Computer Perfect Play

[DSchles]

Short answer: yes, the concealed computer was reckoning every card played.

And no, especially for the shoe game, 3% doesn't seem the least bit low to me. When this study is completed, a lot of people who think they have reinvented sliced bread and play almost "perfectly" are going to be awfully disappointed by what the edge actually is.

Don

[DSchles]

Let's not put the cart before the horse. Let me see how it all turns out.

Don

[DSchles]

Understand that level 4 or 5 counting systems are not "perfect" play.

Don

[DSchles]

Don't be hard on Tarzan ... he's not very good with numbers!!

Don

[davethebuilder]

Understand that level 4 or 5 counting systems are not "perfect" play.

Don

That's true. From memory the advantage of using Optimal Strategy at each TC resulted in only a 1% increase in win rate over using a full set of indices. It will be interesting to see the BJ results.

[DSchles]

"From memory the advantage of using Optimal Strategy at each TC resulted in only a 1% increase in win rate over using a full set of indices."

From memory of what??

Don

[Gramazeka]

"From memory the advantage of using Optimal Strategy at each TC resulted in only a 1% increase in win rate over using a full set of indices."

From memory of what??

Don

Maybe book by John May...

[BJGenius007]

Today your best advantage obtained at deeply dealt shoe games (6deck/0.5 cut off) and a huge spread the best advantage is just over 2% but you must use Hiopt2/ASC full indices.

What kind of bet spread? Typical 20 to 1 spread for 6D, or 1000 to 1 spread?

[davethebuilder]

"From memory the advantage of using Optimal Strategy at each TC resulted in only a 1% increase in win rate over using a full set of indices."

From memory of what??

Don

From my memory of the results that were published in her book and the Optimal Strategy charts that were published at www.spanishcountess.com.

[Three]

How kind of bet spread? Typical 20 to 1 spread for 6D, or 1000 to 1 spread?

For a top non-custom count about 25 to 30 to one.

[Heisenberg]

Yes, I am. We are working on the answer to the precise question that the OP posed. And no, the answer isn't in print anywhere.

I am working with Eric Farmer and Gronbog, and I'm confident that we'll have answers. If the study, along with another than Gronbog is doing with Tarzan, proves interesting, I am thinking of adding another appendix to BJA3 and publishing the findings when the book is reprinted early next year. This is computationally very intensive, so don't expect a quick result.

Stay tuned!

Don

Any news on this ? Thx

[ZeeBabar]

Practical application question to a mathmetician maybe? Based upon the low numbers stated, with a bet spread of $$25-$200 at a DD game, and not playing perfectly, using I-18, thus an edge of 0.5, how much can you expect to earn after you have done played a 100k hands?

The answer might help a lot of recreational players or newbies thinking of full time play.

[RS]

Were these computers actually doing computer perfect play, opposed to just computer play (ie. the computer just counts hi lo, or any other regular count, perfectly)? If what I am talking about is really computationally difficult as Don says, I don't imagine a computer back then having the power to make such calculation in a timely manner. 3% just seems a bit low to me. It seems like the edge a person would get counting a deeply dealt 1D game with good pen.

IMO.....

First off, not sure what Don is doing. I'd assume he's crunching the numbers on a 6 deck shoe. Doing that for "all situations" is going to take a whole hell of a long time. But he's trying to figure out the edge/advantage over the course of all of your play -- that's different from what a computer would be doing in live-play at a deeply dealt SD game. The live-play computer would just be calculating the result for the current hand. I'm not sure what kind of computing power there was back in the 60's or 70's, but I'd imagine it'd still have been very quick.

Figuratively, using a computer at live-play would be like asking a computer to calculate the next prime number after 250,000,000 (which would take about 1 second). Don is trying to calculate all the prime numbers from 2 to 1,000,000,000....or something like that.