BJInfo Open Source UBZ II V0.5

[boneuphtoner]

(with sweet 16, fab 3 indices)

Whoops! I meant Fab 4!!!

 

[nightspirit]

boneuphtoner, good job too!

Your idea reminds me of AM's BF-Count. You choosed also index values which are easy to remember. I like this idea with one index set for all decks, since most of the gain comes from betting in multideck. I believe George C. did the same and for those, who play more 2 deck he had special index set. That our versions perform better, than the canned ones sounds good. But we must bear in mind that they don't split tens and don't double the soft hands, which we included.

I don't generated risk-averse indices, because the gain when using them is only marginal. And our rounded versions performs as well as AM's SVUBZ.
(Dead link: http://img475.imageshack.us/my.php?image=svubzvfullos2did9.jpg) _

_

Regarding your index set, I would skip the +15 indices. Make 10 vs. 10 at +10 for multideck and +5 for handheld games and remove 13 vs.10. I won't double my soft 20. Instead I would inlude other soft hands like A,7 vs. 2, A7 vs. A, A8 vs. 4 and A6 vs. 2. Just my suggestions. Anyway, nice work! Can you post a screenshot?

 

[Mimosine]

Make 10 vs. 10 at +10 for multideck and +5 for handheld games and remove 13 vs.10. I won't double my soft 20. Instead I would inlude other soft hands like A,7 vs. 2, A7 vs. A, A8 vs. 4 and A6 vs. 2. Just my suggestions. Anyway, nice work!

I would agree with this.

I think all these ideas are great. Rounded indices specific for each game or I like the all purpose indices as well. I'd choose the ones NS and I came up with (for 6D or SD&DD) over 1 set of all around indices, but this largely depends on what games you play to. If you play 6D 90% of the time, but once in awhile find a good DD or SD game, then don't waste the time on 2 sets of numbers.

Again to be clear, INSURANCE can never be a rounded number. It should always be precise because it constitutes roughly 1.1% of the advantage gained in index play. The next 17 I-18 plays don't add 1% combined!!!! Thus NS's numbers and my rounded numbers and boneuphtoner's numbers don't impart drastic changes in the output. significant and measurable, yes, but the decision on which numbers to use should always be left to the individual using them.

The two stated goals (or maybe implied) are simplicity in execution and performance BETTER than KO. The reasons being for me is that I know I can do a level 2 count and I love using KO, so this is my(our) attempt at a hybrid between the two.

I think very shortly we'll be at a stage where we can put all these things in this thread together and publish them as a freely available PDF - assuming NS that you're willing to allow this with your work.

That is where I'm leaning for the time being. there is work to be done, and i have not been contributing as much as i'd like, but as it stands we have several UBZ OS variants that achieve the goals stated above. For that thanks for everyone's contributions, especially NS.

And most importantly, UBZ is now a real card counting system (well, we have to get surrender and a bet ramp in order, but the latter can be optimized pretty easy in the absence of index play).

 

[Mimosine]

The cherry on the top would be to get some wonging and -out points with the TC version.

I already have the deck adjusted wongout and wongin points for 6D done.

Here are my initial estimates, again we could try to round them.

6D IRC = -24

WONG IN

5 Decks remain RC = -8 • TC = ~1.20 (bet 2units)
4 Decks remain RC = -6 • TC = ~1.25 (bet 2units)
3 Decks remain RC = -4 • TC = ~1.33 (bet 2units)

WONG OUT

5 Decks remain RC = -28 • TC = ~ -0.8
4 Decks remain RC = -24 • TC = ~ -1.0
3 Decks remain RC = -18 • TC = ~ -1.0

These Numbers are VERY rough.... But we can use optimized bet spreads to determine good wong in points based on when we should bet 2 units. I calculated these numbers using excel, but there is something a little wonky with the calculation, the RC's might be off by +/- 4, or they might be just right.... I'm not 100% confident in them, I know they're close and several conversations with Fred Renzey lead me to believe I'm on the right track.... maybe someone could sim these that after all would be a good indicator if my intuition is good (again )

 

[nightspirit]

The two stated goals (or maybe implied) are simplicity in execution and performance BETTER than KO.

I think that goal is achieved. The rounded versions perform better than KO-preferred but (understandably) not better than KO-Full. But to use only the most important BS deviations saves us, the simplicity! When I find some time in the next weeks, I will try to generate a full set of indices for 2 and 6 decks, such as published in the KO book (just for the sake of completeness). And maybe we could add the one or other index number for single and double deck.

I think very shortly we'll be at a stage where we can put all these things in this thread together and publish them as a freely available PDF - assuming NS that you're willing to allow this with your work.

Are you kidding me??? Thats a fantistic idea! :grin: Most of the work did my computer, special thanks to him!

That is where I'm leaning for the time being. there is work to be done, and i have not been contributing as much as i'd like, but as it stands we have several UBZ OS variants that achieve the goals stated above. For that thanks for everyone's contributions, especially NS.

