually exist, we are not obliged to be "loyal" to "her" in any way. In other words, if "she" hasn't been fair to us, then we are under no obligation to be "fair to her."

And therefore:

If a player has had phenomenal bad luck, so obvious that nobody can deny it, then the next time this player rolls a three or lower (again…), just let this player keep rolling until the die comes up 18 or higher. Turn about is fair play, eh? Or the next time the player rolls an 8 and it's a miss (and a 12 or higher would be a hit), then let the player roll again until a 12 or higher comes up.

Don't let this ad-hoc, emergency rule become a bad habit. It shouldn't be necessary to fudge die results more than once to break up a losing streak. Everybody should reach consensus (and consensus always includes the dungeon master) before this emergency rule is applied. And of course, the option should be available to everyone, not just to this player … and not just to players…

The only reason I could give not to cut this unlucky player a break would be if the monsters, NPCs, and other potentially ruinous effects have been similarly unlucky over the course of the game. Bad luck has a place, but it should be balanced by somebody else's bad luck. If it isn't, then I say, some good luck is called for, even if we need to engineer it to make it happen.

Fortuna is not fair. But human beings can be.

Of course, if you're playing with one or more rules absolutists for whom every rolled result is "sacrosanct," no matter what, then I can't help you.

[Greenish]

Either go with a diceless game, or a caster. Focused on Magic Missile, or something? there is a magic Missile specialist in the "Best of Dragon" book. Or "Best of Dungeon", perhaps? Don't recall the books name atm, but its in there. Might be just the thing for this poor soul.Force Missile Mage PrC, Dragon Compendium Volume I*. But see Eggynack, above.


*The one and only, as it happens.

[Fax Celestis]

8 doesn't seem too far off normal, as roll averages go. Anyway, assuming luck as some sort of actual factor, which this player is somehow bad at, the best solution is probably removing said luck as a factor. Best way to do that is have him reroll as a caster of some kind. Casters have tons of ways of interacting with the game without ever having to touch a die, and when they do rely on luck, it's often spread across enough foes that it averages out, and reliant on the opponent's rolls rather than your own. Could be worth doing, even if only to just get him away from rolling for awhile. The whole thing seems to be hurting his heart a bit.

It's more than a standard deviation of the expected average roll of 10.5.

[Illogictree]

You might want to check that he's been rolling the right dice. Seriously, check that he's not using a d10 or d12 rather than a d20.

If that's not the problem, replace the d20 he's using, see if the problem is with the die.

You might also want to consider letting him roll 2d10 or 3d6+2 rather than a d20, which ought to be more bell-curve like and tend toward the average, which would be better than what he's currently getting.

[eggynack]

It's more than a standard deviation of the expected average roll of 10.5.
Sure, it's definitely abnormal. It's just not the kinda thing where I'd cower away from the screen in stunned silence at this horrible luck, or question the accuracy of the results. It's abnormal, but it's a normal level of abnormal.

[137beth]

If you are using physical dice, it is not actually that uncommon for d20s to be poorly constructed and favor certain numbers. So in that case, have him switch dice.
Otherwise, if you are using computer-generated rolls, make sure he is actually rolling the correct size. Also, make sure that you are using a PRNG that you trust.

[JusticeZero]

Alternately, every other die roll, or even every die roll if needed, reverse it so 20=1, 19=2, etc. Or yeah, switch dice.

[CubeB]

I do think we need to check our dice bot, but the reroll and swapping ideas seem to work. I don't think he really wants to switch classes...

[Doctor Awkward]

Either go with a diceless game, or a caster.
This is the best solution to bad rolls of the dice, but the problem is this isn't the solution for everyone.
Not everyone can get into really immersive roleplaying, and not everyone is good at improvising interesting banter for hours on end.
The people that are will have great fun, the people that aren't will be bored and playing mobile phone games.


