I have been considering switching from Hi-Lo to Zen.
I would be interested in the factors that went into your choice. Was it simulation results? How much was based on the playable games that currently offer the best advantage?
Further, I've been seeing some statements about the different 'flavors' of the Zen count. Particularly that Zen 98 performs somewhat poorer than Full Zen. Which versions did you examine, and which did you choose?
> I have been considering switching from Hi-Lo
> to Zen.
> I would be interested in the factors that
> went into your choice. Was it simulation
> results? How much was based on the playable
> games that currently offer the best
> advantage?
> Further, I've been seeing some statements
> about the different 'flavors' of the Zen
> count. Particularly that Zen 98 performs
> somewhat poorer than Full Zen. Which
> versions did you examine, and which did you
> choose?
I looked at all versions, and settled on the Unbalanced Zen 2 count, as developed by George C. and described in his book, Unbalanced Zen 2.
This is an unbalanced variant of the Zen counts that counts the 3 as +2 (balanced Zen counts it as +1), and like all Zen counts, counts 10 value cards as -2 and aces as -1, thus "semi-neutralizing" the ace. This provides at least some of the value of an ace-neutral count without the hassle of an ace side count.
Being a level 2 count unbalanced by a factor of 4, the pivot point is roughly equivalent to a Hi-lo TC of +2.
Since it is unbalanced, it may be used in running count mode. In other words, like KO, no true count conversion is required.
My decision was based on several factors. One of the biggest was the discussion in Chapter 11 of Blackjack Attack, regarding level 2 counts in general and UBZ2 in particular.
In additon, George C. developed UBZ2 with the double deck game in mind. I play a lot of double deck.
I tried to come up with one set of indices for single and double deck games, but could not find a suitable compromise. So, I use one set of indices for single deck and a different set for double deck. I use about 30 indices for each game.
On those rare occasions when I play shoes, I use a true-count conversion.
Thanks for the response. I obviously misread posts from way back when you'd just made a switch. It is a bit daunting reading the amount of time you reported it took to get back to being comfortable counting while chatting.
Unfortunately I find my time split more closely between double decks and shoes, though I can usually shop around and find shoes with not much over 1 deck cut off. As I understand it a deep-cut shoe hilights how an unbalanced count strays away from giving the actual advantage and correct play choice.
The need for multiple sets of index numbers has concerned me. Is it workable to have a very small set of running count indexes and true-count the rest? Are there easy shortcuts using running count, true-count index, and depth to show whether the play is close enough to require checking the index?
Quick division isn't a problem for me, and I'm seeing good results in working with card stacks to improve my estimates, they had been quite bad. How much does eliminating the need to true-count go into the choice of UBZII over Zen?
> Thanks for the response. I obviously misread
> posts from way back when you'd just made a
> switch. It is a bit daunting reading the
> amount of time you reported it took to get
> back to being comfortable counting while
> chatting.
Well, that's just me. Some people take to counting cards like a fish to water, others (like me) have to put a little effort into learning. Some people frequently change counts, and even use one count for single deck games and a different system for shoes.
I hate those people. :-)
> Unfortunately I find my time split more
> closely between double decks and shoes,
> though I can usually shop around and find
> shoes with not much over 1 deck cut off. As
> I understand it a deep-cut shoe hilights how
> an unbalanced count strays away from giving
> the actual advantage and correct play
> choice.
> The need for multiple sets of index numbers
> has concerned me. Is it workable to have a
> very small set of running count indexes and
> true-count the rest? Are there easy
> shortcuts using running count, true-count
> index, and depth to show whether the play is
> close enough to require checking the index?
The double deck game actually more closely resembles a shoe than a single deck game. The single deck game is the kicker, and my reason for using more than one set of indices. If you rarely play single deck, you can use one set of indices for all numbers of decks. George C. recommends standardizing to a pivot point of +2 by adjusting the IRC, and provides a "generic" set of indices for 2, 4, and 6 deck (they will also work on 8 deck).
> Quick division isn't a problem for me, and
> I'm seeing good results in working with card
> stacks to improve my estimates, they had
> been quite bad. How much does eliminating
> the need to true-count go into the choice of
> UBZII over Zen?
That is the main difference. However, if you read chapter 11 of BJA (chapter 9 in BJA3), you will find that UBZ2 performs nearly as well as, and in some cases better than, the balanced Zen count. The gain from true-counting is surprisingly small.
Perhaps not so surprising, the most significant gains from a true counted system are realized when back-counting/wonging.
Thanks, I think you have me started on the questions that I'll need to now follow up on myself. Yes, I'd gotten pointed this way from BJA2, I believe that when I purchased it less than a month before BJA3 was out (ouch!).
I know I'll end up poking into this question for my own answers now, I guess I'm back to Norman Wattenbertg for more products.
Maybe I do come to this with a prejudice against the unbalanced counts, the thought that of course if you do more you'll get better results.
> Are there easy shortcuts using running count, > true-count index, and depth to show whether
> the play is close enough to require checking
> the index?
