von Marwitz
Forum Guru
Maybe then you should roll it with some nice beautiful girls for yet enhanced enjoyment?But, the precision dice are much prettier, and so they are less inclined to piss me off.
von Marwitz
Maybe then you should roll it with some nice beautiful girls for yet enhanced enjoyment?But, the precision dice are much prettier, and so they are less inclined to piss me off.
The very impressive analysis @Sparafucil3 linked to should be convincing for anybody who is interested in the stats.There are statistics tests for that kind of thing. I'm surprised one of our statistically-minded friends hasn't worked that up.
Amen to that. Great post.The very impressive analysis @Sparafucil3 linked to should be convincing for anybody who is interested in the stats.
As somebody who is "statistically-minded" for a living, the thing to remember is that human beings have essentially no ability to perceive randomness whatsoever. Things that are actually random we perceive as clumped, and the things that we call random are actually things that have been very carefully and non-randomly uniformly distributed. Psychologically, we demand that a random DR generator produce 6,7,6,7,8,7,6,7. But a real random process doesn't do that, they produce clumps and streaks and all kinds of anomalies. What @Tom Morin and @Vic Provost et al. want isn't a random number generator at all; they want a non-random number generator that will give them a seven every time with the occasional 8 thrown in for laughs.
If rolling real dice under a camera lets you enjoy the game more, then by all means do it! But if you perceive any difference between the dice and the dicebot, that is almost certainly the placebo effect, and if it's not, it means there is something wrong with your dice, not something wrong with the dicebot.
Anyway, since people's problems with the dicebot are psychological and come from a universal human cognitive bias, more stats won't really help--what's the saying? You can't reason your way out of something you didn't reason your way into in the first place.
They're not quite that dumb.What @Tom Morin and @Vic Provost et al. want isn't a random number generator at all; they want a non-random number generator that will give them a seven every time with the occasional 8 thrown in for laughs.
He only uses the girls to parade around a large TURN marker whenever the game progresses to a new turn, otherwise it's just too distracting.Maybe then you should roll it with some nice beautiful girls for yet enhanced enjoyment?
von Marwitz
Thanks. Does that mean I’m using up my free random bits each time I play? I only get 1,000,000 random bits per day, so I need to conserve where I can.
Unless something has changed, we used the VASL dice statistics regularly - it broke down each player for each game, by different categories of rolls, number of each result, overall average, etc. We turned the stats in with our playtest reports.The ASL Scenario Archive was updated a while ago. With that update, the old "Best Rollers", "Worst Rollers", "Most Snakes", and "Most Boxcars" board went asunder.
Before DRs were represented by dice-depictions in VASL, numbers were displayed. By copying/pasting the log-files into the particular page of the Archive, the rolls were analyzed and counted. Only sessions with a minimum of 50 DRs were considered for a player.
The nice thing was, that this function allowed you to determine your long time DR average.
...
That bit of history given, I do not know if anything like it exists currently but I do not think so.
von Marwitz
No, we don't want a 7 on every dice roll, that's not it at all, believe me, it is just that after having many instances of the dice bot showing streaks of >9 or <5 which went on for 6+ rolls in the middle of playtests we just decided it was not good for the playtest and to switch out to real precision dice for our rolls. We can still see streaks but they tend to be shorter and it just feels more random. I'm sure the dice bot is following the curve with its results, we just don't want to see lots of high or low numbers seemingly clumped together when we played. Might not happen at all for others but I'm glad to be rolling real dice again as is Tom, we are still very happy with VASL for the most part otherwise. Now we can just bitch at the dice and not the dice-bot. Just trying to make our playtesting as good as we can, Vic.The very impressive analysis @Sparafucil3 linked to should be convincing for anybody who is interested in the stats.
As somebody who is "statistically-minded" for a living, the thing to remember is that human beings have essentially no ability to perceive randomness whatsoever. Things that are actually random we perceive as clumped, and the things that we call random are actually things that have been very carefully and non-randomly uniformly distributed. Psychologically, we demand that a random DR generator produce 6,7,6,7,8,7,6,7. But a real random process doesn't do that, they produce clumps and streaks and all kinds of anomalies. What @Tom Morin and @Vic Provost et al. want isn't a random number generator at all; they want a non-random number generator that will give them a seven every time with the occasional 8 thrown in for laughs.
If rolling real dice under a camera lets you enjoy the game more, then by all means do it! But if you perceive any difference between the dice and the dicebot, that is almost certainly the placebo effect, and if it's not, it means there is something wrong with your dice, not something wrong with the dicebot.
Anyway, since people's problems with the dicebot are psychological and come from a universal human cognitive bias, more stats won't really help--what's the saying? You can't reason your way out of something you didn't reason your way into in the first place.
No, we don't want a 7 on every dice roll, that's not it at all, believe me, it is just that after having many instances of the dice bot showing streaks of >9 or <5 which went on for 6+ rolls in the middle of playtests we just decided it was not good for the playtest and to switch out to real precision dice for our rolls. We can still see streaks but they tend to be shorter and it just feels more random. I'm sure the dice bot is following the curve with its results, we just don't want to see lots of high or low numbers seemingly clumped together when we played. Might not happen at all for others but I'm glad to be rolling real dice again as is Tom, we are still very happy with VASL for the most part otherwise. Now we can just bitch at the dice and not the dice-bot. Just trying to make our playtesting as good as we can, Vic.