If you are using physical dice, it is not actually that uncommon for d20s to be poorly constructed and favor certain numbers. So in that case, have him switch dice.
This is the biggest danger in using physical dice.
Personally all of my favorite dice come from Koplow Games (http://www.koplowgames.com/). Unfortunately they only sell to vendors, so if you want a particular set you'll have to convince your local game shop to order it for you.
Chessex (http://www.chessex.com/) makes good dice as well, and like Kaplow they are (generally) reasonably priced.


Also, make sure that you are using a PRNG that you trust.
And this is the problem with using RNG's and dice roll programs.
Unless you are a Software Engineer, or actually study programming, here's something you probably don't know about computers:

They don't do random.

No, really. A computer is incapable of arriving at a randomly determined number by itself. If you use an RNG function by itself when coding, then the system will apply a mathematical formula to a static number and return an identical result every single time. You need a random number seed to apply to the formula. And if you want the result to change then the seed has to change.
The vast majority of random number seeds use the time showing on the system clock when it comes time to get a random number, because it's something every computer will have that is always changing.
What this means for dice rollers is that rolling at a certain time will always give you the same result, and certain times of day will generate better results than others.
You can increase the complexity of the seed, of course, by also checking the day of the week, or the day of the month, or the month, or even the year, and each iteration will give a closer approximation of true randomness (and the best RNG's do this). But in the end an approximation is all it will ever be, and it's for this reason I prefer physical dice.

[Thanatosia]

Do you use physical D20 or is this Dicebot thing doing your rolling for you? HOw many rolls do you have recorded in your dicebot? Cuz rolling an average of 8 after like 10 rolls is a very different beast then rocking an 8 average after 150 rolls.

Another good emergency measure to someone being 'dice screwed' noticablly worse then average runs of bad luck is to invert all his dice rolls for a bit. All his rolls count as if he rolled 21-(physical Roll): 20s become 1s and 1s become 20s. If bad luck consistantly persists after switching back and forth between dice inversion a bit, and you've rulled out perception bias, then the universe is out to get you, and make your peace with god.

[Fax Celestis]

This is the best solution to bad rolls of the dice, but the problem is this isn't the solution for everyone.
Not everyone can get into really immersive roleplaying, and not everyone is good at improvising interesting banter for hours on end.
The people that are will have great fun, the people that aren't will be bored and playing mobile phone games.


This is the biggest danger in using physical dice.
Personally all of my favorite dice come from Koplow Games (http://www.koplowgames.com/). Unfortunately they only sell to vendors, so if you want a particular set you'll have to convince your local game shop to order it for you.
Chessex (http://www.chessex.com/) makes good dice as well, and like Kaplow they are (generally) reasonably priced.


And this is the problem with using RNG's and dice roll programs.
Unless you are a Software Engineer, or actually study programming, here's something you probably don't know about computers:

They don't do random.

No, really. A computer is incapable of arriving at a randomly determined number by itself. If you use an RNG function by itself when coding, then the system will apply a mathematical formula to a static number and return an identical result every single time. You need a random number seed to apply to the formula. And if you want the result to change then the seed has to change.
The vast majority of random number seeds use the time showing on the system clock when it comes time to get a random number, because it's something every computer will have that is always changing.
What this means for dice rollers is that rolling at a certain time will always give you the same result, and certain times of day will generate better results than others.
You can increase the complexity of the seed, of course, by also checking the day of the week, or the day of the month, or the month, or even the year, and each iteration will give a closer approximation of true randomness (and the best RNG's do this). But in the end an approximation is all it will ever be, and it's for this reason I prefer physical dice.

Well, random.org uses atmospheric noise as its seed. That is as close to truly random as we can digitally get.

[With a box]

How about useing pre-written one time number table?



Well, random.org uses atmospheric noise as its seed. That is as close to truly random as we can digitally get.

Or use radioactive random number generator, law of physics are assure its randomness

[jaydubs]

I'd second (well, third or fourth) the inversion mechanic. You can give him the option to pick either option beforehand without it statistically affecting anything. If it's in person, give him a playing card. When he's keeping it face down, the dice roll is what it is. While he'e keeping it face up, he gets the inversion.