You can do this. CVData can help you come up with different RC index numbers depending on depth. Theoretically, you could be as accurate using this method as normal true counting. However, since quick division isn't a problem for you. . .
Keep in mind the SCORE chapter compares systems using only 22 indices. Don points out, ". . . all systems gain in value if 100 additional indices are employed. But, that gain is not constant across all systems,"(BJA2, p.292). And surely Zen gains steadily on UBZ as you learn more indices. You'd agree, right Parker?
As an example of how this plays out, note that in the charts, Zen rarely beats Hi-Lo by more than 10%. But if you compare both systems using full index numbers (current, state-of-the-art index numbers, not necessarilly the ones listed in original texts), that same measure hovers around 15%. This is true in various versions of Reno and Vegas rules, though I haven't looked much at S17 and LS games.
So, if you plan on learning just the main 20 or so index numbers, it's sort of a toss up, but if you plan on learning more and more as time goes on, you might want to investigate further. I haven't done any sims on UBZ or I'd give a firmer opinion.
Myoo
Thanks to both of you for the ideas. Most new, some supporting possibilities I'd considered.
I find myself on a self-imposed 'timeout'. One backoff and several obvious warnings (I get to the cashier's cage, they answer the phone, then ask for my name...). It seems like a good time to stay out of the local spots for a while and improve my act. But, my act will improve most if I can be devoting less attention to counting and more to my surroundings. So I'll work it on CV. But, if I'm ever going to switch systems now is the opportunity. So I want to nail down what is best for me and the games I see.
> You can do this. CVData can help you come up
> Keep in mind the SCORE chapter compares
> systems using only 22 indices. Don points
> out, ". . . all systems gain in value
> if 100 additional indices are employed. But,
> that gain is not constant across all
> systems,"(BJA2, p.292). And surely Zen
> gains steadily on UBZ as you learn more
> indices. You'd agree, right Parker?
Sure. In any unbalanced system used in running count mode, the indices closest to the pivot point will be most accurate. The further from the pivot, the more an index will be affected by penetration. This isn't the case with true-counted systems.
Of course, UBZ2 can be true-counted as well. I haven't run any sims (yet), but I have a hunch that a true-counted UBZ2 would out-perform balanced Zen by a fair margin. Why? Because the pivot is at a more useful place, and, in order to balance the system, Zen undervalues the 3.
> You can do this. CVData can help you come up
> with different RC index numbers depending on
> depth. Theoretically, you could be as
> accurate using this method as normal true
> counting. However, since quick division
> isn't a problem for you. . .
> Keep in mind the SCORE chapter compares
> systems using only 22 indices. Don points
> out, ". . . all systems gain in value
> if 100 additional indices are employed. But,
> that gain is not constant across all
> systems,"(BJA2, p.292). And surely Zen
> gains steadily on UBZ as you learn more
> indices. You'd agree, right Parker?
> As an example of how this plays out, note
> that in the charts, Zen rarely beats Hi-Lo
> by more than 10%. But if you compare both
> systems using full index numbers (current,
> state-of-the-art index numbers, not
> necessarilly the ones listed in original
> texts), that same measure hovers around 15%.
> This is true in various versions of Reno and
> Vegas rules, though I haven't looked much at
> S17 and LS games.
> So, if you plan on learning just the main 20
> or so index numbers, it's sort of a toss up,
> but if you plan on learning more and more as
> time goes on, you might want to investigate
> further. I haven't done any sims on UBZ or
> I'd give a firmer opinion.
> Myoo
I have the poop to confirm Myooligan's speculation with a little background to boot.
I changed counts last year based on my own evaluation. I played around practicing with several counts knowing full well which had the better PE and BC. Over the last few years I?ve read most of the available BJ books and with SBA and BJRM, I decided to do my own specific evaluation. I know this has been done out the wazoo and overload of info is available in BJA2&3 but hey-I did it my way!
I read Wong?s book and started with hi-lo initially and then switched to KO playing only DD (for about 2 yrs) with 50-60 SBA generated index numbers. Played around some with UBZ2 (based on BJA2, Parker?s and John Auston?s bj21 older posts) with minimal indexes before I decided to do this evaluation. With unbalanced counts, SBA eV maximizing index numbers and BJRM I18 index numbers show you need to vary the index as the deck is depleted (?true fudging? (BJBB2))-and I was doing that but I still wasn?t satisfied. From literature, unbalanced counts are most accurate near pivot which can be detrimental if you play all as I do for DD. I did come to the same conclusion as was posted on bj sites and in several books, index numbers for unbalanced counts are not worth much for off the wall high counts. Standard deviation, negative fluctuations, and just plain bad cards are something we can?t eliminate.
Take this with a grain of salt, but based on the data below and with the seemingly continual demise of good DD games, Ace side count just isn?t worth it. I decided to convert to Zen for use with 6 dk as well as the remaining H17 DD games. A true compromise in that it is close to best in all three: PE, BC, & IC. Though 2 level count takes much longer to implement (talking and counting) not to mention true count conversion (but that?s necessary maybe 15% of the time), only 10-12 index numbers need to be adjusted for the various games (S17 vs H17) and #decks.
