Memory and Counting

[JesterCW]

I'm starting to build a person-action-object system and translate the deviations I already know into a memory palace. The idea/goal will be to have a different palace for each set of major rules that I come across (H17/S17, DAS, etcetc.) and then be able to adapt deviations quickly to the new set of rules. I'm also curious if the associations will come up in counting and help numbers "stick". I feel like when I'm practicing my counting and the speed gets kicked up a notch sometimes I start to drop the "N (negative)" in my head and just flow until things settle down again.

I was curious if anybody here uses and enjoys using memory systems like this (or the Major System / other memory technique). Maybe this is overkill? Just thinking out loud tonight!

Excited to get out on one of my first actual casino trips as an AP later this month. I've played as a ploppy, I've played knowing a little but not being able to keep up with the count, and now I'm just working on my bankroll and being smart around managing it alongside my full-time work. Meanwhile though, going down the rabbit hole on memory stuff...

[Archvaldor]

The issue with memory palaces is not so much the difficulty in acquring the skill-which most people can with practice-it is how to process the information usefully.

A simple level-1 count system does not require an advanced memory or indeed even average intelligence. That will get most of the game in most beatable games that exist.

If you can find a deeply dealt single-deck game then first I want to know how your time machine works, and second you might want to look into the various attempts people have made to create multi-parameter systems. But probably not.

The main use for a memory palace would be advanced card sequencing-type systems, where you try and model the movements of cards through the shuffle in your head. Very difficult to pull of in practice though you could be looking at very high returns if you could.

[Midnight]

I'm starting to build a person-action-object system and translate the deviations I already know into a memory palace. The idea/goal will be to have a different palace for each set of major rules that I come across (H17/S17, DAS, etcetc.) and then be able to adapt deviations quickly to the new set of rules. I'm also curious if the associations will come up in counting and help numbers "stick". I feel like when I'm practicing my counting and the speed gets kicked up a notch sometimes I start to drop the "N (negative)" in my head and just flow until things settle down again.

I was curious if anybody here uses and enjoys using memory systems like this (or the Major System / other memory technique). Maybe this is overkill? Just thinking out loud tonight!

Excited to get out on one of my first actual casino trips as an AP later this month. I've played as a ploppy, I've played knowing a little but not being able to keep up with the count, and now I'm just working on my bankroll and being smart around managing it alongside my full-time work. Meanwhile though, going down the rabbit hole on memory stuff...

I have been working on a new card counting system for the last 8 months that utilizes this very approach. My system is a multi-parameter, multi-level "count" that seeks to maximize all three metrics of counting system performance (BC, PE, IC). I have dubbed the system TOR as I was inspired to create it based on one of the legendary players here who had a similar albeit more complex approach. In my TOR count, I track the card groups {2,3},{4,5},{6,7},{8,9}{Tens}{Ace}. So instead of a single running count like RC 12 in Hi lo, a TOR count DC (deck composition) count would look something like 24-67-93 which is 2 {2,3} removed, 4 {4,5} removed, 6 {6,7} removed, 7 {8,9} removed, 9 {tens} removed, 3 {Aces} removed. This is not a "count" per se, but a map of the composition of the discards. The running counts (and also individual play counts) are all derivatives of this DC count. I call it the swiss army knife system because it allows a player to move between different games and optimize everything for the exact games and rules she encounters (Blackjack, Spanish 21 or other variants). But I digress. Why I had to resort to PAO tables is because, as you can imagine, when practicing with it I was finding it it difficult to remember three different two digit numbers from round to round. So I employed a 100 value PAO table that solved that problem. I assign a person-action-object for each number form 00 to 99 and this allows me to store the number in memory as an image instead. This has worked extremely well and I no longer have any trouble with the DC count from round to round. In fact, it worked so well that I then started using it to remember whole new sets of indices for the various play counts that I came up with. In my approach I store both the play count for a particular rule set for H17 and S17 in a single two digit number. As an example, my PAO object for the index for 16 V T is Santa Claus (Santa is represented by 00 in my table) which means the index for H17 is 0 and S17 is also 0). But my PAO object for 11 v A is Batman who is assigned 02 in my table. This means that the H17 index is 0 and the S17 index is 2. So when I am learning new indices I just associate the particular index with the image which is way, way easier. I recently put this to the test and was able to learn an entire set of new indices is just 2 days with not much effort. I have not seen the need to use memory palaces in anyway for card counting, but have begun some preliminary research into adapting this method for shuffle tracking and Ace sequencing that would need clean memory palaces inorder to work. I was able to impress a family member a week ago by perfectly recalling the exact sequence of a shuffled deck of cards after just running through them briefly in about three minutes, lol. I did this by using my PAO table and putting each PAO triplet in a separate location in one of my memory palaces. This allowed me to store the full deck in the palace with just 9 images and recall them by walking through the palace in reverse order. They think I'm a magician, lol. But all of these are still in research and development phase. I often tell persons I spend much more time researching Blackjack than actually playing it (outside of CV Blackjack that is) but will report back more on all of this once I get all the bugs ironed out.