[CubeB]

Do you use physical D20 or is this Dicebot thing doing your rolling for you? HOw many rolls do you have recorded in your dicebot? Cuz rolling an average of 8 after like 10 rolls is a very different beast then rocking an 8 average after 150 rolls.

Another good emergency measure to someone being 'dice screwed' noticablly worse then average runs of bad luck is to invert all his dice rolls for a bit. All his rolls count as if he rolled 21-(physical Roll): 20s become 1s and 1s become 20s. If bad luck consistantly persists after switching back and forth between dice inversion a bit, and you've rulled out perception bias, then the universe is out to get you, and make your peace with god.

It is a dice bot that we use, because it's IRC based. There might be something fishy with how it's set up. As far as the person's history, he has been playing this character for a little over a dozen gaming sessions.

I'll suggest maybe inverting the dice for a while (and seeing if it's perception bias), or getting someone to check the seed.

I've also posted the actual party averages here, for science purposes. Note that Cobalt and Ara are newer additions, (three sessions old as opposed to 12+)


Bob (Swordsage): 8.79
Cobalt (Incarnate): 12.32
Ara (Wizard): 11
Mary (Warlock): 11.91
Skunk (Ranger): 9.96
Vanel-Xo (Homebrew Thingy): 11.82

[Fax Celestis]

Can you post the code for the bot? Use pastebin or something. Don't even need the whole bot, just the die roller.

[CubeB]

Can you post the code for the bot? Use pastebin or something. Don't even need the whole bot, just the die roller.

I'll try and get it for research purposes, though the person who runs the bot is currently asleep like a normal person. I think it might be date/time based though. I'll get back to you when I can.

[NNescio]

8 doesn't seem too far off normal, as roll averages go. Anyway, assuming luck as some sort of actual factor, which this player is somehow bad at, the best solution is probably removing said luck as a factor. Best way to do that is have him reroll as a caster of some kind. Casters have tons of ways of interacting with the game without ever having to touch a die, and when they do rely on luck, it's often spread across enough foes that it averages out, and reliant on the opponent's rolls rather than your own. Could be worth doing, even if only to just get him away from rolling for awhile. The whole thing seems to be hurting his heart a bit.


It's more than a standard deviation of the expected average roll of 10.5.


Sure, it's definitely abnormal. It's just not the kinda thing where I'd cower away from the screen in stunned silence at this horrible luck, or question the accuracy of the results. It's abnormal, but it's a normal level of abnormal.

Let's handle this rigorously. Here we go, Math!

(Please don't run away from the four-letter word. Also, I would appreciate if someone could double-check.)

Assuming a fair d20, it is has a uniform, discrete probability distribution. The probability of any roll is 1/20.

The expected value, or 'average' is μ = E(X) = (1/20)*(1 + 2 + 3 + ... + 20) = (1/2)*(1 + 20) = 10.5

The standard deviation is σ = sqrt [(1/20)*(1^2 + 2^2 + 3^2 + ... + 20^2) - 10.5 ^ 2] = sqrt {[(20^2)-1]/12} = [sqrt (133)]/2 ≈ 5.766

Of course, the above just gives the standard deviation of individual die rolls (across the whole
'population', or all possible outcomes), and not of averages of collections of rolls.

Suppose we take a sample, that is, a collection of die rolls. the sample size, n, is equal to the number of die rolls. Each such collection has an 'average' or mean x̄ (x bar). The standard deviation of the distribution of sample means (that is, how the 'averages' of the collections of rolls are spread out), σ_mean, depends on the sample size, n, and the population standard deviation, σ:

σ_mean = σ / (sqrt [n])

Or, in other words, the more die rolls we take, the more we expect the 'average' of the collection of rolls to be close to the ideal 'average', 10.5, that is, they deviate less.


Do you use physical D20 or is this Dicebot thing doing your rolling for you? HOw many rolls do you have recorded in your dicebot? Cuz rolling an average of 8 after like 10 rolls is a very different beast then rocking an 8 average after 150 rolls. ...

