CVCX vs CVData results

[fwb]

When analyzing a new table I've come across, I usually start out in CVCX to get an optimal betting ramp for a rule set. Then I put that strategy into CVData, but for the life of me I cannot get the outcomes to come out the same. Usually CVCX shows better win rates with roughly the same std dev. What could I be doing wrong? Are there differences between the two sims I'm missing here? I like playing around in CVData for the more advanced features after I get a betting ramp from CVCX to start with, but even if I make sure to set up all the rules identically to begin with, the results are always notably different.

Example:
In CVCX, I put in the following rules: Single deck H17 DAS, 3 players, 50% pen, split to 4 hands, Backcount 0, spread $25 to $100, play two hands, min chip size $25. Use Hi Opt II, illus. 18 indices from -1 to 99, with half deck resolution (rounded) and sidecount aces. This shows me an optimal betting ramp (starting at 1) of: 25,25,50,50,75,75,100,75,100.
Winrate shows $110.13/hr, Std Dev/hr: $791.34, SCORE: 193.70

In CVData, I put in the exact same rules and betting strategy and get the following results:
Winrate: $80.52/hr, Std Dev/hr: $734.82 , SCORE: 120.8

Thanks for any help...

 

[PrinceDragon]

You didn't do anything wrong.It's just the way that Norm design his software.

In CVCX:WR=Ave.bet *IBA*Hds/Hr

In CVData:WR=Total.bet/Hr*TBA

CVData's result are more accurate and realistic,personally,I just add those 2 numbers up and divide by 2 to get my estimate WR (maybe I shouldn't)

Hope this helps

P.D.

 

[QFIT]

CVCX and CVData have exactly the same accuracy. Something is different. I would need to see the sims.

 

[MJ1]

I believe CVCX only provides an estimate when it comes to giving performance of backcounting strategies, especially when the strategy calls for playing 2 hands.

In this case, CVData yields a more accurate answer as it does not rely on post sim calculations which assume TC freq. remain unaltered when a player wongs in.

Note: For play-all games, CVCX and CVData will provide identical results irrespective of how many hands you play.

Feel free to report back when you guys figure out the root of the problem.

 

[fwb]

CVCX and CVData have exactly the same accuracy. Something is different. I would need to see the sims.

Here is a zip archive with a doc of screenshots of my configuration and results, along with HST exports of both sims: (Dead link: http://www.yousendit.com/download/Z01NeFlVdkdRR2RjR0E9PQ)
(link expires in 7 days)

Anyone get the same results? This time I tried enabling "Remove seat effect" but it made the gap even wider.

 

[QFIT]

Ahh. As MJ1 said, for backcounting CVCX assumes that a player leaves when you sit. This gives a different answer. It is not possible post-sim to calculate accurately the effect if the number of hands changes. That requires CVData.

 

[fwb]

Hmm, good to know. But I still can't seem to get consistent results, even with no backcounting/wonging. It actually made the results even less similar. Can anyone let me know if they are seeing the same thing when they take a CVCX generated betting strategy and throw it into CVData?

 

[QFIT]

CVCX and CVData give identical results when no backcounting or multiple-seats are used. CVCX must estimate in those cases while CVData gives actual results. Send me the sims and I'll look.

 

[MJ1]

CVCX and CVData give identical results when no backcounting or multiple-seats are used. CVCX must estimate in those cases while CVData gives actual results. Send me the sims and I'll look.

QFIT, do TC Freq have anything to do with optimal bet derivation? If so, how can CVCX provide an accurate optimal bet schedule if it assumes that TC Freq remain fixed when backcounting is set?

Off the top of my head, I believe CVCX provides an optimal bet schedule that is suppose to maximize SCORE. SCORE is based upon SD and WR. SD and WR are related to TC Freq which tell us how often we place various sized wagers. If the software uses incorrect TC Freq to determine SD and WR, how on earth could it possibly provide an optimal bet schedule if the SCORE being maximized isn't even the correct SCORE in the first place?

All I am trying to convey here is that there is most likely a bet schedule that can produce a higher SCORE for wongers (maybe as much as 10%-20%). How to figure that out post sim, well I have no idea. Do you think this is an area worth exploring?

MJ

 

[QFIT]

Yes, optimal betting ramp calcs takes into account TC frequencies. And yes post-sim calculators like CVCX and BJRM cannot correctly determine those frequencies. Fortunately, backcounting optimal betting strategies are far less sensitive to TC freqs than play-all.

Switching from single-hand to two hands at a specified count is far more sensitive. However, in this case CVCX is able to estimate the effect on TC freqs and adjusts accordingly.

It should also be noted that the effect on TC freqs. very much depends on the number of players.

