Double Deck Exit Count QuestionAs I am playing the 10 and five dollar games, my bedspread, obviously looks a bit ridiculous as such I’d like to know at the 26 and 52 penetration levels, are there Hilow or Keio eggs accounts where I should exit until the shuffle.

Playing the lower limit double deck, H17 DAS GAMES with 67% to 75% penetration games at about 10 casinos here in Las Vegas for the next 30 days.  I have done this many times in the past but not for 30 straight days. Max bet is typically $70 or 2×$50.  I do not go higher than that for longevity reasons at max spreads.  I have been backed off approx 10 times at various joints thru the years, but somehow only one casino in Summerlin seems to remember me easily.  As I am playing the $5, $10 and $15 games with minimal cover play, my bet spread after 20-30 minutes usually looks a bit ridiculous.  As such I’d like to know EV calculated exit (HiLo or KO) counts at the 26 and 52 penetration levels, etc, i.e. where I should exit until the shuffle?

Thanks in advance!!

Don't know how you would ever propose to pull off constantly exiting if the count goes negative. Almost impossible to play a DD game that way. In any event, it would always be

TC = -1 or less.

Don

Thank you Don.  I definitely could not and do not exit like that, I agree it would be impossible to pull off.  I would estimate I exit about 25% of games before the shuffle after playing off the top, doing so for bathroom breaks, to take a fake phone call, or to order a drink from a waitress.  I then re-enter at the shuffle.  With an average of two other people playing at the table, I do not think exiting one out of four shuffles looks obvious.  But then again, I could be wrong, some of my mid game backoffs could have been triggered by doing so prior.

Thanks for the answer Sir.

I think you have to bite the bullet and play most DD in play-all mode. I did so for forever in Las Vegas, but, admittedly, it was a very long time ago.

Don

In https://www.blackjack-hack.com you can simulate exiting the shoe at different counts. Below the betting with the count table you can select a count to exit the shoe with Option: Quit shoe at count less than or equal
 

You can set your game conditions, number of decks, penetration, count, betting spread and strategy and try exit points and see results. 
 

If you exit at -3 or -4 with hi-lo I believe it comes out that you exit one out of every 3 shoes or so. But with your fairly high bet spread you mentioned you may be able to play through the negative counts. Or you could exit at around -8 or or -9 or below to cover the really bad conditions without frequent exits.

Phil

Is this BJ simulator reliable ?

With simple sims, I can't reconcile anything with what I get from CVData ...

Phil, this is an important comment from someone whose opinion matters. You have to be able to validate your work vs. the industry standard. If you do simple sims, where all the conditions are identical, and your output doesn't agree with Norm's, then you have to be able to justify why that would be the case.

Maybe GMan can give us an example of the discrepancies he's finding.

Don

Eureka and apologies :-)

This is an happy situation where I found my suspicions were wrong !

I used the base game: 6 decks, S17, DAS penetration at 4.5/6 with the sweet 16 to compare with the canned CVCX sim.

First, the number of "rounds" may be misleading. You cannot run more than 2,500,000 rounds (25,000 hours) per sim, but you can get the average for mutiple runs. My PC is slow so I patiently did 16 runs for 40,000,000 rounds (40,000 visits of 10 hours or 1,000 hands).

What confused me earlier is the way the results are displayed. The $ won per hour (100 rounds) are close enough $53 vs $55

The % Won/Lost (called EV% in CVCX) is lower because it takes into account all splits and double, the Total Bet Advantage is used instead of the Initial Bet Advantage.

Same for the average bet where it takes into account all splits and doubles, producing a higher average bet.

When you add variance and a short sample, this could lead to misinterpretation of results like I did the first time.

I would also like to ask what is the best interpretation I could have of "Risk Adjusted Return" box.It comes from $won/session divided by SD/session x 100.

In my sims we get $530.22/$3856.8 x 100 = 13.75

Is it some DI per session ?

"The % Won/Lost (called EV% in CVCX) is lower because it takes into account all splits and double, the Total Bet Advantage is used instead of the Initial Bet Advantage."

Phil, if this is the case, I suggest you change it so that your output conforms to that of all other blackjack researchers. NO ONE does it your way, so it's going to be detrimental to your reputation's growing, if you're out of step with how the BJ community interprets results. This is important!