--

Alright, so how do we know if there's something wrong with the rolls? Strictly speaking, you can't be 100% sure, as that would require an infinite number of trials. We can, however, be say, 95% sure. This is known as the 95% confidence level in statistical analysis, and is generally the most commonly chosen (you see this in scientific papers whenever somebody mentions α = 0.05)

We now construct two hypotheses:

H0: The dice were fair, that is, the sample mean (dice 'average') did not diverge significantly from the population mean (ideal 'average') .

H1: The dice were not fair, that is, the sample mean diverged significantly from the population mean.

H0 is known as the null hypothesis. (Also, the above is a two-tailed test, as we see later).

Now, how do we know which one to reject or accept? Well, we use a selection criteria (explained later). Per the Central Limit theorem, the sample means for any population distribution approximates a normal distribution (looks like a bell-curve) for a large enough sample size (generally taken to be n ≥ 30). The curve looks like this:

http://i61.tinypic.com/2poqdm8.png

Here, the value -1.96 and 1.96 are the critical z-values (a number associated with the normal curve). As you can see, data that fall within the two values are within the 95% 'margin of error'. Data that fall to the left or right of the critical values are outside the margin of error, and we can be 95% sure that the dice were not fair.

Our selection criteria is simple; if the sample means ('dice average') has a z-value under -1.96 or above 1.96, we reject the null hypothesis and declare that the dice are unfair. Otherwise, the dice are fair.

The z-value is calculated as follows:

z = (x - μ)/σ_mean

Since we can't exactly determine the value of σ_mean as n was not provided by the OP, we have to modify things a little:

z = (x - μ)*(sqrt [n])/σ

n = [z*σ/(x - μ)]^2

We pick z = -1.96 (the lower critical value, 'though strictly speaking it doesn't matter as we are going to square the result later on). We have determined earlier that σ = 5.766, μ = 10.5. x is the unlucky 'average' roll obtained by the OP's friend. Then:

n = [-1.96*5.766/(8 - 10.5)]^2 ≈ 20.44.

For any collection of d20 rolls with at least 21 rolls, having an average of 8 or below is enough to declare that the dice were unfair, at a 95% level of confidence. I'm sure we make far more rolls than that during the course of the game, so yeah, something is probably wrong with the dice (or whatever PRNG being used).

Okay, you might ask, but there's a 5% chance you're wrong!

Sure. So let's crank things up a bit, shall we? We pick a confidence level of 99.99%. This is pretty much the highest you'll ever see in scientific papers.

The critical z-values are given by -3.29 and 3.29. Plugging in the above formula:

n = [-3.29*5.766/(8 - 10.5)]^2 ≈ 57.58.

I'm sure we make more than 58 rolls in say, three sessions or so. So, yeah, something is definitely wrong with the dice.

(And this time we're 99.99% sure!)

[eggynack]

Nifty mathery, that. However, you should probably revise x up to 8.79, which is apparently the actual average roll. Additionally, I'm not sure if the other players are using the same roller, and if they are, then those results should possibly be averaged in. In any case, when I fit things into the equation you provided, it came out as 43.679. Still less dice rolls than were likely done, but a bigger number at least.

Edit: The other error bars give 123 rolls, incidentally, assuming the revised x.

[prufock]

He is becoming frustrated, depressed, and tempted to just let his character die because the dice just aren't falling in his favor.
Protip: Changing characters will not affect die rolls.

Reversing his roll (ie, his actual roll = 21-d20) is a decent option, allowing him to choose before each roll which it is.

Rolling 2d10 instead of d20 will give a more average result.

If you use either of these, let the other players take the same option.

[Feint's End]

Protip: Changing characters will not affect die rolls.

Reversing his roll (ie, his actual roll = 21-d20) is a decent option, allowing him to choose before each roll which it is.

Rolling 2d10 instead of d20 will give a more average result.

If you use either of these, let the other players take the same option.

But caster don't roll as many die. That's why rerolling is a reasonable idea.