 

[MJ1]

Fortunately, backcounting optimal betting strategies are far less sensitive to TC freqs than play-all.

Thanks, I was not aware of this. But if you had to take an educated guess, by roughly what % do you think SCORE could be enhanced if the proper TC Freq were used for the backcounting optimal bet schedule? 1-3%? 4-6%?

Switching from single-hand to two hands at a specified count is far more sensitive. However, in this case CVCX is able to estimate the effect on TC freqs and adjusts accordingly.

Right. I tested this feature a while back when I compared the output to that given by CVData. Results were virtually identical. Nice job.

Do you plan on extending this feature to KO? It would be a worthwhile improvement.

It should also be noted that the effect on TC freqs. very much depends on the number of players.

For switching back and forth between 1 and 2 hands or wonging?

MJ

 

[QFIT]

Thanks, I was not aware of this. But if you had to take an educated guess, by roughly what % do you think SCORE could be enhanced if the proper TC Freq were used for the backcounting optimal bet schedule? 1-3%? 4-6%?

Sorry, no idea. I developed CVCX by chance. BJRM already existed. I had no intention of repeating something that someone else had done. But, I had this neat idea of a parallel simulator, something that had never been done before in BJ. A program that could accurately run thousands of different sims at once and instantly switch between them and chart the differences. I could not resist the concept and built it. And I am glad that I did as it has enormously aided my research and later created Chapter Ten of BJA. But it has the same flaws as BJRM when it comes to backcounting.

There are so many variables when it comes to backcounting. Not only is it affected greatly by the number of players, but if people enter and leave, which they will do in a game that allows backcounting, that also changes results.

Right. I tested this feature a while back when I compared the output to that given by CVData. Results were virtually identical. Nice job.

Do you plan on extending this feature to KO? It would be a worthwhile improvement.

Wonderful. I am really happy to hear this. This particular feature drove me nuts for a month and I had to restart the project a couple of times. But, I cannot figure out any method of applying this in any manner to unbalanced strategies.

For switching back and forth between 1 and 2 hands or wonging?

MJ

For both. The method I use in CVCX for switching between one and two hands takes into account the number of players. That multiplied the problem substantially. But I wouldn't release the product until I figured out how to do this. In my mind, this was a major breakthrough. But, I just listed it as another feature.

 

[fwb]

CVCX and CVData give identical results when no backcounting or multiple-seats are used. CVCX must estimate in those cases while CVData gives actual results. Send me the sims and I'll look.

Please take a look at the link I posted earlier, is that what you need?

Here is a zip archive with a doc of screenshots of my configuration and results, along with HST exports of both sims: (Dead link: http://www.yousendit.com/download/Z01NeFlVdkdRR2RjR0E9PQ)
(link expires in 7 days)

Anyone get the same results? This time I tried enabling "Remove seat effect" but it made the gap even wider.

 

[QFIT]

Yes, it has backcounting.

 

[fwb]

But if I try that exact setup without backcounting/wonging, it still gives me significantly different results. Do you see any reason why?

I also tried using only one hand, which made the results very close, but not exact. Winrate was about the same but std dev was off by about $100.

 

[QFIT]

I would need to see the sims EXACTLY as you ran them. There are huge number of variables. In CVData, from the menu select Tools then Export and e-mail the export file. This includes all settings.

 

[fwb]

edit: nvm

 

[QFIT]

The HST file is the sim output file. The file I am asking for is the input, the configuration and all strategy files. You create this by clicking on Tools then Export.

 

[fwb]

Ahh I think I found my mistake, I didn't change my betting ramp in CVData when changing to not to wong out. I didn't realize how much your optimal betting ramp changes between wonging out or not, interesting. Now I'm seeing nearly identical results. Std dev is still off negligibly by about $20, which I'm assuming is due to small details CVData accounts for which CVCX does not. Which made me curious, how does CVCX treat at the following options which you can specify in CVData?

-Remove seat effect
-Cards delt face down (CVCX assumes all cards face up?)
-True count divisor (cards delt/seen/in tray)
-True count division and deck estimation (CVCX just lists one "True count calculation")
-Count recalculation points
-Other players' playing strategies

 

[QFIT]

-Remove seat effect

CVCX assumes that you sit at third base. CVData allows you to specify the seat. Remove seat effect causes you to keep switching seats to average out the results.

-Cards dealt face down (CVCX assumes all cards face up?)

CVCX assume face down for SD and DD, up for shoes. In CVData you specify.

-True count divisor (cards delt/seen/in tray)

CVCX assumes 'in tray'

-True count division and deck estimation (CVCX just lists one "True count calculation")

CVCX allows most of the CVData options.

-Count recalculation points

Where do you see this?

-Other players' playing strategies

CVCX assumes a typical player strategy. This has almost no effect.