BJInfo Open Source UBZ II V0.5

Mimosine

Well-Known Member
#1
Hey Gang,
In light of recent interest in making a workable UBZII I've slapped together a UBZII variant based on the original UBZII manual, "KO Blackjack," Advice from Fred Renzey and AutoMonkey (see: http://www.blackjackinfo.com/bb/showthread.php?t=6666 )

The goal - develop rounded indicies that are easy to use and remember for SD, DD, and 6D games.

the method - common pivot point of 0 which is equal to TC = +2 at all points in the deck. To accomplish the goal -- use different IRCs. To simplify betting and indices we need to develop a Key Count (defined where we first have the advantage) and for V0.5 we will define two additional points; first - "Advantage" where we have a TC on average of +4, and a second upper level "Rogue" where we have a TC on average of +6. These 3 points; KeyCount, Advantage, Rogue will varry based on the number of decks, but also should be very easy to remember and incorporate into play.



Code:
Tags									
2	3	4	5	6	7	8	9	 T	 A
1	2	2	2	2	1	0	0	-2	-1
Initally i'm trying to get these numbers to work for Key Count, Pivot Point, and Advantage:

SD -2, 0, +2
DD -3, 0, +3
6D -6, 0, +6

The Rogue number is still in its infancy.

the Key Count, Advantage, and Rogue numbers very very roughly correlate
to the average TC listed below.

Code:
           [U]  IRC     KeyCount  [b](P)ivot[/b]  (A)dvantage (R)ogue[/U]
    SD        -4        -2        [B]0[/B]         2
    DD        -8        -3        [B]0[/B]         3
    6D       -24        -6        [B]0[/B]         6         8
  AVG TC       0         1        [B]2[/B]         4         6
   Bet        1u        2u     1/2 Max   Max bet   2 Hands

Simplified Index plays, using KO Blackjack as inspiriation:

Take Insurance at +1.

Code:
	UPCARD									
[U]Hand	2	3	4	5	6	7	8	9	10	A[/U]
17										
16								A	K	
15									P	
14										
13	I	I								
12	P	P	K	I	K					
11										P
10									A	P
9	P					P				
8				P	P					
10,10				R	R					
A,8			R	A
in All cases the P = 0. "I" will be the IRC (easy enough to remember) K and A represent -x and +x based on the number of decks. e.g. 6D K = -6, A = +6.

The mysterious Rogue value is for the player willing to split 10s, play 2 hands, and double down on A,8 V 4 --- all some of the most risky plays.

I will be developing this further, once I get a simulator, and maybe some feedback. to me this is a logical extension of KO blackjack, that is easy for me to grasp. Looking at Automatic Monkey's Simian Varriant UBZ, or George C's or Fred Renzey's KISS indicies, I'm not quite ready to digest so many different index numbers - thus I'm trying to make a simplified rounded index matrix.

Betting being the most important part of counting, my numbers there too are also a first approximation.

thoughts to get me to a V0.6 ?
 

TENNBEAR

Well-Known Member
#2
Great Work, I think you are on to a unbalaced system that will improve the overall strength in the games with weak penetration.
I am planning to upgrade from KO to UBZEN2 soon.
 
Last edited:

Knox

Well-Known Member
#3
Renzey said:
The Count Sytem Performance Comparisons on pg 194 of BJBBII showed the EVs of UBZII to be +0.72% vs. +0.68% for Full Matrix KO. UBZII was tested using the exact format I suggested in my last post.

Is that enough to change? A personal call.
My call right now is no, its not worth it. The beauty of KO is it is so incredibly easy. It takes minimal effort. In contrast, counting some cards as 2 points and some as 1 makes my head hurt just thinking about it.

So KO is easy and I get a little bored with it. Fine, I just add some of the additional index plays. Not quite the "Full Matrix KO", but a customized version, with some of my own indices and rounding, that works for me and captures most of the benefit. Chop out a few of the full matrix plays on the high positive and negative extremes and you still have a very easy-to-play yet powerful system.

Be sure to consider the simplicity of a system in which you make virtually no errors versus one more complicated. What will your error rate be? Will you still be able to monitor your surroundings, drink if you choose, chat with the PB, dealer, and ploppies while still having a good time playing?