2d10 has a higher average than a d20 .... 3d6 would do the trick though. Either of those changes the chance to crit quite dramatically (1/100 for the 2d10 and 1/216 for the 3d6) though so depending on his liking this might not be an option.

[Extra Anchovies]

2d10 has a higher average than a d20 .... 3d6 would do the trick though. Either of those changes the chance to crit quite dramatically (1/100 for the 2d10 and 1/216 for the 3d6) though so depending on his liking this might not be an option.

Unearthed Arcana has suggested adjustments to threat ranges for the 3d6 bell-curve system. If your group is composed of people who can all add quickly, 3d6 can be a lot of fun. Removing some of the randomness means less incentive to take 10, which means more dice rolls, and dice rolls are fun.

[DoomyDoom]

I'm sure we make more than 58 rolls in say, three sessions or so. So, yeah, something is definitely wrong with the dice.
Checked the math, looks good to me, so I second this conclusion. Out of curiousity I took all the averages OP provided and threw them together in the same sheet, sorted by |x-μ| to check the actual difference from the average... guess what - only wizard and ranger are close enough to average to conclude the dice are fair. Everyone else is outside 95% confidence interval. I assumed ~50 die rolls to account for people who only played for 3 sessions, but if you crank up the amount of rolls to ~500, only wizard's and ranger's results will remain plausible (within 99% confidence, but even then over 95%). So while Swordsage is rolling implausibly low, 3 out of 5 players seem to be actually rolling way too high.

[NNescio]

Nifty mathery, that. However, you should probably revise x up to 8.79, which is apparently the actual average roll. Additionally, I'm not sure if the other players are using the same roller, and if they are, then those results should possibly be averaged in. In any case, when I fit things into the equation you provided, it came out as 43.679. Still less dice rolls than were likely done, but a bigger number at least.

Edit: The other error bars give 123 rolls, incidentally, assuming the revised x.

Thanks for the update. Can't average all the rolls together though, as I'm not sure how many rolls each player had made. I suppose I can make some assumptions that each person makes around the same number of rolls (with the two new players making 3/12 = 1/4 of the rolls), but that just... doesn't seem like a good idea. Also, generally in statistics we try to avoid lumping data from multiple sources together, as they may have been obtained in a different manner (different PRNG seeds, in our case. Or maybe someone is cheating with hack-fu, but I doubt that).


Checked the math, looks good to me, so I second this conclusion. Out of curiousity I took all the averages OP provided and threw them together in the same sheet, sorted by |x-μ| to check the actual difference from the average... guess what - only wizard and ranger are close enough to average to conclude the dice are fair. Everyone else is outside 95% confidence interval. I assumed ~50 die rolls to account for people who only played for 3 sessions, but if you crank up the amount of rolls to ~500, only wizard's and ranger's results will remain plausible (within 99% confidence, but even then over 95%). So while Swordsage is rolling implausibly low, 3 out of 5 players seem to be actually rolling way too high.

Thanks for going over my math. And good job doing the other calculations. I'm starting to suspect the PRNG is improperly seeded based on system clock (versus the actual system time), with the game being conducted online with players from multiple time zones, so geographic location has became a factor.

Of course, this is just a wild hunch.

[CubeB]

Thanks for the update. Can't average all the rolls together though, as I'm not sure how many rolls each player had made. I suppose I can make some assumptions that each person makes around the same number of rolls (with the two new players making 3/12 = 1/4 of the rolls), but that just... doesn't seem like a good idea. Also, generally in statistics we try to avoid lumping data from multiple sources together, as they may have been obtained in a different manner (different PRNG seeds, in our case. Or maybe someone is cheating with hack-fu, but I doubt that).



Thanks for going over my math. And good job doing the other calculations. I'm starting to suspect the PRNG is improperly seeded based on system clock (versus the actual system time), with the game being conducted online with players from multiple time zones, so geographic location has became a factor.

Of course, this is just a wild hunch.

Hm...

Looking over stuff.

Swordsage and Homebrew Guy live in Texas, but Homebrew Guy is from out of the state (From the east coast) and also doesn't roll that often, so his computer might be off, or it could be a lack of rolls.

