VASL dice bot...

[Jay White]

How many of you guys use vasl a lot and get a lot of DR streaks? Like, a lot of low DR's one after the other (under 5), or a lot of high rolls (over 9)

I noticed Sam B has a huge file of vasl (3?) DR's on his page for statistical testing. Has anyone looked at this data? I dont know how you test for "micro-streaks", but I'd like to know what people find out.

Also, does anyone know much about the actual dice bot used in vasl? Is there a _really good_ randomizer, and does vasl use it?

Am I just super paranoid?

I'd prefer to see real techie answers to this post, but if you feel like saying "thats BS" or "I know what you mean", go for it. It won't help answer the question though Its way too easy to "think" somethings wrong with vasl, but the DRs in one game are insufficient to say how the dice bot works overall. I think.

Thanks in advance,

-Jay

 

[Anonymous]

Hi Jay,

I know what you mean but it's BS....

couldn't resist. I am not too concerned about the vasl bot, because we didn't have too much "strange" results and even if so, it goes for both players so it's at least even.

And now open the gates for the techies here

Greets,

Andreas

 

[Klaus Fischer]

Hi Jay,

I always wonder about the Dicebot 8)
In a recent VASL my opponent (hi Kevin) rolled about 4 CH and about on the resolving side had a lot of 2's and 3's, whereas I had only high rolls.
However in my Turn 3 (only) I also had a nice streak of good rolls.
It appears that the DB can get "stuck" in a streak of rolls
But hey! It's just a game - or...?
Ciao :wink:

Klaus

 

[Jazz]

I noticed Sam B has a huge file of vasl (3?) DR's on his page for statistical testing. Has anyone looked at this data? I dont know how you test for "micro-streaks", but I'd like to know what people find out.

Also, does anyone know much about the actual dice bot used in vasl? Is there a _really good_ randomizer, and does vasl use it?

Am I just super paranoid?

I'd prefer to see real techie answers to this post, but if you feel like saying "thats BS" or "I know what you mean", go for it. It won't help answer the question though Its way too easy to "think" somethings wrong with vasl, but the DRs in one game are insufficient to say how the dice bot works overall. I think.

Thanks in advance,

-Jay

I have noticed streaks in the past. It's gotten to the point where I will look for an inconsequential DR to make after rolling 'cars (and looked for juicy attacks right after rolling 'eyes). I suspect (and it's only a suspicion) that the dicebot is keeping track of the distribution of rolls and is trying to force an "expected" distribution of DR results? Unless you do something with the system clock, random is not that easy a thing to do on a computer.

A techy answer would be tough. In effect, instead of an event being 1 DR, you would need to do an analysis of an event consisting of however many DR's you want to consider a "streak". Well beyond my meager sadistical skills....

 

[SamB]

The data on my site was VASL4. Both VASL3 and VASL4 use Java's random number generator.

If you -really- want to investigate, write to Sun Microsystems about their random number function.

Sam

 

[GazMarsh]

Information on the Java random number generator:

http://java.sun.com/j2se/1.3/docs/api/java/util/Random.html

 

[Bryan Holtby]

There is another thread somewhere in these forums about the bot. It is full of techie information.

Many players feel that the bot is busted, including me, so you arent alone

 

[Largus Means]

All the players whining about the dice bot, suck it up. When your throwing real dice and you hit a "10", "10", "11", "9", "12", "11" you don't cry about the cubes not working. And it does happen. The dice bot is fine, you can't roll "3"s and "4"s all day long. And at the end of the game, 90% of the time you can look back and say, "blew it there" or "did that wrong" and that the dice weren't to blame. And for that last 10%, yeah maybe the dice did screw you, but that happens in real F-t-F games as well.

Largus
6+1

*** DR = 6,4 *** <Largus Means>
ne
damn dice bot has got my number again

 

[Bryan Holtby]

Apparently you didnt look at the other thread, or you would see the bot issue to be valid instead of putting up some "suck it up" ludicrious response to a resonable question.