Don't get me wrong, innovation is a good thing and I wish you well on this endeavor. Just a few things to thing about...
 
#4
UBZ-MIMO looks pretty good. I can suggest a few minor improvements -

1. Make the A value the same for 1-2D and then simply Ax2 for 6-8D.
2. Include the R value for 1-2D at +4 and again Rx2 for 6-8D
3. Use R value for certain stiffs also, like maybe 14v10 and 15v9&A (just guessing which ones but you get the idea.)

zg

Ps - I DO recommend splitting 10s!

------------------------------
FROM GRINDER'S WAREHOUSE -
UBZ11, DD, Specific vs Composite Indices
HERE - http://web.archive.org/web/20041221174949/frontpage.inxpress.net/grinder/simpage/ubz11ddindice.htm
 
Last edited:

nightspirit

Well-Known Member
#5
The table shows the index numbers for the conditions Mimosine gave in her post above. They were generated with the help of CVIndex. For DD I don`t know which pen. to use, so I picked 1.4/2. For 6 decks I think 4.75/6 is a good compromise. Would be nice if anybody could take a look on the numbers and validate it. So far I'm not so familiar with the software.

Sims for SD I haven't run so far. Maybe we can use AM numbers.

The next questions raised yet by Mimosine: in which range should we round the index numbers for simplicity? (if the table below is correct) For example, should we round numbers from -3 to +4 to zero (pivot).
And hands like 12 vs. 5, 12 vs. 6, 13 vs. 2 and three simple ignore?



Each column represents a new sim. The Late Surrender indexes were only generated in the last two columns. (The option was checked but I didn't include 15 vs. 9, A and 14 vs.10), so BS calls for LS 16 vs. 9,10,A and 15 vs. 10) I simply forgot about it in the first sims.
 

Mimosine

Well-Known Member
#6
nightspirit said:
The table shows the index numbers for the conditions Mimosine gave in her post above. They were generated with the help of CVIndex. For DD I don`t know which pen. to use, so I picked 1.4/2. For 6 decks I think 4.75/6 is a good compromise. Would be nice if anybody could take a look on the numbers and validate it. So far I'm not so familiar with the software.
that would be Mr. Mimosine to you :whip: :)

nevertheless i'm going to generate two strategy tables maybe later today using your sim as a reference. then we can see what the results are using your numbers, my simplified 2 rounded number chart, and a less rounded 5 number chart (with KK, pivot, IRC, Advantage, and Rogue numbers). i suspect they will probably all perform about the same. let's keep up the work on 6D for now, then get cranking on DD.

good work!


edit: P.S. it looks like for insurance decisions the pivot, i.e. 0 (tc = +2) will be perfect for ALL LATE SURRENDER!!! this would be great.
 
Last edited:
#7
Great stuff, with the Rogue number. One other thing you might want to add is the opposite of the Rogue number for SD and DD games where you are not Wonging out- a count so low that you are not doubling on 11, hitting a lot more stiffs, not splitting etc. Such indices have value in that they allow you to make the same amount of money while putting less money at risk.
 

Mimosine

Well-Known Member
#8
Automatic Monkey said:
Great stuff, with the Rogue number. One other thing you might want to add is the opposite of the Rogue number for SD and DD games where you are not Wonging out- a count so low that you are not doubling on 11, hitting a lot more stiffs, not splitting etc. Such indices have value in that they allow you to make the same amount of money while putting less money at risk.
an excellent suggestion.

i'm sure we will hit you up for your expertise to develop the SD & DD strategy - and will rely heavily on your UBZ-SV. i think starting with 6D is the way to go, as you have already formulated a lot with respect to sd and dd games, while 6D remains largely under-explored and remains the most viable game for someone at my betting level.
 

Mimosine

Well-Known Member
#9
Rounded Index Numbers for UBZ-OS (Unbalanced Zen - Open Source), 6D along with NightSpirit's numbers. In going for simplicity - one set of numbers for H17 or S17. These are the numbers I plan to sim when possible.

IRC = -24, tags as listed above.

Broken down:
1) Moron - all index plays at RC = 0,
except Insurance which is the most important index play, +4.