That's not worth mentioning. It was a great idea by you! And without your interest in that project nobody would have cared about it. You really have a brilliant understanding of the unbalanced counts and you proved again that

"in general the published versions of systems are NOT the superior versions" Quote zengrifter

And most importantly, UBZ is now a real card counting system (well, we have to get surrender and a bet ramp in order, but the latter can be optimized pretty easy in the absence of index play).

Yeah, I found it undervalued all the time! The surrender indices are almost completely, only 15 vs. 10 is absent. The concept of key and pivot values got somehow lost. The bet ramps are not the problem, since the sims are already there.

 

[nightspirit]

I already have the deck adjusted wongout and wongin points for 6D done.

Here are my initial estimates, again we could try to round them.

6D IRC = -24

WONG IN

5 Decks remain RC = -8 • TC = ~1.20 (bet 2units)
4 Decks remain RC = -6 • TC = ~1.25 (bet 2units)
3 Decks remain RC = -4 • TC = ~1.33 (bet 2units)

WONG OUT

5 Decks remain RC = -28 • TC = ~ -0.8
4 Decks remain RC = -24 • TC = ~ -1.0
3 Decks remain RC = -18 • TC = ~ -1.0

These Numbers are VERY rough.... But we can use optimized bet spreads to determine good wong in points based on when we should bet 2 units. I calculated these numbers using excel, but there is something a little wonky with the calculation, the RC's might be off by +/- 4, or they might be just right.... I'm not 100% confident in them, I know they're close and several conversations with Fred Renzey lead me to believe I'm on the right track.... maybe someone could sim these that after all would be a good indicator if my intuition is good (again )

Very cool! I'm not that familar with CVData's shoe depth function, but where there's a will, there's a way.

 

[boneuphtoner]

Show me how!

Can you post a screenshot?

I'd love to, but I tried copying and pasting the CVCX charts, and when I pasted them into this message, all I got was a jumbled series of columns containing the raw data. Any ideas?

 

[Mimosine]

I'd love to, but I tried copying and pasting the CVCX charts, and when I pasted them into this message, all I got was a jumbled series of columns containing the raw data. Any ideas?

export as a tab delineated file from excel. then copy paste, then muck about with it a bit, and finally you might get something presentable.... these forums have really WEAK html support. otherwise you could make some really nice tables using the <table> feature...

 

[nightspirit]

I'd love to, but I tried copying and pasting the CVCX charts, and when I pasted them into this message, all I got was a jumbled series of columns containing the raw data. Any ideas?

I use Darkshot2 for screenshots. Then you can upload it to one of these free picture hosting sites, like imageshack or photobucket.

For which spread and penatration you would like to see the bet ramps, Mimosine? Any preferences?

 

[boneuphtoner]

Lets see if this works

Here are a few comparisons of my 25 indices compromised UBZ with the level 2 Mentor and RPC Counts (with Sweet 16 and Fab 4). Included comparisons are 6, 8, and 2 decks. My UBZ-Ultimate Strategy is shown in blue.

 

[boneuphtoner]

Here's more

Here is a single deck comparison. Again the same indices were used for all decks except for single deck where all of the +10 indices become +5. Everything else is the same. Also included is a 6 deck comparison of my UBZ strategy (in blue) with some of the level one counts and the canned UBZ sim (in yellow).

 

[nightspirit]

Hey boneupthoner, cool! Looks good! Have you tried to run a sim with the same indices like the other counts use them, as well?

I run a sim using the sweet 16 and fab 4 for the UBZ-OS, the good "3" version.

(Dead link: http://img452.imageshack.us/my.php?image=ubzsweet16fab4vhilovkoud0.jpg) _

_
There is no gap anymore:cry: ! And here one with the Illustrious18 (Dead link: http://img411.imageshack.us/my.php?image=ubzillustrious18vkofullxk8.jpg) _

_

Splitting teens adds a bit, but how often can we do it?
I think after school is necessary...

 

[zengrifter]

That leveling off and dip in the 1D sim is curious. zg

 

[boneuphtoner]

Here are a few more comparisons. One is a chart with my UBZ strategy stripped to the Sweet 16. As you can tell, performance does suffer, it no longer outperforms, but it still is in line with the balanced level 2 Mentor and RPC. Also shown below is my full rounded 25 indices UBZ as compared to the Mentor and RPC full uncompromised indices. It sure holds its own IMHO!

 

[zengrifter]

Interesting comparisons.

The optimized composite UBZ is the easiest of the level-2 systems without any compromise of performance. Now who is or will be using it or a close facsimile thereof.

Another comparison, please: Add a true-counted UBZ to the unbalanced line-up. Then lets compare Mentor, ZEN and TCUBZ separately.

A dedicated newbie could start with the UBZ count and an adjusted BS - graduate to 16-25 indices - and if still aspiring learn to true-count and learn new indices.