Hitting a series like you said isnt the issue with real dice or the bot. The issue is what FOLLOWS that streak. With real dice, you get maybe one low roll, one high roll to follow your streak, with the bot, you could very likely end up with a streak of DR 5 or less.

Or, you could find youself in the dreaded 2/12 cycle, in which the higher one roll is, the lower the next roll. I have been and watched a victim of this cycle, imagine the Germans AND the Russians not getting a single hero in 5 Red Barricades CG dates! Tons of berzerkers though.

How about this oddity. Playing 3 CG dates in RB, totaling 17 turns, and seeing only 4 SAN activations (either a 1 or 2 rolled after SAN). Considerably less than the 33% one would expect, it was actually closer to 4%.....While this is possible with cubes, it is yet another bizarre example of something wierd in VASL botland.

While I love VASL, and quite frankly I likely wouldnt be playing ASL at all if it werent for VASL, the dice bot often drives me nuts.

 

[da priest]

All the players whining about the dice bot, suck it up. When your throwing real dice and you hit a "10", "10", "11", "9", "12", "11" you don't cry about the cubes not working. ...Largus
6+1

Wrong, I've been getting those rolls for over a decade, once in a great while I get average rolls and I'm elated... otherwise I just whine...

 

[Bryan Holtby]

*** DR = 6,6 *** <Rob P.>
*** DR = 6,4 *** <Rob P.>
*** DR = 1,2 *** <Rob P.>
*** DR = 1,2 *** <Rob P.>
*** DR = 4,5 *** <Bryan H>
*** DR = 6,6 *** <Rob P.>
*** DR = 1,1 *** <Rob P.>

Just a sample of the standard idiocy of the VASL dicebot........

 

[Bryan Holtby]

Just to add to the above, I also malfed two guns in advance fire that turn, as well as rolling a 12 with a mineroller AND rolled a 12 on a bog check in wire. I might add that I made 11 rolls that turn, 5 of which were boxcars.

for those of you who say that there is nothing wrong with the dicebot....you dont play enough.

 

[SamB]

Apparently you didnt look at the other thread, or you would see the bot issue to be valid instead of putting up some "suck it up" ludicrious response to a resonable question.

Personally, I am note convinced there is a problem. If a problem exists (from what I've read of these discussions) it seems to me unlikely that it would have a significant affect on an ASL game. (with what, 200 DRs??)

The dice -can- and -do- decide the game sometimes. But its not a usual happening...

The "sample" above is ludicrously small to draw any conclusions. And all you guys whinning don't help either. IMO no one has shown that there is a significant problem with the randomness of the dice bot.

JMO, YMMV...

Sam "I lose with or without the dice bot" Belcher

 

[Pitman]

Kinney swears up and down there is no problem.

 

[Bryan Holtby]

Perhaps I can never be convinced that it is fine. I have seen SO many results like I have posted that I wouldnt believe Einstein if he told me the math in the program was sound.

The trouble is Sam, that the above events HAVE decided the scenario, the above DR's I posted and commented on have left 1 bogged FT tank on wire, I wrecked PT mineroller, an OBA immobilized tank and 2 burning wrecks, and two additional tanks with malfed MA's. Which oddly enough, were at the leading edge of the attack and will shortly be burning from 88mm AT fire. How does that NOT decide a scenario?

I had lots, and I do mean LOTS of tanks in the opening OtO CG game, over 30 of them....now I have 18 fully functional AFV's....after 1.5 turns. The dicebot has decided the opening scenario. This is NOT the first time that this has happened, it isnt even the second or the third..... Im not complaining about the number of tanks lost, I expected heavy casualties, I just find it hard to take that I lost 4 to boxcars when I made so few rolls.

While I agree that the sample above is too small to draw any conclusions, it is but another example of how the bot swings from extremes. I have often bitched about how few hero's are seen in VASL. Play a RB CG game and see how often you get a German hero vs a zerker, you will be surprised how many zerkers you get.

