Adding AA78mTc side count to High Low

[seriousplayer]

The books are only $3 or $4 if you buy it online. I am not making any money from my books and I am not trying to pedal my books. I have published this information as a courtesy at no charge. If you do not like it then do not use it. I never asked anyone to buy my books - it is your option. If they want more information then they can spend $3 or $4 for my book. And if you do not want to buy the books then don't. Buy a $4 cup of coffee at Starbucks instead.

$3 or $4 for your book is too much. You are trying to use other people covered published work to persuade other people of your own work. You have no prove or simulations. That is taking action that is ineffectual and does not lead to progress. You have been doing that the whole thread posting useless information post after post. Yes, you are trying to pedal your own book. I am not going to throw $3 or $4 aways.

I mentioned that I taught Carla, my friend who is not that bright and just of average intelligence, nothing special and who knew nothing about counting, the KO with AA89mTc and in less than 2 weeks she was 100% efficient in keeping both counts (I checked her on both counts) and she keeps both counts in her head. These are +/- counts. You are keeping only two integers in your head. And with the KO, estimation of decks remaining is not that important as it is with the HL.

All I can tell you is that Carla and I have been using KO with AA89mTc for over 4 years now. We have no problem with the counts which can be played for hours on end with no exhaustion and no errors.

Fairytale!

KO with AA89mTc is not ridiculous. What is "ridiculous" is keeping the level 2 HO2 with ASC where you have to estimate decks remaining very accuracately and this is very hard to do which is why not many people use it and stick with the HL

What is “ridiculous”? As I mention keeping both a secondary count of AA89mTc to improve playing and in addition to keeping both a secondary count of 45m79c or 5m7c to improve betting. One or the other secondary counts should improve both playing and betting but not both. As I already mention in post # 162.

I did not give optimal bet spread. I just tried to explain an reasonable betting approach. My original post had nothing to do with betting.

If that is the case than you are giving nonsense information. I think you should give people who brought you book a refund.

[bjanalyst]

I really see no need to question my 4 years of experience with Carla in successfully using KO with AA89mTc by calling it a Fairytale. That is implying that I am lying about 4 years of actually casino use. If you doubt my work that is one thing but doubting my actually experiences or what I am saying is another and also is kind of insulting.

And everyone builds on what others wrote. The saying goes on the "shoulders of giants'. Newton improved on Galileo and Einstein improved on Newton. No one starts from scratch.

So I used information that was published and then I did my own analysis using that information in ways that was not done before.

And I told you that another user has agreed to do simulations so just be patient.

And I suggested just keeping the KO with AA89mTc or HL with AA78mTc which is what this post was all about to begin with. Because others asked about betting efficiency is the only reason I brought in the 5m7c and 45m79c. That should not have been part of this post.

The betting efficiency of the KO is 96.5% for S17, DAS, LS which is really quite good. That is why I decided to concentrate on increasing playing efficiency in this post which was suppose to be about only the HL with AA78mTc or KO with AA89mTc.

And you wanted one side count that helps with both playing efficiency and betting efficiency. If you just used the 45m79c or 5m7c only, your betting efficiency is about 99% and there is some increase in playing efficiency as well such as hit/stand and surrender hard 15 and hard 16 but not as much increase in playing efficiency as AA78mTc or AA89mTc has. So you have a choice, increase betting efficiency to 99% with 5m7c or 45m79c or no increase in betting efficiency but a much greater increase in playing efficiency using AA78mTc or AA89mTc.

But wait, you don't believe any of my analysis anyhow and categorize everything I have said and done as a Fairytale.

That is why I am anxious to get these simulation results as that is the only way it seems that what I have said or done will be accepted or not.

[seriousplayer]

And everyone builds on what others wrote. The saying goes on the "shoulders of giants'. Newton improved on Galileo and Einstein improved on Newton. No one starts from scratch.

So I used information that was published and then I did my own analysis using that information in ways that was not done before.

Wrong!! I already done a similar study myself but I did it with a level 3 count. I have exhaustive simulations myself to back that up. What you are doing is not new!!

[Three]

You have no prove or simulations.

In a quantitate sense yes.

That is taking action that is ineffectual and does not lead to progress.

No. His theory is solid. Whether it is worth it is a matter of opinion. He just worked the problem backwards. He had his count and figured the side count to make perfect insurance. He should have started at a strong insurance count and worked backwards to pull it apart into the strongest component counts for betting and playing decisions, while realizing it can be various multiples of one count added to another to get any decision count for bets or plays. The component counts themselves don't need to be strong counts for anything. It is the combined counts that you use for making decisions that need to be strong. You are trying to create the strongest family of combined counts to cover betting and the highest correlated counts to playing EoRs for your most frequent plays, especially positive that you will have a larger bet out on average when the deviation is made or have an index near TC 0. Betting strength is a top priority. You can stick with a level one count if that is your preference but that makes for less accurate betting even if it has a high BC. The low cards have a big discrepancy in betting EoRs.