2) Simple - all index plays at -6 or +6 (Defined as the KeyCount and Advantage numbers).
Ins = +4
Splitting 10's = +10

3) Full - more complete set of rounded numbers for those inclined. Here we incorporate the IRC, Rogue and the Pivot (RC = 0, TC = +2). So you have 5 numbers to remember -24, -6, 0, +6, +10.

4) UBZ-NS NightSpirit's Sim numbers to a reasonable approximation across all the different games he simmed (e.g. H17, S17).


Code:
          Moron    Simple        Full    UBZ-NS

   INS        4         4         4         4

  12 v 2      0         6         6         4
  12 v 3      0         -6        0         -3
  12 v 4                          -6       -11
  12 v 5                         IRC       -24
  12 v 6                         IRC       -20
  13 v 2                         IRC       -18
  13 v 3                         IRC       -27
  16 v 9      0         6         10        10
 16 v 10      0         -6        -6       -12
 15 v 10      0         6         6         6

  8 v 5       0         6         6         7
  8 v 6       0         -6        0         -1
  9 v 2       0         -6        -6        -6
  9 v 7       0         6         6         7
 10 v 10      0         6         6         7
  10 v A      0         6         6         6
  11 v A      0         -6        -6        -6

 A,8 v 5      0         -6        0         -3
 A,8 v 6      0         -6        -6        -7

  XX v 5                10        10        11
  XX v 6                10        10        10

i plan on simming these with Powersim, once i read the manual a little more - i probably have to generate a bet spread to get real numbers. I'll probably do a 1-10 spread for all. It will take some time. if someone wants to crank these out using CVCX/CVData please do! :) i guess i'll also have to sim H17 and S17, but I won't sim LS for the time being.
 
Last edited:

Mimosine

Well-Known Member
#10
ROUND 1 sims:
Using PowerSim6
Decks: 6
Cards: 312
Players: 4
Shuffle Point: 234
Maximum Rounds per shoe (32767 = infinity): 32767
Burn Cards: 1
Maximum Dealer Rounds (0 = infinity): 0
Holecard(1) or Upcard(0) last: 1
European No Hole Card: 0
Maximum Split Hands: 4
Double After Splits: 1
Resplit Aces: 0
Extra Hits on Split Aces: 0
Double on hard 10 or 11 only: 0
Double on split aces: 0
Dealer hits soft 17: 0
Late Surrender: 0
Pick up hands from Last(1) to first(0): 1
Display hands: 0
Display shoes: 0
Strategy File for Player 1: UBZmoron.str
Strategy File for Player 2: UBZsimple.str
Strategy File for Player 3: UBZfull.str
Strategy File for Player 4: UBZns.str

Tag for 1: -1
Tag for 2: 1
Tag for 3: 2
Tag for 4: 2
Tag for 5: 2
Tag for 6: 2
Tag for 7: 1
Tag for 8: 0
Tag for 9: 0
Tag for 10: -2

Count is unbalanced
Initial Running Count: -24

(Crude 10-100 spread)
Bet Spread: -20: 10 -19: 10 -18: 10 -17: 10 -16: 10 -15: 10 -14: 10 -13: 10 -12: 10 -11: 10 -10: 10 -9: 10 -8: 10 -7: 10 -6: 20 -5: 20 -4: 30 -3: 30 -2: 40 -1: 40 0: 50 1: 50 2: 60 3: 60 4: 80 5: 80 6: 100 7: 100 8: 100 9: 100 10: 100 11: 100 12: 100 13: 100 14: 100 15: 100 16: 100 17: 100 18: 100 19: 100 20: 100

Results:

10000000 rounds Bankroll = 10000
1: $.133445/$22.2079 = .600889% var = 1539.35 SCORE = 11.56818 ROR = .176616
2: $.150129/$22.2079 = .676018% var = 1554.995 SCORE = 14.4944 ROR = .145014
3: $.152476/$22.2079 = .686587% var = 1557.115 SCORE = 14.93083 ROR = .141077
4: $.157337/$22.2079 = .708475% var = 1550.543 SCORE = 15.96537 ROR = .131408

not very impressive... i don't think.... comments? From the SCORE UBZsimple and UBZfull indicies are pretty good. I need to run this against KO with the same spread to see how it fairs.... var seems HUGE, but i don't really know if it is.... hmmm.
 

nightspirit

Well-Known Member
#11
Hey Mimo, had some time today and run a few sims for you. For 4.5/6 S17 DAS 2Billion rounds with CVCX

Code:
          Score
Moron     16.88
Simple    18.62
Full      18.89
ubz-NS    18.93

KO pref  16.40 
UBZII    16.71   (both canned sims)






The last picture shows a comparsion version of the canned sims vs. ubz-os-full. That UBZ-OS performs so well here is because KO doesn`t split the tens and both soft double are also not contained, I think. You won't be able to split the 10's all the time, so see this comparision with a grain of salt. Furthermore, we don't know how well UBZ-OS-full performes for other rules. That's the next step :)
 
Last edited:

nightspirit

Well-Known Member
#12
Today I resumed the 2 deck idea and ran some sims. I tried to round the index numbers according Mimosines plan, my previous sims and Monkeys SVUBZ. And thats what i used for 2 decks S17, DAS, LS 2000 billion rounds:

Code:
          index number
12 v. 2          2
12 v. 3          0
12 v. 4         -2
12 v. 5         -8
12 v. 6         -8
13 v. 2         -8
13 v. 3         -8
16 v. 9          2
16 v. 10       -8
15 v. 10        2

8 v. 5           2
8 v. 6           0
9 v. 2          -2
9 v. 7           2
10 v. 10        2
10 v. A         2
11 v. A         -2

A,8 v. 5        -2

A,8 v. 4         8
X,X v. 5         8
X,X v. 6         8

Surrender
15 v. 9           0
15 v. 10          0
14 v. 10          0


 
Last edited:

Knox

Well-Known Member
#14
Guys, I am on the verge of a software purchase for sims. I can't take it anymore, not having one. I want to be able to sim any custom system that I develop. Help me double check and make sure I buy the right software first time around. Is Qfit the only way to go and if so which program?

Thanks
 

nightspirit

Well-Known Member
#15
Knox said:
Guys, I am on the verge of a software purchase for sims. I can't take it anymore, not having one. I want to be able to sim any custom system that I develop. Help me double check and make sure I buy the right software first time around. Is Qfit the only way to go and if so which program?

Thanks
I only use Qfit's suite and don't know where other simulators have their advantages and disadvantage etc... But with Qfit's software I have all i ever need. And if you got problems or questions Norm's support is outstanding as well.

Among a lot of other things can you sim or redesign your custom systems with CVData and with CVCX you get quick the optimal bet ramp. I think Automaic Monkey, our "Lord of sims" ;) can give you a lot more information.
 

Mimosine

Well-Known Member
#16
NS,

Really good work. I ran some heavy duty powersim sims a few weeks back and was really disappointed with the results. I couldn't get any of the UBZs to outperform KO. I think i tracked it down to an error in my index play file, but was so fed up i put it on the back burner...

your results have rekindled my interest! that is a pretty big difference using simple UBZ vs. KO! wow!

a lot of my index numbers were rounded using intuition and an excel spreadsheet that tracks TC as a function of RC. ZG and others have suggested using rounded index numbers for 2D and SD that are HALF the value of 6D. I think this would work great and be easy to memorize... Insurance would have to be accurate of course since it is the most important.

but we could use -3 and +3 for the other index numbers, the IRC would stay the same for the H 12v4, 13v2, 13v3, and we would have to adjust TT V5,6 to a lower number i think maybe half of the rogue index number.... but more than +3.

does this jive with the results you're seeing?

i need to buy CVCX finally. powersim doesn't cut it, and finally i have a mac that can run windows so i can pump pump pump out some sims on my nice Duo 2.4GHzzzzoooooom!
 

nightspirit

Well-Known Member
#17
Mimosine said:
NS,

Really good work. I ran some heavy duty powersim sims a few weeks back and was really disappointed with the results. I couldn't get any of the UBZs to outperform KO. I think i tracked it down to an error in my index play file, but was so fed up i put it on the back burner...

your results have rekindled my interest! that is a pretty big difference using simple UBZ vs. KO! wow!
Thanks, you'r welcome! I also hope that I doesn't made a mistake with the settings. Would be nice if another member could cross check the results. I think to use CVData for such sims would garuantee a higher accuracy. Could be one method to check it.