Wizard is UK.

Warlock and Incarnate are on the East Coast.

I do not know where the Ranger lives.

[Hippie_Viking]

Doesn't computer rng based on time use milliseconds? Or is that only possible with numbers up to 100? Would it be possible to do bigger numbers with this somehow?

[CubeB]

I can roll up to d999999?

[Segev]

An option, which is statistically the same, would be to have people the swordsage is attacking roll their AC, and have him have a static to-hit.

He tells you his attack bonus plus 10, and then YOU roll for his target, subtracting 10 from the AC and then adding a d20. (Maybe a d20-1, since technically this shift is statistically in the favor of the target because it improves the expected average by .5 and removes .5 from the theoretically expected average of the swordsage).

This takes the die out of his hands, at the least. And, as DM, if you roll secretly, you could always fudge it if you think he's missing too much.

[NNescio]

Doesn't computer rng based on time use milliseconds? Or is that only possible with numbers up to 100? Would it be possible to do bigger numbers with this somehow?

Windows stores system time in two formats. SYSTEMTIME is basically Year/Month/Day/Hour/Minute/Second/Milliseconds. FILETIME records the number of 0.1 microsecond ntervals/ticks (or 100 nanoseconds/0.0001 milliseconds) the system is away from 1 January 1601 00:00:00 UTC.

Unix, IIRC, stores time as the number of seconds away from 1 January 1970 00:00:00 UTC, with some weird handling of leap seconds.

Macs, I dunno.

[CubeB]

An option, which is statistically the same, would be to have people the swordsage is attacking roll their AC, and have him have a static to-hit.

He tells you his attack bonus plus 10, and then YOU roll for his target, subtracting 10 from the AC and then adding a d20. (Maybe a d20-1, since technically this shift is statistically in the favor of the target because it improves the expected average by .5 and removes .5 from the theoretically expected average of the swordsage).

This takes the die out of his hands, at the least. And, as DM, if you roll secretly, you could always fudge it if you think he's missing too much.

Hi, Seg.

That could work, actually. It's another short term solution, but I really want to ensure the dice aren't skewed at all.

[Fax Celestis]

Hi, Seg.

That could work, actually. It's another short term solution, but I really want to ensure the dice aren't skewed at all.

You could allow him to use a different roller. Try random.org or invisiblecastle.com.

[prufock]

But caster don't roll as many die. That's why rerolling is a reasonable idea.
Presumably he created a swordsage because he wanted to play a swordsage, just an effective one.

[Zirconia]

When we had an issue like this IRL with physical dice, after 5 fumbles in one session, the DM let the player use a deck of cards to "roll". Pull the face cards out, red = low, black = high and you have 1-20 x2. The upside was at least when you pulled a 1, you knew that was one less you would have to face, and you were one closer to the nat 20. :)

[eggynack]

Presumably he created a swordsage because he wanted to play a swordsage, just an effective one.
Maybe, but it's still worth asking. If he doesn't go for it, then the issue definitely shouldn't be pushed, but it's a pretty good solution if it works out. Even if the dice thing can be fixed, it could be nice to get away from the whole luck issue for awhile.

[DoomyDoom]

Ah, it actually just hit me. Are all of you using the very same dice roller? Then shouldn't we actually check the average for ALL players? The Central limit theorem, which we used before, assumes a sufficient amount of random variables, as well as "well-defined" average and variance. So, what if that RNG OP's group is using is just bad until we get a certain amount of tests?
So, assuming average M=10.5, standard deviation S=5.766 (same as before), let's aggregate those guys:

