LearningCurve
LearningCurve
Joined: May 5, 2020
  • Threads: 1
  • Posts: 2
May 5th, 2020 at 1:32:08 PM permalink
Hi,

I was wondering if anyone has a reference or resource that can help me develop an MS excel version of the wizard of odds blackjack hand calculator? Do you know if he posts the source code for it?

I would like to be able to record cards that have been played and have the spreadsheet calculate the best play. I'm curious to see if the recommended play will deviate much from basic strategy. I have tried to include a link but this forum tool keeps stripping the link, below is the best I can do! You have to go to the wizard of odds website and paste in that path.

/games/blackjack/hand-calculator/
gordonm888
gordonm888
Joined: Feb 18, 2015
  • Threads: 38
  • Posts: 2567
May 5th, 2020 at 3:21:25 PM permalink
Welcome to the forum. The forum does not allow new members to post links, videos or photos until they have been on it for a month. We all found this frustrating when we first joined.

You are asking about developing an Excel version of a composition-dependent BJ hand calculator. I have done that. I am not aware of any internet resources to help with that. It took me a long time to write (and to debug, OMG) and I necessarily use some approximations when I model splitting and re-splitting of pairs. (and, NO, I am not trying to sell you a service or a calculator!)

I actually recommend using either the Wizard's calculator written by J.B. (on the WOO site), or the composition-dependent calculator on BJstrat.com. The bjstrat calculator has an advantage over the WOO calculator: it shows the probabilities for the dealers outcomes (for getting 17, 18, 19, 20, 21 or Busting) before the player draws anymore cards. However, the WOO calculator does show more digits of accuracy.

If you are determined to try an excel version of a composition-dependent BJ hand calculator:

1. Recognize that it is actually ten separate calculators: one each for each of the dealer's possible face up cards.
2. I would suggest starting first with the "vs 9" or "vs 10": calculators because I think they are the easiest (beware of the fact that dealer checks for an ace with a Ten showing in the US version of BJ, which is information that needs to be accounted for.)

There are many good programmers on this forum, so I think I'll sit back and let some of the other experts offer some advice about the approach. Maybe I'll learn something too.
Last edited by: gordonm888 on May 5, 2020
So many better men, a few of them friends, were dead. And a thousand thousand slimy things lived on, and so did I.
charliepatrick
charliepatrick
Joined: Jun 17, 2011
  • Threads: 32
  • Posts: 2114
May 5th, 2020 at 8:58:25 PM permalink
I vaguely remember trying this out and, as Gordon has said, it's easier to start with vs 10 and vs 9. Personally I worked out all the possible hands a player could make (initially ignoring splits) and then worked out their chances against the dealer's upcard. I then looked at whether each hand should draw, stand or double. (As I'm based in the UK I initially didn't look at peeking, but that's a factor you need to consider, e.g. when looking at the probabilities when drawing the next card.)

I didn't get this far but for splitting I guess you'd then look at say (8-8) and rework the calculation assuming a number of 8's (i.e. resplits) have already been removed from the deck. I doubt if anyone then looks at the second hand being split knowing what cards the first hand used! Spltting Aces is obviously easier as they can either get another Ace (which resplits) or one card only.
LearningCurve
LearningCurve
Joined: May 5, 2020
  • Threads: 1
  • Posts: 2
May 6th, 2020 at 9:40:03 AM permalink
Hi gordonm88,

Thanks a lot for the reply!

I have jumped into the website great resource, it's hard to believe that the developer doesn't play blackjack!

I am more just curious to see if having a composition dependent tool will positively affect my overall return on the shoe.

In the possiblywrong blog post about distribution-and-variance-in-blackjack-part-2 it's interesting that the author notes that optimal strategy (CD) has a larger variance at higher EVs than basic strategy but it appears that optimal strategy in general has a better EV range (-6% to 10%) vs basic strategy ( -12% to 6%)
gordonm888
gordonm888
Joined: Feb 18, 2015
  • Threads: 38
  • Posts: 2567
Thanks for this post from:
charliepatrick
May 6th, 2020 at 3:49:18 PM permalink
Quote: charliepatrick

I vaguely remember trying this out and, as Gordon has said, it's easier to start with vs 10 and vs 9. Personally I worked out all the possible hands a player could make (initially ignoring splits) and then worked out their chances against the dealer's upcard. I then looked at whether each hand should draw, stand or double. (As I'm based in the UK I initially didn't look at peeking, but that's a factor you need to consider, e.g. when looking at the probabilities when drawing the next card.)

I didn't get this far but for splitting I guess you'd then look at say (8-8) and rework the calculation assuming a number of 8's (i.e. resplits) have already been removed from the deck. I doubt if anyone then looks at the second hand being split knowing what cards the first hand used! Spltting Aces is obviously easier as they can either get another Ace (which resplits) or one card only.



Yes, I have handled splits just as you surmise. There is a recurrence formula you can use to account for the first three splits but not splitting the fourth time, and then basically use the same outcome probabilities for each of the hands resulting form the resplit (with the eights removed.).

I do rigorously write out all the hands for splitting AA pair only once (the usual rule) -there are only 169 of them.

I have also written out all the player hands for splitting (once) a 99 pair and an 8-8 pair. With DAS and vs 2-6, this is easier than you might think (even when trying to make hit/stand trade-offs on post-split hard 12 and 13 vs 2-6.) Of course, splitting pairs multiple times is very hard to analyze without making recurrence approximations.

In general, introducing lots of extra equations to analyze split pairs is really hard to justify. The pair split decisions that are most susceptible to be changed by shoe composition are 22 and 33 pairs versus small dealer upcards - and those are also the hardest to analyze rigorously.

In fact, I think all of this composition dependent math is pretty unimportant for 6 deck and 8 deck shoes, other than the trivial cases of Insurance and 16 vs 10. The glamor of "composition dependent analysis" was associated with single deck BJ played to a deep penetration - a game that has pretty much gone the way of the dodo bird.
So many better men, a few of them friends, were dead. And a thousand thousand slimy things lived on, and so did I.

  • Jump to: