racquet
• Posts: 411
Joined: Dec 31, 2014
March 8th, 2017 at 7:04:01 PM permalink
Is this the proper place to post a question/comment on the Blackjack Card Counting Game on Wizard of Odds? If so, it looks to me as if there are some bugs in the software when warning on strategy errors. I'm not sure if it applies any index plays based on the true count, but for sure telling me that hitting a 13 against an 8 when the count is +1 can't be good advice. There are other obvious errors, and at some point the "analyze hands" percentages or odds are easily wrong. I've even seen where ALL the options are reported with the same result - all zeros.

I'd like to say that it seems like the dealer gets far too many blackjacks, even on a negative count. But that's probably providing the same level of real world outcomes that we all see and bitch about all the time. But still...
JB
• Posts: 2089
Joined: Oct 14, 2009
March 8th, 2017 at 8:27:26 PM permalink
Quote: racquet

hitting a 13 against an 8 when the count is +1 can't be good advice

Hitting 13 vs. 8 is basic strategy. There isn't even an index play for it in the I18.

The advice offered in the game is always based on the perfect analysis considering the current state of the shoe at that time.

There are, however, times when communications glitches occur. Behind the scenes, there is a chain of communications that occurs to perform the analysis. If something goes wrong in this chain of communications, then the analysis will show all zeroes or will say "Analysis error". If you get non-zero results, you can be sure it is showing the correct analysis.
Wizard
• Posts: 26712
Joined: Oct 14, 2009
March 8th, 2017 at 8:42:11 PM permalink
Quote: racquet

but for sure telling me that hitting a 13 against an 8 when the count is +1 can't be good advice.

For sure, that is correct advice.
"For with much wisdom comes much sorrow." -- Ecclesiastes 1:18 (NIV)
racquet
• Posts: 411
Joined: Dec 31, 2014
March 12th, 2017 at 4:22:06 PM permalink
I misstated the earlier example - but whatever it was, it was wrong.

Here's one I have on my screen right now: Dealer shows a 2, player has 2-3. True count is -1. The game says: Hitting is not the best play. Analysis says: Surrender: -0.50000, Stand: +0.639719, Hit: -0.642205.

So I did not take the advice, and insisted on hitting. I got a 9, so my hand is now 14 against the dealer 2. I select stand and the games says: Standing is not the best play. Analysis says: Stand: -0.336923 Hit = 0.318846.

I have also seen the game reporting an analysis where all the options are 0.000000.

I'm using the game to hone my skills with count and I18 variations. So "false-positive" strategy errors are annoying and distracting. They are occurring often enough and in such obvious situations that I can't think I am the first person to notice and report them. In this case I haven't played three shoes and see these errors.

Perhaps some registers are not getting cleared when the end of the shoe is reached?
racquet
• Posts: 411
Joined: Dec 31, 2014
March 12th, 2017 at 4:52:54 PM permalink
Sorry to be a pest, but...

How can a running count of +5 and "decks remaining" of 1.7 yield a true count of +1? 5 divided by 1.7 = 2.94.

I don't know if the running count or decks remaining is accurate, as I am not keeping the count in my head and don't have a shoe or a discard rack to use as a guide, but I assume that what the program displays is in fact the true state of things. Even with a "by-eye" approximation that you would have to use in the field, TC 5 divided by 2 decks left = 2.5.

Are there other simulation packages out there, even available for sale as opposed to this free one?
Wizard
• Posts: 26712
Joined: Oct 14, 2009
March 12th, 2017 at 4:56:12 PM permalink
Quote: racquet

I misstated the earlier example - but whatever it was, it was wrong.

Here's one I have on my screen right now: Dealer shows a 2, player has 2-3. True count is -1. The game says: Hitting is not the best play. Analysis says: Surrender: -0.50000, Stand: +0.639719, Hit: -0.642205.

