Is it just me, or is the Paizo Dice Roller a lot worse than normal lately. . .

[DM Beckett]

I've noticed this sort of above and beyond the normal, but it seems that the Dice Roller on the boards might be messed up.

I've seen an abnormally large number of Nat 1's, 2's, and 3's this last week or two. or just very low. To the point it's starting to affect my games.

Rolls (not picking and choosing):

Depending on the Dice used, I'm bolding the ones between 1/4 and under 1/2. These are also all the rolls in the games, not just mine.
Diplomacy and Intimidate: 1d20 + 9 ⇒ (2) + 9 = 11
Attack: 1d20 + 3 ⇒ (1) + 3 = 4
Damage: 1d6 ⇒ 1
Relf: 1d20 + 13 ⇒ (6) + 13 = 19
Relf: 1d20 + 13 ⇒ (3) + 13 = 16
Planes: 1d20 + 12 ⇒ (12) + 12 = 24
know planes: 1d20 + 19 ⇒ (5) + 19 = 24
KN: planes: 1d20 + 8 ⇒ (6) + 8 = 14
trip: 1d20 + 18 ⇒ (8) + 18 = 26
atk1: 1d20 + 14 ⇒ (4) + 14 = 18
dmg1: 1d10 ⇒ 3
atk2: 1d20 + 14 ⇒ (20) + 14 = 34
confirm: 1d20 + 14 ⇒ (2) + 14 = 16
dmg2: 1d10 ⇒ 1
ref: 1d20 + 11 ⇒ (3) + 11 = 14
Bite v. favored enemy?: 1d20 + 13 + 4 ⇒ (7) + 13 + 4 = 24
Damage, again favored enemy added: 1d8 + 10 + 4 ⇒ (3) + 10 + 4 = 17
Trip, if not already tripped by Xiaobo: 1d20 + 13 + 4 ⇒ (14) + 13 + 4 = 31
Rapid Shot #1, PBS, Deadly Aim, Favored Enemy: 1d20 + 17 - 2 + 1 - 3 + 4 ⇒ (13) + 17 - 2 + 1 - 3 + 4 = 30
Damage, PBS, Deadly Aim, Favored Enemy: 1d8 + 4 + 1 + 6 + 4 ⇒ (3) + 4 + 1 + 6 + 4 = 18
Rapid Shot #1, PBS, Deadly Aim, Favored Enemy: 1d20 + 17 - 2 + 1 - 3 + 4 ⇒ (6) + 17 - 2 + 1 - 3 + 4 = 23
Damage, PBS, Deadly Aim, Favored Enemy: 1d8 + 4 + 1 + 6 + 4 ⇒ (7) + 4 + 1 + 6 + 4 = 22
Iterative Shot, PBS, Deadly Aim, Favored Enemy: 1d20 + 12 + 1 - 3 + 4 ⇒ (11) + 12 + 1 - 3 + 4 = 25
Damage, PBS, Deadly Aim, Favored Enemy: 1d8 + 4 + 1 + 6 + 4 ⇒ (4) + 4 + 1 + 6 + 4 = 19
aoo: 1d20 + 14 ⇒ (18) + 14 = 32
dmg: 1d10 ⇒ 4
Biite Xiaobo (AC 24): 1d20 + 13 ⇒ (20) + 13 = 33 <POSSIBLE CRIT>
Crit Xiaobo (AC 24): 1d20 + 13 ⇒ (12) + 13 = 25 <CRIT>
Damage: 2d8 + 10 ⇒ (3, 4) + 10 = 17
Claw Thoth (AC 26): 1d20 + 13 ⇒ (17) + 13 = 30 <HIT>
Damage: 1d6 + 5 ⇒ (4) + 5 = 9
Claw Thoth (AC 26): 1d20 + 13 ⇒ (1) + 13 = 14 <MISS so no Grab>
Refl: 1d20 + 13 ⇒ (6) + 13 = 19
Bite Dariamus (AC 26): 1d20 + 13 ⇒ (16) + 13 = 29 <HIT>
Damage: 1d8 + 5 ⇒ (3) + 5 = 8
Claw Ezrek (FF AC 19): 1d20 + 13 ⇒ (14) + 13 = 27 <HIT>
Damage: 1d6 + 5 ⇒ (4) + 5 = 9
Claw Osric (FF AC 21): 1d20 + 13 ⇒ (8) + 13 = 21 <HIT>
Damage: 1d6 + 5 ⇒ (2) + 5 = 7
trip1: 1d20 + 16 ⇒ (13) + 16 = 29
trip2: 1d20 + 13 ⇒ (2) + 13 = 15
atk1: 1d20 + 14 ⇒ (16) + 14 = 30
dmg1: 1d10 ⇒ 6
atk2: 1d20 + 14 ⇒ (6) + 14 = 20
dmg2: 1d10 ⇒ 3
Concentration: 1d20 + 13 ⇒ (10) + 13 = 23
Attack: 1d20 + 12 ⇒ (10) + 12 = 22
Damage: 1d6 + 12 ⇒ (6) + 12 = 18
Attack: 1d20 + 12 ⇒ (3) + 12 = 15
Damage: 1d6 + 12 ⇒ (1) + 12 = 13
swift kirin know planes vs dc 15+CR : 1d20 + 19 ⇒ (7) + 19 = 26
atk assist: 1d20 + 6 ⇒ (15) + 6 = 21
Claw: 1d20 + 2 ⇒ (5) + 2 = 7
Dmg: 1d4 + 2 ⇒ (2) + 2 = 4
Attack: 1d20 + 7 ⇒ (3) + 7 = 10
Damage: 1d10 + 10 ⇒ (5) + 10 = 15
Kryssa AoO: 1d20 + 7 ⇒ (11) + 7 = 18 <HIT>
Damage: 1d10 + 10 ⇒ (6) + 10 = 16
Maldoc AoO: 1d20 + 7 ⇒ (1) + 7 = 8 <MISS>
Rapier: 1d20 + 5 + 2 ⇒ (20) + 5 + 2 = 27
Damage: 1d6 + 1 ⇒ (3) + 1 = 4
Maybe this time?: 1d20 + 5 + 2 ⇒ (10) + 5 + 2 = 17
Damage: 1d6 + 1 ⇒ (3) + 1 = 4
Claw 2: 1d20 + 2 ⇒ (16) + 2 = 18
Dmg: 1d4 + 2 ⇒ (4) + 2 = 6
CLW wand: 1d8 + 1 ⇒ (3) + 1 = 4
Perception: 1d20 + 4 ⇒ (4) + 4 = 8
CLW: 1d8 + 1 ⇒ (1) + 1 = 2
Perception check: 1d20 + 5 ⇒ (3) + 5 = 8

Spoiler:

out of 69 rolls these last few days:
7 = Nat 1's (with 3 of those on a d20)
15 = Nat 2's-4's (with 10 of those on a d20)
26 = Below 1/4(ish)
20 = Nat 10 - 20 (d20 only, out of 43)

[Faelyn]

You're not the only one that has noticed this issue... My WoTR group has been rolling consistently low through most of the encounters. Especially recently though...

[Oladon]

It's just you.

69 rolls is nowhere near enough to determine a pattern. Allow me to direct you here.

That was done a while ago, though I have no reason to think anything has changed with the roller since then. Regardless, if you really want to show that there's a lower-than-expected average, you'll need to do something along those lines, including the Chi-squared test at the end.

[Pygon]

In my VTT, people complained that players seemed to roll low a lot, so I secretly coded the randomizer to return (21 - d20) instead of (d20), which gives the same results, just flipped.

No one noticed. So, I changed it back.

[Hayato Ken]

I think the dice roller here is actually better than normal dice by now.
In our home games we have some guys that roll so high on average that it´s already ridiculous, even if you give them new dice.
The distribution here is just average and fights can take longer than three rounds that way, when no one hits. A good thing in my eyes.

Just imagine a movie where fights would be over that fast and everyone would hit all the time.

[Tacticslion]

Well, for me, at least, this is the second time I've had substantially differing (and worse) dice results between preview and post in the gaming area.

The other time happened here/here (which I worked out a solution for here) with a weird double post - you can clearly see the values I typed out don't mesh with the values rolled up. Also a weird double-post-thing with two different identities, and... I really don't know what happened there.

I mean, neither have been terribly important, but... it's rather weird. Also not really an exciting thing if I can't trust my dice values before I post. EDIT: The fact that both were, categorically, worse than what I thought I was posting is disturbing as well.

[Mattastrophic]

Pygon wrote:

In my VTT, people complained that players seemed to roll low a lot, so I secretly coded the randomizer to return (21 - d20) instead of (d20), which gives the same results, just flipped.

No one noticed. So, I changed it back.

Oh, that's clever. I like it.

-Matt

[rknop]

It's intentional that the die rolls are not finalized while you're previewing. It's a basic anti-cheating measure, because one could edit die rolls in and out and keep previewing until you got what you liked. When you preview, you see *a* roll, but not *the* roll.

Once you've posted the post for real, then the rolls are final. If you edit the post, they won't change.

I know on some other message boards, you can see the history of (real, not preview) edits. (Presumably, you could still cheat a little bit by editing your post and rearranging which roll went where.)

[Ross Byers]

rknop wrote:

It's intentional that the die rolls are not finalized while you're previewing. It's a basic anti-cheating measure, because one could edit die rolls in and out and keep previewing until you got what you liked. When you preview, you see *a* roll, but not *the* roll.

Once you've posted the post for real, then the rolls are final. If you edit the post, they won't change.

This is not how it works. There is no logical difference, from a cheating perspective, of previewing an post repeatedly and editing it after it is posted, unless you are assuming the GM and other players are watching the thread 24 hours a day.

The preview should reflect the same values as the final post, because it is better to be able to figure out the effects some something like a natural 1 or a crit without having to edit the post.