[JesterCW]

Good show! That makes a lot of sense. I like the storing of the two rule sets between H17/S17 in a pair. I’ll still need to address negatives somehow.

[First Normal Form]

JesterCW, it's a bit off-topic, but on a slow day you might enjoy the poker book, Unleash Your Hidden Poker Memory. The author applies memory techniques to gather and retain data on his opponents. The poker advice itself isn't necessarily great, but it's an interesting idea from a memory maven.

[Secretariat]

@ MIDNIGHT!
Looks like a modified Tarzan count but to make it clear, let's say that in the middle of 6-deck shoe the numbers of cards played for each group are
23 45 67 89 Tens Aces
24-23-28-16-44-12
What would your count looks like?

[Secretariat]

I feel like when I'm practicing my counting and the speed gets kicked up a notch sometimes I start to drop the "N (negative)" in my head and just flow until things settle down again.

Speed should be your priority. Practice with noise, music, TV on. If you can't count at home under difficult conditions you probably won't keep up in a casino. Plus, I would focus on one set of rules, first.

[Midnight]

@ MIDNIGHT!
Looks like a modified Tarzan count but to make it clear, let's say that in the middle of 6-deck shoe the numbers of cards played for each group are
23 45 67 89 Tens Aces
24-23-28-16-44-12
What would your count looks like?

Ha! Ha! You have a very keen eye Secretariat. TOR is a modified Tarzan count and he is the legendary player I was referring to in my original post (that's where the "T" comes from in the name T.O.R.). In your example, my TOR count would be 43-8D-42-E. One of the drawbacks of using the Deck Composition approach to counting is, as your example illustrates, the numbers quickly get larger and more unmanageable as you progress through the shoe. To handle this and simulate the effect of cancellation that you would have in a typical counting system, I employ what I call "Even Distribution Identities (EDI)". The EDI's are as follows:


quarter deck (or 13 cards): 22-22-41
half deck (or 26 cards): 44-44-82
3/4 deck (or 39 card): 66-66-12.3 (The period is just a separator between the tens and aces groups)
Full deck (or 52 cards):88-88-16.4 (The period is just a separator between the tens and aces groups)


In all of these groups the cards are in an exact even distribution with respect to their relative numbers. So in the case of say the half deck EDI, anytime you have a TOR count of 44-44-82 you know that the numbers in those groups are exactly what you would expect in 26 cards. In counting, we don't care about even distributions for betting (TRUE 0) we want to know when the shoe has deflected away from this so we can estimate our advantage or disadvantage. Therefore you can subtract any multiple of any one of the EDI's from the actual TOR count to reduce it to its simplest form which simulates the effect of cancellation in a traditional counting system. This also eliminates the need to side count various ranks as it appears was the case in Tarzan's original system. All the information is stored in the DC count and can be derived from it. TOR has what I call an "optional" side count which is denoted by the "E" in the count 43-8D-42-E above. This is a count of the number of times the EDI has been applied up to this point in the shoe. When tracking this mentally, I keep these with letters instead of numbers as to not confuse myself and separate it completely from the TOR triplet. However when actually playing all I do is keep this with a chip and add one chip to a designated stack for each reduction. What could is this information for? You can use it if your betting count is an unbalanced one like KO or to know exactly where you are in the shoe or how many cards have been dealt to the exact card. So in example above I am using the half deck EDI so I know that exactly 5 half decks have come out (E being the fifth letter of the alphabet). To get the full picture of the exact number of cards that have been dealt up to that point you would just add the TOR count values with the E * 26. So 4 + 3 +8 + (-4) + 4 + 2 + (5*26) or 147 cards have been dealt or nearly 3 decks into a 6 decker. However if I don't care about this auxiliary information I can drop the optional side and just stick with the main TOR count. I use TKO for betting so I would add 4 + 3 + 8 - (4 + 2) + 5 for a RC of 14 @ 3.0 decks remaining which is equivalent to TRUE 1.