Do you have a screenshot?
"For with much wisdom comes much sorrow." -- Ecclesiastes 1:18 (NIV)
racquet
• Posts: 411
Joined: Dec 31, 2014
March 13th, 2017 at 3:56:36 PM permalink
I'm not sure how to do a screenshot. If I do <ctrl>PrtScrn I get my whole desktop display. I can then drop that into Word and crop away the parts of the desktop that aren't the window of the simulator, but I can't figure out how to get that into this reply. Do I attach a file somehow? I have seen images from people's phones here, but not screenshots. I'd take a picture of the screen, but then there's some skill in getting that into this reply that must be just as foreign to me. I'll keep trying - but give me some clues? Thanks.
racquet
• Posts: 411
Joined: Dec 31, 2014
March 13th, 2017 at 4:27:42 PM permalink
Never mind. I did a little searching here and found some directions. Here's an example where the running count is +7 with 2.7 decks left. 7 / 2.7 = 2.6. Even assuming a by-eye approximation as you would be doing in real time, a true count of 7 divided by 3 decks remaining is at least 2. I did not keep track myself to see if the true count is correct, and of the cards played to see if "decks left" was correct. I adjusted the rules to 6deck, S17 but other than that all the rules are the default. I think I see the counts adjust as individual cards are played, but in this case the cards in this round net to a running count of 0, so nothing has changed as of the start of dealing this round.

racquet
• Posts: 411
Joined: Dec 31, 2014
March 13th, 2017 at 5:27:58 PM permalink
Here's an example of what I think is an incorrect "best play". Stand a 16 against a dealer 8? This can't be an index play far down the list from the I18. A TC of -3 would seem to me, if anything, maybe say to HIT the sixteen INSTEAD of stand, since there are more low cards out there, and therefore a better chance of not busting. Oh yeah -- the TC is wrong here as well -- running count of -12 decks left of 3.1 - TC = -3.87. Using the by-eye real-world method: -12 / 3 = -4.

For some reason the IMG code does not work here like it did in my previous post. So here's the link:

http://imgur.com/Zv5EAOd
beachbumbabs
• Posts: 14268
Joined: May 21, 2013
March 14th, 2017 at 4:56:06 AM permalink
Quote: racquet

Here's an example of what I think is an incorrect "best play". Stand a 16 against a dealer 8? This can't be an index play far down the list from the I18. A TC of -3 would seem to me, if anything, maybe say to HIT the sixteen INSTEAD of stand, since there are more low cards out there, and therefore a better chance of not busting. Oh yeah -- the TC is wrong here as well -- running count of -12 decks left of 3.1 - TC = -3.87. Using the by-eye real-world method: -12 / 3 = -4.

For some reason the IMG code does not work here like it did in my previous post. So here's the link:

http://imgur.com/Zv5EAOd

I added .jpg to your IMG and it posted. No idea why it needed that, but it worked.
If the House lost every hand, they wouldn't deal the game.
JB
• Posts: 2089
Joined: Oct 14, 2009
March 14th, 2017 at 9:29:59 AM permalink
Thanks for the screenshots, I will investigate. Clearly something isn't right.
JB
• Posts: 2089
Joined: Oct 14, 2009
May 24th, 2017 at 4:38:15 AM permalink
The intermittent problems with the strategy advice have finally been resolved.

Previously, it was not uncommon to get incorrect results or an "Analysis Error" for the scenario being analyzed.

What was happening was the game would frequently show you results from somebody else's analysis due to a design flaw. So the results were being calculated correctly, but they weren't necessarily for the scenario you analyzed. "Analysis Error" would appear when two simultaneous calculations collided, causing the calculator application to crash on the server.

Each calculation is now properly isolated, so you only see the results for your scenario, as intended.

The blackjack hand calculator is implicitly fixed as well, since it relies on the same application as the card counting game.

I'm still not sure what caused the problem with the running count depicted in the screenshot on page 1, as the calculations (in the code) appear to be correct, and I have not been able to reproduce it. The only thing I can think of is if the initial running count in the options was changed to something other than zero, in which case things won't appear to add up.