There ARE ways to cheat this system, because Paizo.com is a messageboard, not a dice-rolling service. Making it impossible to cheat would require things like blocking the ability to delete posts, or showing the complete edit history of a post, that make it function worse as a messageboard.

If you really want cheat-proof dice, you should be using a service that promises that, like Invisible Castle. (There are still ways to cheat or manipulate Invisible Castle.)

[DM Beckett]

Yes, you can view the results of rolls by previewing the post, and even continue to edit them as needed, previewing the post again and again before actually submitting. I utilize this a lot when I run games, so I can speed up combat.
Attack 1 (vs AC 16): 1d20 + 4 ⇒ (3) + 4 = 7 <miss>
Attack 2 (vs AC 16): 1d20 + 4 ⇒ (2) + 4 = 6 <miss>
Attack 3 (vs AC 16): 1d20 + 4 ⇒ (14) + 4 = 18 <HIT>
Damage: 1d8 + 4 ⇒ (4) + 4 = 8 ouch!!!
Attack 4 (vs AC 16): 1d20 + 4 ⇒ (9) + 4 = 13 <miss>

And there are ways to still cheat.

It used to be that you could make a bunch of posts until you got a roll you wanted, and then just delete all the other posts. Thankfully that one is gone, because unless you kept track of the number of posts in a game's page, or just happened to be online in the few seconds it took to delete other posts, no one would know.

Anyway, we still seem to be getting an abnormal amount of low rolls, but, maybe it's just us. I don't know.

[rknop]

Hum, OK. I had thought that die rolls on previews were thrown away and regenerated when the post was submitted for real. That's certainly how it works at myth-weavers (where I do more of my PbP gaming), and I thought I'd seen it work that way here.

The logical difference between editing after previews and editing after submitted posts is that (at least at M-W, though I guess not here) are twofold. First, if you change the order of die rolls in a submitted post, a "DIE ROLLS WERE CHANGED" warning shows up with the post, but not if you do it only after a preview. Second, the history of submitted edits is stored, so that you can go back and look at previous submitted versions of a message. (At least, I thought they were... I'm not seeing it. Damn, I'm getting old, and my memory is going. I'm pretty sure I've used a message board where moderators could see the full history of changes to posts. Most wikis work that way, of course.)

[LazarX]

You do understand that it's not a true random die roller? That's it's based on a seed number generated from the number of seconds from the previous midnight? That's how it's generally done because it's not really possible to do it any other way on a budget less than the NSA's.

[thejeff]

LazarX wrote:

You do understand that it's not a true random die roller? That's it's based on a seed number generated from the number of seconds from the previous midnight? That's how it's generally done because it's not really possible to do it any other way on a budget less than the NSA's.

True, but largely irrelevant. (Though I doubt thats what it's actually based on. Rolls at the same second of the day will always be the same? That's a really bad seed mechanism.)

Yeah, it's a pseudorandom number generator. So is almost any computerized RNG any of us have every used. Functionally, a good one is as random as dice. Theoretically exploitable in ways dice are not, but less likely to be biased by worn edges or weighting.

[Ross Byers]

There are many ways to get a seed for a random number generator, and I can assure you that it is not 'seconds since midnight'.

[BigDTBone]

I assumed it was the last 3 digits of the gigahertz clock on the server. Which may yield a slightly uneven distribution but would still be very good for randomness.

[Vic Wertz]

The random seed is "the number of milliseconds since the last messageboard post discussing whether or not the die roller is sufficiently random."

[rknop]

LazarX wrote:

You do understand that it's not a true random die roller? That's it's based on a seed number generated from the number of seconds from the previous midnight? That's how it's generally done because it's not really possible to do it any other way on a budget less than the NSA's.

Actually, not so hard any more. Regular everyday laptop and desktop computers are capable of generating truly random numbers. It's a lot slower than the pseudorandom numbers you're talking about, which is why for most purposes people still use pseudorandom numbers. But "true" random numbers from an "entropy pool" are a standard part of mainline OSes nowadays.

Also, PRNGs ("Pseudo Random Number Generators") are very good. Unless you feed the same seed twice, for a halfway decent PRNG you won't notice the fact that it's not really random until you've generated the billions of numbers necessary to start seeing the sequence repeat. Sometimes people will throw out warnings about computer random number generators because they aren't "truly" random, but that's a red herring. For all purposes useful for gaming, they are, and, indeed, they're likely to be more uniformly distributed than any actual physical dice you roll. (Scientists use PRNGs in simulations all the time; there are a lot of people who've spent a lot of time thinking about what makes a good PRNG.)

[rknop]

Vic Wertz wrote:

The random seed is "the number of milliseconds since the last messageboard post discussing whether or not the die roller is sufficiently random."

...so the seed is always between 0 and 5? :)

[thejeff]

Yeah, even a relatively lousy P-RNG is fine for our purposes. If you're using it for serious cryptography you'll have vulnerabilities, but that's a whole different story.