I included 3 unbalanced counts (KO, UBZ2, & AKO) and 4 balanced counts (Mentor, HiOpt2, AO2, & Zen).
Rather than just the I18 indexes, and since I have no aversion to memorizing large number of indexes, I used ALL the SBA generated index numbers in the evaluation. I used indexes generated for pen of 70 for all the counts and no ?true fudging? for unbalanced. Penetration of 62 reduces the KO SCORE for game 1) from 60.22 to 41.34-about a 30% reduction. Didn?t bother with the other counts.
The following parameters were chosen to compare the various counts for the 4 distinct DD games listed below in the corresponding numbered rows (1,2,3,4):
SBA random seed: 1234567890
100 x 10^6 rounds for initial screening-then 500x10^6 for the top 3 counts,
70 card pen,
Resplit up to 3 times for 4 hands,
1 cd to split aces
3 players-counter at seat 3,
True count floored,
Optimal BJRM $10 units spread 1-6,
Play only 1 hand,
13.5% RoR,
Ace side count for AO2 and HiOpt2 for betting valued at ?2,
Ave bet not tabulated-generally about 2 units.
No surrender
DD games (rows)
1) H17 DS rsa 1 cd
2) H17 DS
3) S17 DS rsa 1 cd
4) S17 DS
Generated data (columns) for comparison:
a) BJRM SCORE
b) BJRM hourly win rate expressed as % of initial bets: W/L%
bx) SBA TBA (for the 500MM sims only)
c) SBA TBA from Insurance
d) SBA Insurance contribution to TBA %
KO, a, b, c, d
(1), 60.2, 1.010, 0.092, 7.436
(2), 55.1, 1.000, 0.1, 7.774
(3), 78.5, 1.180, 0.09, 7.409
(4), 72.9, 1.130, 0.09, 7.306
Ave, 66.68, 1.080, 0.093, 7.481
UBZ2, a, b, c, d
(1), 64.32, 1.05, 0.1027, 7.76
(2), 58.68, 1.003, 0.1046, 7.846
(3), 82.26, 1.179, 0.0953, 7.58
(4), 76.21, 1.162, 0.1014, 7.695
Ave, 70.37, 1.099, 0.101, 7.720
AKO, a, b, c, d
(1), 62.88, 1.082, 0.1072, 8.387
(2), 58.98, 1.026, 0.1007, 8.14
(3), 82.99, 1.244, 0.0994, 8.014
(4), 77.09, 1.194, 0.1052, 8.35
Ave, 70.49, 1.137, 0.103, 8.223
Mentor, a, b, c, d
(1), 70.56, 1.119, 0.0924, 7.457
(2), 63.04, 1.049, 0.0959, 7.59
(3), 83.91, 1.21, 0.0918, 7.433
(4), 79.03, 1.182, 0.0932, 7.526
Ave, 74.14, 1.140, 0.093, 7.502
HiOpt2, a, b, bx, c, d
(1), 69.39, 1.3, 1.095, 0.1091, 7.85
(2), 64.16, 1.24, 1.041, 0.1107, 7.895
(3), 92.93, 1.47, 1.26, 0.1066, 7.854
(4), 83.17, 1.41, 1.19, 0.108, 7.857
Ave, 77.41, 1.355, 1.147, 0.109, 7.864
AO2, a, b, bx, c, d
(1), 69.39, 1.3, 1.09, 0.0964, 7.78
(2), 61.62, 1.22, 1.017, 0.0991, 7.925
(3), 89.11, 1.45, 1.247, 0.0951, 7.832
(4), 80.82, 1.4, 1.183, 0.095, 7.756
Ave, 75.24, 1.343, 1.134, 0.096, 7.823
Zen, a, b, bx, c, d
(1), 69.22, 1.33, 1.096, 0.1022, 7.523
(2), 65.29, 1.27, 1.046, 0.1044, 7.607
(3), 90.06, 1.49, 1.237, 0.0998, 7.507
(4), 82.99, 1.43, 1.186, 0.101, 7.527
Ave, 76.89, 1.380, 1.141, 0.102, 7.541
Thanks for giving me more to think about.
From my brief use of hi-opt 1 when I returned to counting I know I won't be deploying any side counts. I'd used +- before, a blend from Revere and Thorpe so hi-lo the obvious choice.
I also know I don't want to switch counts again. The process is painful enough for me that I'll have to convince myself both on authority and with my own trials that a count is the best for me both for initial use and for where it can be expanded to. Adding TC to an unbalanced count is one way to grow in a count, adding more indexes is another.
A side question. Griffin's book shows gain from perfect full-memory strategy at about 1.3% for single deck, .4% for 4 deck and .3% for 6 deck. Ever seen the figure for double deck, or maybe I missed his description of how to calculate that for X decks?
Posting Permissions
-*-
Bookmarks