TOR is an acronym for Tarzan on Roids. The "On Roids" part comes from the ability of this variant of Tarzan to elevate the Betting Correlation up to the range we are used to in our familiar systems like Hi-Lo and KO (over 96%) while still retaining all the power of Tarzan's approach for playing your hand. Tarzan's original system had a relatively weak betting correlation in the neighborhood of ~.86 due to the grouping of low (2-5), middle (6-8), and high (tens). He would derive his betting count from low - high and then make adjustments with his side count of aces and 9 but even then the correlation for betting still does not reach the level of Hi-Lo as counting the 6 and 7 as neutral cards hurts betting too much. My estimates is that the P.E. of TOR is somewhere in the neighborhood of the .90 described as possible by Griffin in TOB pg. 59 (I calculated it to be .92). However once I get all the simulations completed by one of our esteemed and prominent researchers here, will know for sure. This coupled with perfect insurance, Near perfect 16 V Ten (98% correlation), near perfect 12 vs 3 (99.12% correlation) in fact near perfect every playing deviation, should make TOR a powerful system in theory, and as close to having a combinatorial analyzer with me at the table as I could get.

[JesterCW]

Speed should be your priority. Practice with noise, music, TV on. If you can't count at home under difficult conditions you probably won't keep up in a casino. Plus, I would focus on one set of rules, first.

Absolutely. Variance was on my side on my last trip to the Casino but I got a lightning fast dealer and ended up head to head multiple times with them before I went to another Casino across the street. I lost the count almost immediately— and although I would regain it I knew it was no longer accurate and ended up just playing basic with any obvious deviations. It went better at the other place but I had a slower dealer and a few sweet (but slow) ploppies that sat down with a BS card, must have been in their 70s and it was almost 2am. They did well!

I digress though, my current training looks something like this:
-100 cards a day min for AnkiApp (deviations, some more annoying division)
-CVBlackjack Games (2deck/6deck) with my deviations loaded in
-CVDrills (quick counting, true count conversion, and flash drills mainly — some whole table counting)
-Actually counting down decks of cards with two cards removed to check at the end
-Dealing to myself and my roommate
-I’ll add in more distractions for counting, and try to keep up the speed/accuracy

Agreed to focus on my current game/deviations as that’s what I’ll be playing for a bit. Given my current work situation, my bankroll is really what’s stopping me. I have a Computer Science degree I apparently burned out from a bit and am really just saving for the most part right now while I practice and learn things. Like PAO systems . I might play soon on a replenish-able bankroll of around $1-2k a month but want something to start with more than $600 (as tempting as it may be and I’ll admit I’ve gone with $200 and gotten lucky, but I wouldn’t have lasted 10 min with some splits/doubles or any bigger bets so I was gambling — don’t want to do that again!).

@midnight — I’m just getting back from a work trip and am switching jobs so life has been hectic — but I’ll chime in on your system when I get a bit more brain power! I’ll say it seems like something you’ve put a lot of thought into and it shows. Right on!

[Secretariat]

@ MIDNIGHT
Interesting stuff Midnight and quite a feat of mental gymnastic.
I am looking forward to the sims and how you worked out the indices.
So for 24-23-28-16-44-12 (cards played = 147 rounded to 3.0 decks)
your count is 43-8D-42-E.
For that specific deck composition Tarzan count would be 47-44-44 (12) and thus translated into 3-0-0-12 (RC3 or TC1)
You get to the same result.
I fail to see how you get to perfect insurance (quickly). Can you explain?
One strong point of the Tarzan count is instantaneous perfect insurance deck compo of 33.3% + tens)

EDIT 2
Looking back, I see that there are 12 more 67s out of play than 89ss (28vs16=12).
So the count is stronger than TC1.
This warrants a TC2 bet and perhaps a TC3 bet. Can you make that adjustment in your betting?