I will continue to bitch about the bot until I see a definite change in the boxcar/snakeye syndrome. It happens WAY too often.

I am no stats guy, nor a programmer. Unfortunately my steady opponent had his stats class on the day after pub night, so he didnt retain much of what was taught.

I dont think the issue is with the overall average of the roll, its the high/low swings, the 11/3 or 12/2 or 12/3 or 11/2 swings that drive me nuts.

I just ran another dozen or so rolls.....and guess what....

*** DR = 5,2 *** <Bryan H>
*** DR = 5,3 *** <Bryan H>
*** DR = 6,6 *** <Bryan H>
*** DR = 6,6 *** <Bryan H>
*** DR = 1,1 *** <Bryan H>

and another....


*** DR = 2,4 *** <Bryan H>
*** DR = 2,3 *** <Bryan H>
*** DR = 1,1 *** <Bryan H>
*** DR = 6,5 *** <Bryan H>
*** DR = 5,4 *** <Bryan H>
*** DR = 2,5 *** <Bryan H>


How is it not wrong that the bot swings to extremes? And just what the hell are the odds of rolling snakes after boxcars? Im sure its not as simple as 1 in 36......

Im so convinced there is an issue that I am looking for another program to replace the dicebot. Failing that....I guess I will purchase one of those minicams.....attempt to convice my regular opponent to do the same, and will then resort to using dice.

I guess I am now off to fabricate a dice tower, and if it doesnt look as pretty as others, you can rest assured that it will be stong enough to bash the crap outa my keyboard if I see another 12/2 result on the VASL bot!

 

[Chris Milne]

The odds of rolling snakes after boxcars (or vice versa)? 1 in 36. Unless you have sentient dice.

I've worked with Sam's large spreadsheet of dicebot rolls to see if there is any sequencing of high/low, low/low, or high/high rolls. Unfortunately, even with that many rolls, the sample size isn't huge, but it's not shown anything particularly abnormal. There is the odd blip that I think derives from the small sample size (I think after rolling a 3, a roll of 10 came up unusually often).

 

[SamB]

How is it not wrong that the bot swings to extremes? And just what the hell are the odds of rolling snakes after boxcars? Im sure its not as simple as 1 in 36......

This little rant is not adressed only to the guy who wrote the above, but everyone that complains about the VASL dice bot.

The premise that the dice bot is flawed should be easy to prove. I've provided a spreadsheet with 14,000 dice rolls on my website. If you think the dice bot is flawed, read up on probability and statistics and PROVE IT.

Until you do that, you are just *itching about your dice. A time honored tradition in ASL and wargames in general.

I've seen the kind of swings you talk about OVER and OVER again when using real dice. I've seen games decided by the dice. Complaining about instances of a few dozen rolls are totally bogus and will never prove anything.

Put up or shut up. If there's a problem it should be EASY for anyone with any background in staticsics (I don't have much formal education) to prove. This isn't rocket science. I haven't seen any mathmatical proof of a problem.

put up or shut up....

Sam

 

[Anonymous]

Not quite Sam...

Probability and statistics have very little to do with "random" number generation on computers - look for information on linear congruential generator algorithms.

Basically what it comes down to is that these algorithms, given certain seeds, generate crappy, short and repetetive sequences...

They're not called 'pseudorandom' rather than 'random' numbers for nothing you know.

 

[GazMarsh]

Oops - that was me, forgot to log in.

 

[Nat Mallet]

Pseudorandom number generators generate the same string of characters giving the same seed. So the seed is the dice bot's weakest link.

My question is, why isn't dostime good enough? dostime is basically the number of milliseconds the computer has been up, which can be a fairly big number. If that's used as the seed, then there's no way to predict or influence the number that will be generated.

Unless you're able to time actions in miliseconds, of course.

Nat