Shuffle Tracking - I have heard it said repeatedly: "Unbalanced counts are NO GOOD for shuffle tracking". I don't know about that - on more than one occaission I spent hours STing with one or another top BJ-math person who was using KO and neither made any mention to me that their play would be compromised. Is there an argument in defense of STing with unbalanced counts?

What about that 'dip and level' in the previous 1D chart - what is that? 'Cut-card effect'? zg

 

[boneuphtoner]

I honestly have no idea whats up with the dip and level observation you brought up about the 1D sims....what I can say is that ALL of the CVCX SD SCORE sims I've seen show this....all of the canned sims run by Norm and the all of the ones I've run..... I gotta ask Norm about that.

 

[bj bob]

Look how one favors Mentor and one favors RPC.

What about that 'dip and level' in the previous 1D chart - what is that? 'Cut-card effect'? zg

I noticed that same "dip and launch" on the graph too, Zg. Very interesting! Off the top of my head, the difference between the 19 and 15 shuffle points is a whopping +45 in the SCORE. Seems to be the exact difference between Ro6 and Ro7. Next thing I need to figure out is what exactly that translates into in $'s. Suppose your're playing $25-$75 and how much EV is gained by being dealt that "one more hand". Also, would it be worth it to toke the dealer (bribe:devil: ) a buck or two at that point or perhaps artificially extend a marginal hand before the mandatory shuffle point, e.g. continue to hit an a-7 v. 3 rather than stand.

 

[nightspirit]

Another comparison, please: Add a true-counted UBZ to the unbalanced line-up. Then lets compare Mentor, ZEN and TCUBZ separately.

A dedicated newbie could start with the UBZ count and an adjusted BS - graduate to 16-25 indices - and if still aspiring learn to true-count and learn new indices.

I wanted to sim the wong-in/out adjusted version by Mimosine but it seems that my computer makes his last breath at the moment (simmed to death )
Qfit's software doesn't contain the UBZ in true count mode but i found a comparision by Cacarulo (Dead link: http://www.advantageplayer.com/blackjack/forums/bj-main/webbbs.cgi?read=9427) _Ranking_ There is also a comment by Fred Renzey in this thread.

Here is a comparision between Mentor and Zen (canned sims) by Qfit.

(Dead link: http://imageshack.us) _

_

Here are a few more comparisons. One is a chart with my UBZ strategy stripped to the Sweet 16. As you can tell, performance does suffer, it no longer outperforms, but it still is in line with the balanced level 2 Mentor and RPC. Also shown below is my full rounded 25 indices UBZ as compared to the Mentor and RPC full uncompromised indices. It sure holds its own IMHO!

I'am very impressed by it! Damn good work!

 

[zengrifter]

Qfit's software doesn't contain the UBZ in true count mode but i found a comparision by Cacarulo (Dead link: http://www.advantageplayer.com/blackjack/forums/bj-main/webbbs.cgi?read=9427) _Ranking_ There is also a comment by Fred Renzey in this thread.

Cacarulo's comparison shoots down AutoMonk's contention that RPC is superior to ZEN in shoe games.

Actually, Brett Harris established ZEN's superiority over RPC in shoes - EVEN IN ENHC games - nearly 6-years ago.

It also shoots down my theory that TUBZ would outperform ZEN.

The QFIT sim above is in contradiction to Caraculo's ranking, Mentor outperforming ZEN. zg

 

[boneuphtoner]

Here is an explanation about the SD behavior

Hi All,

ZG had requested an explanation for the ususual behavior in the CVCX single deck sims. All of the canned sims plus the ones I've run show this leveling off as one progresses to deeper penetrations, followed by a subtle dip in SCORE. I asked qfit's Norm Wattenberger about this, and this is how he explained it...after thinking about it, it really make sense:

This occurs in single deck games with a fixed penetration with multiple players. When you set a penetration; the dealer doesn't shuffle at that exact point. The shuffle occurs at the end of that round. If you change the penetration by adding one card; then the effect is to increase the number of rounds dealt in some percentage of shuffles. Add yet another card and the percentage changes again. But, the effect is not linear. And it isn't always a good thing. Suppose you increase the penetration just enough to eke in another round. Another round is good. right? Well, if you just managed to eke in another round; that means you probably didn't deal as many cards per hand in the previous rounds. Which means the cards were high and the count is low. So that extra round is not favorable. In other words, the cut card effect comes into play with a vengeance at some particular penetrations in single deck games with a fixed penetration with multiple players and that causes a dip in SCORE at those penetrations. Add a few more cards and you will no longer be at the point where you are just ekeing in low count rounds and the effect will mitigate. Switch to multiple decks, heads-up play or fixed rounds and the effect goes away.

I've always said that single deck is a fundamentally different game. Nothing in BJ is linear (except Insurance.) Strange effects that appear only as minor perturbations in shoe games can become glaringly obvious in single deck.