Hi, Norm!
VERY interesting.....
Hmmmm... more research on Random Number Generators is called for....
I would be surprised if there were any user control over the RNG used in BJ678. I'll look harder at that issue.
Thanks! Very helpful!!!
SiMi
Some more on this. For a standard LV strip game counting, EV standard error is about .09 for 20M hands. It's about .012 for 1B hands. If you are consistently getting result differences of only .01 between 20M and 1B hands, the RNG is quickly degrading and the software simply cannot accurately sim more than 20M, or some lower, number of hands. That isn't enough hands.
"I don't think outside the box; I think of what I can do with the box." - Henri Matisse
Norm
Yes, I believe I see what you're saying. Those differences are huge - an order of magnitude!
I AM consistently getting result which differ by only .01 between 20M and 0.5B hands. I rarely run more than 0.5B hands on BJ678 because the documentation says that value gives "superb" results. Often, I'm satisfied with a rough estimate so I'll use 20M to save time. Actually, I see very little difference after about 10M hands but I go to 20M just to be safe. I'm probably missing some setting on the program...
I am now researching the limits of the RNG used in Java (there are at least 2 different utilities for that). This is a big issue for simulators (not so much for game PLAYING) and I hadn't actually gotten to that point in my programming yet. I remember reading somewhere (you may have written it!) that it's necessary to develop techniques other than just playing the game at warp speed for time-effective sims. I suspect I'll get there very soon. But, I've been eager to verify the sim results from BJ678 for a long time!
I recall a free open source sim written in XBASIC that generated some interest and controversy a while ago. I started looking at it about a year ago but decided it would be more satisfying to just write my own stuff rather than decipher someone else's code and then port it to some other language I prefer.
Again, thanks for your time! It's VERY VERY HELPFUL and VALUABLE!!
SiMi
In case anyone is still reading and cares...
Quick research shows the Mersenne Twister is a good candidate for implementing your own PRNG with a drop-in subclass available for Java to replace Sun's Random util:
http://en.wikipedia.org/wiki/Mersenne_twister
http://www.cs.gmu.edu/~sean/research...neTwister.java
Guys, thanks for all the great info and the warm welcome! SiMi thanks a lot for breaking it down for me the way you did, you helped out tremendously! I guess no more OK casinos for me and the better half! I just thought id give it a shot since I read a few books and have been practicing a lot at home but boy was I wrong! I guess it's back to the kitchen table for us, at least till we get it all right and build the bankroll a little more. Speaking of, I'm sure you vets get this question a lot but what would be a good starting BR to go into card counting at least part time?? Hey metronome! I see your from the Big D as well, where do you usually play if you don't mind me asking? Thanks again guys hope to hear from y'all soon!
Bookmarks