Yes indeed and the dice are just part of the puzzle in playtesting, we try so hard to 'break' the scenario and make sure BOTH sides actually have a decent chance to win but when the dice don't seem to be 'fair' as described above we just had to do something and playing with the precision dice at least takes one more thing beyond our control out of the equation. Now the dice, at least in our minds, are what they are and so far seems the odd streaks of all high or low numbers for awhile have disappeared to just the usual several rolls of such, which we can live with. It is all about making these scenarios as tight as we and our reinforced squad of playtesters can help mold them. We are not always right and sometimes we miss something that someone like Steve Pleva will figure out later. So be it.I could take the pains to dig out the cognitive dissonances and biases that are the root behind your impressions and then try to find the English translations for them so that you can look up explanations that are better framed than I could put them. But as my interest in cognitive dissonances is more a hobby than a profession, I will leave it at that.
Suffice to say that the lack of streaks is an indicator for things not working at random. That you will judge improbable things much more likely than they really are, especially if you have encountered them personally and even more if emotions (fear, joy, anger) are involved. That you will overrate events that confirm your prior view and underrate those that oppose it.
So if you always had that hunch that the dicebot wasn't doing things properly and are unnerved by long freakish streaks that irritatingly screwed up a promising playtest session all of a sudden, you might just be in the thick of it.
That said, if you feel more comfortable with a camera and dice, go ahead. ASL is about enjoyment and as long as both sides are more happy with dice, all the better.
With regard to playtesting, using your method instead of the dicebot will neither do good or harm IMHO, since it does not make a difference. Playtests and games are rather screwed up by the occasions in which you roll high or low rather than streaks.
von Marwitz
Yeah, it does end up being pretty close to a normal distribution. I guess my only complaint (and it's not a big one) is that there seem to be streaks that happen that look to be an effort to ensure that normal distribution?Hey Tom!
Hope your doing well and designing more scenarios for VotG. What do you mean by dicebot hijinks? Just for sh!ts and giggles, while watching a movie this evening, I started punching the 2-dice button on VASL. I actually punched the damn thing 10,000 times. Here's what the data showed:
5,000 rolls
Die 1 -- 3.53
Die 2 -- 3.52
Total -- 7.05
10,000 rolls
Die 1 -- 3.53
Die 2 -- 3.51
Total -- 7.03
That's pretty close to a statistical average in my book.
Cheers,
Mark
Hi Dave, it would be nice to have back if it can be fit into your already busy schedule.Just a note about the archive... the vasl analyser stopped working after the vasl output changed to include the graphics rather than the actual numbers of the rolls.
Because, whilst it was up, I would frequently get emails asking why the log file wouldn’t upload any more.
I was going to revamp it to read vasl log files but the stats page of vasl became so much better that there didn’t seem a lot of point.
If there’s a demand I could look at re-doing it but I’m not sure there is, really.
Here is an idea:I was going to revamp it to read vasl log files but the stats page of vasl became so much better that there didn’t seem a lot of point.
If there’s a demand I could look at re-doing it but I’m not sure there is, really.
This is what we believe we observe as well. The dicebot feels...streaky, especially if you do rolls very quickly in a row. Like the RNG seed gets stuck. It's exciting and also discouraging to see 4 snakeyes in a roll, or 10 rolls in a row that are 9+, etc.We can still see streaks but they tend to be shorter and it just feels more random.
I recommend you go read the link I posted. The statistical sampling they did in there is FAR more sophisticated that what you imply here. WRT "timing" are you suggesting the VASL dicebot is aware of the importance of the DR or are you saying the timing of when you roll the 1,1. People have repeated the ASL maxim "it's not what you roll as much as it is when you roll it" a couple of times. If you asking about "how often an 8 follows a previous 8", I refer you again to the article I linked. This is addressed. -- jimThis is what we believe we observe as well. The dicebot feels...streaky, especially if you do rolls very quickly in a row. Like the RNG seed gets stuck. It's exciting and also discouraging to see 4 snakeyes in a roll, or 10 rolls in a row that are 9+, etc.
AFAIK the only VASL dicebot stats people are pulling out show a gross mean with a give sample size, which always points to ~7. Which is all well and good, but what about actual distributions and sensitivity to timing of rolls (how quickly or slowly they're made relative to the previous roll), or which button is used (IFT/TH/MC/R/etc), or hell...which version of java they are running (because I assume the rolls are made client side).
We still use the dicebot because it's incredibly convenient, but we often squint at it skeptically due to weird behavior sometimes. /shrug
You can mitigate many of these concerns if you turn on random.org for die rolls.This is what we believe we observe as well. The dicebot feels...streaky, especially if you do rolls very quickly in a row. Like the RNG seed gets stuck. It's exciting and also discouraging to see 4 snakeyes in a roll, or 10 rolls in a row that are 9+, etc.
AFAIK the only VASL dicebot stats people are pulling out show a gross mean with a give sample size, which always points to ~7. Which is all well and good, but what about actual distributions and sensitivity to timing of rolls (how quickly or slowly they're made relative to the previous roll), or which button is used (IFT/TH/MC/R/etc), or hell...which version of java they are running (because I assume the rolls are made client side).
We still use the dicebot because it's incredibly convenient, but we often squint at it skeptically due to weird behavior sometimes. /shrug
We did turn on random.org and that is the bot giving us the wild high and low streaks.You can mitigate many of these concerns if you turn on random.org for die rolls.