[Midnight]

@ MIDNIGHT
Interesting stuff Midnight and quite a feat of mental gymnastic.
I am looking forward to the sims and how you worked out the indices.
So for 24-23-28-16-44-12 (cards played = 147 rounded to 3.0 decks)
your count is 43-8D-42-E.
For that specific deck composition Tarzan count would be 47-44-44 (12) and thus translated into 3-0-0-12 (RC3 or TC1)
You get to the same result.
I fail to see how you get to perfect insurance (quickly). Can you explain?
One strong point of the Tarzan count is instantaneous perfect insurance deck compo of 33.3% + tens)

EDIT 2
Looking back, I see that there are 12 more 67s out of play than 89ss (28vs16=12).
So the count is stronger than TC1.
This warrants a TC2 bet and perhaps a TC3 bet. Can you make that adjustment in your betting?

Hey Secretariat, good catch. I made a slight error above when I calculated the TKO RC from the TOR DC Count. As I was using the half deck EDI I was supposed to use E = 10 (i.e. 5 * 2) which would have resulted in a RC of 19 and not 14. This is equivalent to TC2 as you state. I have been doing lots of research lately using the quarter deck EDI for what I call TOR-Shuffle and TOR-sequence so have this firmly in mind. But back to your question on perfect insurance. As stated earlier, in TOR all the betting counts and individual play counts are derived from the TOR DC count. For this particular play, i.e. insurance, I use what I call a "K mapping" (or "K pattern" or "K value") to get from the TOR count to the play count for insurance which is the unbalanced 10 count. The K mappings or K patterns are integers that when applied to the zero EDI (i.e. 00-00-00), result in the TAG values of the desired play count. For insurance the K values that map the TOR count to the unbalanced 10 count are 11-11-(-2)1. We would then compute the running count for the play as the sum of these K * G products (where G is the TOR value for the respective group). So for our example, Multiplying each group by the respective K value and then summing results in 4 * 1 + 3 * 1 + 8 * 1 + (-4) * 1 + 4 (-2) + 2 * 1 + 10 (for E =10 as this is unbalanced with half deck EDI). This sums to RC = 15 for the unbalanced 10 count. For that count which has 100% correlation with the insurance EORs, you take insurance with any running count greater than 24 anywhere in the shoe for the 6D game. So with this particular Deck Composition we would not insure if faced with a dealer ace up.

The EDI concept and K mappings is what allows me to convert the TOR DC Count to almost any particular count that I want. If I want TKO for betting I use the K values 11-10-(-1)(-1). If I feel lazy and wanted to skip the custom play counts and just use HI OPT II for all plays I use the K values 12-10-(-2)0. With a DC count like TOR you are not locked into any particular system. As long as you know the indices for whichever system you want to use you can use a K mapping to go from TOR to that system on the fly right there at the table. This is why I call it the swiss army knife system. Its versatility allows it me to do all kinds of things that I can't do with a traditional system like Hi-Lo. For example, TOR allows me to play both the main game and countable side bets at the same time without having to abandon one for the other as is often the case. Most times when you find a countable side bet, if its EORs are not highly correlated with your main count you have to make a decision on which one is more valuable and just go after that one. With TOR I can have my cake and eat it to. Once I find appropriate K values for the side bet, it is relegated to just another play with its own play count. The win rate for the main and the side bet then become additive instead of mutually exclusive. Anyway its late and I'm tired so I'll use another post tomorrow to methodology I used to compute the indices.

[JesterCW]

Just a simple addition to a thought I had earlier around negatives and a PAO system. I’m now using three modifiers to represent if the first, second, or both digits of the number pair are negative. (Wrapped in a rope, chain, and barb-wire) — for three number pairs I read them from top to bottom on the image. Seems to be working.

[Midnight]

Just a simple addition to a thought I had earlier around negatives and a PAO system. I’m now using three modifiers to represent if the first, second, or both digits of the number pair are negative. (Wrapped in a rope, chain, and barb-wire) — for three number pairs I read them from top to bottom on the image. Seems to be working.

Hey Jester, that is an excellent idea if I should say so myself. I had a rather cumbersome way of handling this very problem in my PAO system for indexes but I think yours is better. Hope you don't mind but I'm stealing this to add to my TOR PAO images. Thanks for sharing the idea.