Acting like ignoring that won't cost a lot in betting accuracy because the BC is about the same shows a lack of understanding of what you are doing. The high BC means your relative bin averages can be expected to be accurate but says nothing about how wide the range of actual deck composition advantages is around that average. Betting accurately is about having a much narrower range there, or more accurately a smaller SD for the collection of advantage for the data points that the sim collects in the betting bin. If you use a level 1 count your betting bins will either be TC bins or fractional TC bins. Multi-level counts may combine more than one TC bin to get a betting bin. If both use integer TC betting bins, the multi-level counts get added accuracy benefit from having a lower frequency for the corresponding TC bin since the data is stretched out across about twice the TC range.

A level 1 count can do the same with integer TCs by dividing by the number of half decks remaining rather than the number of full decks remaining to get the TC. That will stretch the data across twice the TC range and should lower the SD for the integer TC betting bins which should increase betting accuracy some. But it doesn't fix ignoring the wide range of betting EoRs for the low cards. It just levels the field concerning how much you spread out the data across TCs. Like for a floored TC of +4 will now be broken up into TC +8 and +9. Counting all the low cards the same still has its inherent inaccuracy, but the higher frequency end goes into TC +8 and lower frequency end goes to TC +9. This allows you to make a smoother ramp with more betting bins that doesn't have as big a potential for having to jump bets too much causing you to make a decision of whether to cut EV by not making the required jump for heat reasons or possibly draw attention from big bet jumps. The higher flexibility for making more bets using integer TC betting bins will increase SCORE (decrease n0). So a level one count dividing by the number of half decks remaining when calculating the TC gets some the extra a level 2 count has, but no matter what you do the loss in accuracy from treating all the low cards the same will affect the level one counter some. Of course the other option is to go to one decimal accuracy accuracy for TCs and dividing into fractional TC bins.

Sorry. I got side tracked there by what you need to consider when working the problem back from the strong insurance count to component counts.

[bjanalyst]

Thanks for your comments.

If you keep KO with 5m7c your betting efficiency for S17, DAS, LS game is 99.0% as compared with KO betting efficiency of 96.5% and there is also some increase in playing efficiencies over using just the KO.

When adding AA89mTc to the KO there is a large increase in playing efficiency but no increase in betting efficiency.

I have calculated a table for KO + k1*(5m7c) + k2*(AA89mTc) where you now have betting efficiency of 99% and there is also additional increases in playing efficiency over the KO with AA89mTc. But I am not going to even touch on that now as there is enough controversy with KO with AA89mTc or HL with AA78mTc.

Thanks for not bashing my calculations but instead giving constructive suggestions.

I hope that other users on this website will refrain from bashing HL with AA78mTc until the simulations are done.

[seriousplayer]

So a level one count dividing by the number of half decks remaining when calculating the TC gets some the extra a level 2 count has, but no matter what you do the loss in accuracy from treating all the low cards the same will affect the level one counter some. Of course the other option is to go to one decimal accuracy accuracy for TCs and dividing into fractional TC bins.

True, I found this out by simulating a version of Hi-lo with side count of 7s. CVDATA is not good at generating decimal true counts indices which would affect the simulation results.

[Gronbog]

When Gronbog contacts me about the simulations I will ask him what he wants to simulate either

(1) just insurance and hard 12 v 2, 3, 4, 5, and 6

(2) Just the 12 Illustrious 18 that AA78mTc helps the HL with

(3) The more complete strategy changes listed in the tables below.

I generally like to know what the improvement is for each level of extra effort, so I would suggest simulating all of the above.

It's the holiday season and I'm busy spending time with family and friends. The planning of the simulations will begin in the new year.

[Stealth]

bjanalyst,

Your professed advantages are without the validation of simulations as has been stated and recognized. The player who has volunteered to do your sim is a well-respected member that is higly competent in all things technical blackjack. Take advantage of his offer and listen carefully to the results.

One of the major impacts to the value of your suggestions will be frequency of occurance. To oversimplify an example, the combined frequency of 12 v 2, 3, 4, 5, 6 is less than 3.35% in a six deck H17 Das game. So, if a large part of your improvement comes from these plays it will only impact ~3% of the hands played! Further understanding of these occurances at the positive true counts will add to your enlightenment. The impact at negative true with small bets is not the same as the impact at a positive true with a much larger bet.

A lot of this data is available on qfit . com at https://www.card-counting.com/blackjack-data.htm

This data is not meant to replace the needed sim but it may give you some insights that you seem to need.