Bob (Swordsage): 8.79
Cobalt (Incarnate): 12.32
Ara (Wizard): 11
Mary (Warlock): 11.91
Skunk (Ranger): 9.96
Vanel-Xo (Homebrew Thingy): 11.82
Average = SUM/6 = 10,96666667
Assuming, again, 50 rolls per person we get n=300 rolls.
Testing against 95% confidence (z=1.96) we get:
z = (x - M)*(sqrt [n])/S = (10.97 - 10.5)*sqrt(300)/5.766 = 1,401821673 < 1.96 !
So, uh, unless I made a blunder since I'm very tired, dismissing RNG as bad might be a little too fast. It's hard to believe that some of your players got only high rolls out of it and some got only low ones thought... I don't know, maybe the RNG is faulty in a sense that it doesn't average until 300 rolls or something... or that 1% chance really happened to you (as in the rolls do average out, but SS got only low ones). If so, congrats on not being statistical average, I guess :smalltongue:

[eggynack]

Ah, it actually just hit me. Are all of you using the very same dice roller? Then shouldn't we actually check the average for ALL players? The Central limit theorem, which we used before, assumes a sufficient amount of random variables, as well as "well-defined" average and variance. So, what if that RNG OP's group is using is just bad until we get a certain amount of tests?
I mentioned the possibility of averaging results. The counterargument, and it's a reasonable one, was that even on the same RNG, different players could be getting different results due to living in different time zones, as RNG's are usually based on the system clock. I'm also not all that sure how a particular RNG could be bad until you run it enough times.

[Madwand99]

There is only one main strategy anyone can use to avoid bad dice luck: you need to pre-generate rolls and make sure the distribution is fair. Using the suggested method of drawing cards from a deck works. You can also just use a spreadsheet (if online, google docs has a nice one) and RANDBETWEEN to generate a set of rolls, then use an averaging function to verify the rolls average to 10.5. The player can either take their rolls directly from the sheet in-order, or if some randomness is desired, roll a d6 (or any die size desired), count off unused rolls, choose the roll selected, and mark it off so it won't be used again. This will guarantee your player has average luck.

[DoomyDoom]

I mentioned the possibility of averaging results. The counterargument, and it's a reasonable one, was that even on the same RNG, different players could be getting different results due to living in different time zones, as RNG's are usually based on the system clock.
Assuming the actual code runs client-side, sure. On an off-chance that a particular implementation uses server-side RNG we get a given seed that doesn't change for players. Given that OP mentioned it's an "IRC based dice bot", the whole thing becomes plausible, since IRC (chat client) probably simply receives text with roll result. This, however, absolutely blows up the whole "number of rolls" theory since RNG seed of IRC bot probably doesn't change all that often, so we can assume as big of a sample as we want, so the hypothesis of a "biased until n rolls" formula (which was what I meant by "bad until enough runs" RNG) can be more or less dismissed.

[Fax Celestis]

I mentioned the possibility of averaging results. The counterargument, and it's a reasonable one, was that even on the same RNG, different players could be getting different results due to living in different time zones, as RNG's are usually based on the system clock. I'm also not all that sure how a particular RNG could be bad until you run it enough times.

If it's run off a bot, though, wouldn't it use the bot's local system time?

[jjcrpntr]

So, I have a friend (A Swordsage), who is getting really frustrated, because he can't hit. Literally, he cannot.

It's an e6 Eberron Game. The party consists of a Human Transmutation Specialist, a Shifter Ranger, a Human Warlock, a Warforged Incarnate/Totemist (Meeeee), some... weird homebrew Wordcaster thing no one understands, and the aforementioned Human Swordsage.

The issue is, for whatever reason, he can't seem to roll higher than a 3.

Seriously, he misses with touch attacks. He is practically dependent on our Hero Point mechanic. According to our dice bot, his lifetime roll average is 8.

He is becoming frustrated, depressed, and tempted to just let his character die because the dice just aren't falling in his favor. No one else has this problem.

Is there anything I can do to help him improve his chances of at least hitting?

I had a guy in my group that I think over the course of 4 months has maybe hit 6 times. I'm not even kidding. Dude rolls 1,2, or 7 on just about any attack roll. Then gets 18-20 on saves/skill checks. Luckily for us/him he thought it was funny but we all felt bad for him. It was particularly troubling for a few games where he was really the only melee, heck practically the only damage they had in the party.