I want to sim some index numbers for various counts using SBA and am wondering how many hands is generally considered sufficient. The default setting is 200 million, which runs real quickly on my computer (less than an hour). I tried setting it to 10 billion, which is taking a bit longer (several days, and counting). What is considered a sufficient number of hands to come up with accurate index numbers?
> What is considered a
> sufficient number of hands to come up with
> accurate index numbers?
I think that depends on the play in question. Some can be decided in only a few million hands, others are still a toss-up after 10 billion. Doing 10 billion hands for all indices is probably overkill unless you have a VERY fast computer, a LOT of time to kill, or only a few indices to run.
SBA lets you set the confidence interval (in no. of standard deviations) as well as the number of hands. It stops when it is within the confidence interval or at the maximum number of hands, whichever comes first. I use 4.5 SDs and 10 billion hands for generating high-quality indices overnight.
CVIndex just has three settings (from memory: "Quick and dirty", "Standard" and "Beat to Death"). If Norm is reading this, I'd be interested to know what these settings actually correspond to?
> CVIndex just has three settings (from
> memory: "Quick and dirty",
> "Standard" and "Beat to
> Death"). If Norm is reading this, I'd
> be interested to know what these settings
> actually correspond to?
I used three English language settings so I wouldn?t have to include a math lessonPlus it allows added flexibillity. CVIndex runs different numbers of hands for different circumstances. For example type of index, level of strategy, balanced vs. unbalanced, risk averse vs. EV maximizing, resplit, DAS, all affect the number of hands. The ?accurate? setting can run as many as two billion but generally tops out at about 700 million. Of course if the index is obvious it runs far fewer. Confidence level is about 4.5. ?Beat to death? can run ten times as many and ?quick and dirty? maxes at about a fifth as many but normally runs far fewer.
I initially ran the index numbers using the default settings of 200 million hands and 3.5 SE. This was really quick (less than 1 hour) and I wanted to step it up a bit, so I maxed it out at 5 SE and 10 billion hands, plus tried to generate 3 full sets of indexes simultaneously (1,2 and 6 deck). This ran for about 3 days on my 1 month old PC, and was still not close to being done, so I nixed those and restarted the 1 deck index generation last night at 4 SE and 1 billion hands. It was almost done by the time I left this morning (10 hours). Just non-DAS splitting to go.
I plan to re-run the index generator for any index numbers where the max hands was reached before the 4 SE threhold, using 10 billion hands. Doing it for just those close call indexes will hopefully save alot of time. Of course, this will also only be for indexes that I actually plan to potentially use (-10 to +10). If the toss-up is between -25 and always split, then I really don't care.
That unless your PC has more than one CPU, running three generations at once isn't any faster than running one after the other. Each will run at one-third speed.
> That unless your PC has more than one CPU,
> running three generations at once isn't any
> faster than running one after the other.
> Each will run at one-third speed.
I figured that out the hard way
From now on, I will stick to running one sim at a time.
Posting Permissions
-*-
Bookmarks