[Gronbog]

My question is : Are those "Optimal Beting Spreads"?

I did not give optimal bet spreads. I just tried to explain an reasonable betting approach.

Part of the output of the simulations will be a spreadsheet containing the frequency, EV and variance at each true count from which the optimal Kelly bets will be calculated. The spreadsheet behaves similarly to the output of CVCX, so bjanalyst's suggested betting ramp can be input for comparison against the optimal bets. The (c-)SCORE, Win Rate, standard deviation, RoR and N0 are updated as the betting ramp is changed, in a way similar to CVCX.

[seriousplayer]

Part of the output of the simulations will be a spreadsheet containing the frequency, EV and variance at each true count from which the optimal Kelly bets will be calculated. The spreadsheet behaves similarly to the output of CVCX, so bjanalyst's suggested betting ramp can be input for comparison against the optimal bets. The (c-)SCORE, Win Rate, standard deviation, RoR and N0 are updated as the betting ramp is changed, in a way similar to CVCX.

Gronbog,

I acknowledge you for doing the simulation for two different system variants. One variant with 45m79c and the other with 5m7c. Then the AA79mTc with Hi-lo and AA89mTc with KO. Well, both variants of the 45m79c and 5m7c can be done using CVDATA without modifying the software itself but the AA79mTc with Hi-lo and AA89mTc with KO needs modification in CVDATA.

[Three]

betting efficiency

I am not trying to nit pick, but BC is a correlation not an efficiency. Mathematically there is a big difference. I have already explained that in this thread so I won't do it again. Betting efficiency would be more like the EV of a betting system using BS for its optimal bets for a set spread for a standard RoR, and BR compared to computer perfect betting EV for BS. There are no standards for measuring BE because nobody defined them. System efficiency would be EV for optimal bets with the systems playing deviations. I am not aware of any standards for any of these yet to be defined stats. But you can compare things by setting a constant for BR, RoR, and spread for the counts being compared and the game they are being considered for. The supposition about BE standards are getting into a grey area, but the concept of the differences between BC and what BE would be is correct.

So BC is how well the count tags correlate to the full deck betting EoRs. While BE would be the performance of the count system for a set standard spread, RoR, Basic Strategy, and BR using the count system's optimal bets for these constants. I feel using indices would give you a system efficiency rather than a betting efficiency. What ends up happening is you get to bet more optimally if you have a stronger betting efficiency. System efficiency is what is relevant to APs.

[lij45o6]

Lemme jump in concerning the secondary 4579SC: If one is to do this, then why, oh why, would you use High Low? Simplify your system to a similar system!

High Low[A-T]: {-1, 1, 1, 1, 1, 1, 0, 0, 0, -1}
4579SC[A-T]: {0, 0, 0, 1, 1, 0, -1, -1, 0}

Combined[A-T]: {-1, 1 , 1, 2, 2, 1, -1, 0, -1, -1}: 0

While the above is balanced, I would recommend going to a better Level 2 system:

Highly Optimal 2[A-T]: {0, 1, 1, 2, 2, 1, 1, 0 , 0, -2} (use an ASC or a BASC to compute betting)
Zen[A-T]: {-1, 1, 1, 2, 2, 2, 1, 0, 0, -2} (ASC to adjust for Ins and playing in 1/2D games, +3D use Zen only)

Think of brevity and simplicity.

[BJGenius007]

Lemme jump in concerning the secondary 4579SC: If one is to do this, then why, oh why, would you use High Low? Simplify your system to a similar system!

High Low[A-T]: {-1, 1, 1, 1, 1, 1, 0, 0, 0, -1}
4579SC[A-T]: {0, 0, 0, 1, 1, 0, -1, -1, 0}

Combined[A-T]: {-1, 1 , 1, 2, 2, 1, -1, 0, -1, -1}: 0

While the above is balanced, I would recommend going to a better Level 2 system:

Highly Optimal 2[A-T]: {0, 1, 1, 2, 2, 1, 1, 0 , 0, -2} (use an ASC or a BASC to compute betting)
Zen[A-T]: {-1, 1, 1, 2, 2, 2, 1, 0, 0, -2} (ASC to adjust for Ins and playing in 1/2D games, +3D use Zen only)

Think of brevity and simplicity.

The reason BJanalyst is using the complicate new system is because he has two choices to play different scenarios. In most scenarios, he just plays Hi-Lo. But for 12 v 2-6, insurance and Lucky Lady, he uses the combined system (HiLo + 89aceaceFaces). But as Three pointed out, the obvious drawback is that this proposed new system doesn't improve on most common hands like 16 v 7-A, 15 v 7-A etc.