Don

Just saw this, will take a look. Thanks.

Phil

By base game do you mean the game that my simulator initializes with? Just so I can make some runs with the same inputs. Can look into this without using the same inputs, but may as well have that.

Are using the count and betting levels it initializes to, or what count and betting levels you are using? Any values should be fine. Also my simulator comes up with hit soft 17.

Thank you,

Phil

Oh, also it is very likely slow due to my web server and code. All the hands and calculations run on a server.

Thank your for being patient and doing so many runs!

Phil

I've been looking through my code and far as I can tell the % of $ Bet is the % won or lost of the total $ bet. The total $ bet includes doubles, splits, and insurance bets and has been this way for quite a while. I added Total to the description to clarify and added the total $ bet just above it.

I will look through some more to see if I can find anything.

I am including the total initial bet and average initial bet along with the total bet in the table further down on the right side of the screen.

Thanks,

Phil

For the Risk Adjusted Return I am comparing the average amount earned or lost per visit vs. how much the outcome of a visit can vary. It's an attempt to see how reliably one can expect the average amount as compared to how different the actual outcome visits can be. So the higher the average amount to the variation (standard deviation) in outcomes gives a higher index.

Thank you,

Phil

I found a problem where I was not adding in the results of insurance bets in the case of splits. The amount of the insurance bet was being added to the total $ bet but the result was not tallied to $ won or lost where the decision on the hand was or would have been a split.

I fixed this. The % $ bet could be higher if insurance bets are significant net wins since the problem was that it was included in the total $ bet but not $ won. This would make the difference under question even larger so will continue to look to see if there is anything else.

Thanks,

Phil

Thanks for many clarifications.

The "base game" I was referring to is: 6 decks, S17, DAS and penetration at 4.5/6 with the use of the Sweet 16 index.

A 1-16 spread was used on one hand:

TC < 1 = $25

TC = 1 = $50

TC = 2 = $100

TC = 3 = $200

TC >= 4 = $400

"For the Risk Adjusted Return I am comparing the average amount earned or lost per visit vs. how much the outcome of a visit can vary. It's an attempt to see how reliably one can expect the average amount as compared to how different the actual outcome visits can be. So the higher the average amount to the variation (standard deviation) in outcomes gives a higher index."

This is interesting but why not used the well established DI and SCORE ?

Thanks, I will make some runs with these inputs and see if I can find why I might be getting some different % returns than expected. 

I probably don’t understand how to apply SCORE to these simulations. Not sure if I would have to set an optimal betting ramp and get a 13.5% risk of ruin and see what the return is with 100 hands per hour and a $10k bankroll. Since the betting levels are user input maybe i would have to compare it to an optimal betting ramp, not sure.

I don’t think I get in 100 hands an hour in the casino. Seems that there are lots of things going on that slow the game down, and a lot of full tables. Not sure how it is in most places, and not sure it matters for comparative calculations but could be different, not sure. 

I had forgotten about the Desirability Index. Return over standard deviation seems similar but will look at the DI some more.

Thank you,

Phil

I wasn’t aware of this problem at all.  I verified every numbers and found them correct. Are you saying there was a tiny problem? If you can, please use a real card example to show where the problem was. Thank you for this work!

With 8,8 vs. Ace or A,A vs. A, if the count was at or above the count to take insurance i would take insurance and add it to the total $ bet but I would never evaluate the outcome of the insurance. It could have also happened with other splits against Ace if basic strategy was modified to add additional splits against Ace.

When I was looking into the questions here I was checking the total $ bet and adding the initial bets, doubles, splits, and insurance bets won or lost and it was not coming out to the total $ bet that was displayed and used for the % won or lost. I looked at how many times insurance had been taken it was sometimes greater than the number of insurance bets won or lost and traced it back to not evaluating and recording the outcome of the insurance decision in these cases. 
 

It didn’t normally show up with lower bets since insurance was normally being taken with higher counts, bets. And it didn’t always show up when running the simulator if insurance had not been taken when dealt 8,8 or A,A during the run. 

Thank you,

Phil

The only other situation is 9,9 vs. A. Player splits @ HiLo TC=+2 for hit-17 games but @ TC=+3 for stand-17 games.