Mimosine said:
a lot of my index numbers were rounded using intuition and an excel spreadsheet that tracks TC as a function of RC. ZG and others have suggested using rounded index numbers for 2D and SD that are HALF the value of 6D. I think this would work great and be easy to memorize... Insurance would have to be accurate of course since it is the most important.

but we could use -3 and +3 for the other index numbers, the IRC would stay the same for the H 12v4, 13v2, 13v3, and we would have to adjust TT V5,6 to a lower number i think maybe half of the rogue index number.... but more than +3.

does this jive with the results you're seeing?
Yes, sounds good! I must have overlooked your Key Count values.. (I realized that I confused A,8 vs. with A,8 v. 5 first :confused: ) I will substitute the -2 and +2 values with -3 and +3, the rogue value could be +10 for 6 decks and +5 for 2 decks. I only choosed +8 because of AM's risk averse indices and it would be easy to remember as +IRC. Anyway, I think your intuitive values will perform as well, if not better and I will try to run the sims according your idea in the next days.

Mimosine said:
i need to buy CVCX finally. powersim doesn't cut it, and finally i have a mac that can run windows so i can pump pump pump out some sims on my nice Duo 2.4GHzzzzoooooom!
Yeah, get it! With that processor you could produce sims a waaaayyy faster than I (I don't have the heart to mention which processor I use :sleep: ...).
 
Last edited:

nightspirit

Well-Known Member
#19
According Mimosine's idea for 2 decks, I substituted my original "2" index number from above with "3", made some other small changes and compared both with eachother. The "3" performs slightly better. But personally i would stick with the "2". See below.
Code:
                 index 
12 v. 2          3
12 v. 3          0
12 v. 4         -3
12 v. 5         -8
12 v. 6         -8
13 v. 2         -8
13 v. 3         -8
16 v. 9          3
16 v. 10        -8
15 v. 10         3

8 v. 5           3
8 v. 6           0
9 v. 2          -3
9 v. 7           3
10 v. 10         3
10 v. A          3
11 v. A         -3

A,8 v. 5        -3
A,8 v. 6        -3

X,X v. 5         8
X,X v. 6         8

Surrender
15 v. 9           0
15 v. A          -3
14 v. 10          0

insurance         1
(Dead link: http://img509.imageshack.us/my.php?image=ubzos2d3vs2ox6.jpg) _
_

Then I generated some index numbers for single deck. Depending on rules and penetration some will change slightly, but since we round them anyway, that doesn't matter much. I generated the same as for DD and 6D. I know it's not possible to use all because of rule restrictions like double on 10,11 only.
For S17
Code:
                 index 
12 v. 2          2
12 v. 3          1
12 v. 4         -1
12 v. 5         -3
12 v. 6         -2
13 v. 2         -2
13 v. 3         -3
16 v. 9          3
16 v. 10        -4
15 v. 10         1

8 v. 5           2
8 v. 6           1
9 v. 7          -1
9 v. 7           2
10 v. 10         1
10 v. A          1
11 v. A         -4

A,8 v. 5        -1
A,8 v. 6        -2

X,X v. 5         3
X,X v. 6         4

Surrender
15 v. 9           1
15 v. A          -2
14 v. 10          0  

insurance        1
If the dealer H17, hit on 12 vs. 6 when RC < -5 and double down 11 vs. A when RC > -4.
For handheld games I would choose only one strategy and use them for both. My suggestion would be:
Code:
                 index
12 v. 2          2
12 v. 3          0
12 v. 4         -2
12 v. 5         irc
12 v. 6         irc
13 v. 2         irc
13 v. 3         irc
16 v. 9          2
16 v. 10        irc
15 v. 10         2

8 v. 5           2
8 v. 6           0
9 v. 2          -2
9 v. 7           2
10 v. 10         2
10 v. A          2
11 v. A         -2

A,8 v. 5         0
A,8 v. 6        -2

X,X v. 5        +irc
X,X v. 6        +irc

Surrender
15 v. 9           0
15 v. A          -2
14 v. 10          0

insurance 
SD    0
DD    1
What do you think? To memorize 6 for 6 decks and 2 for 2 decks sounds fairly simply, or? :)
The next steps could be to develop optimal betting ramps for SD, DD and 6D. Then we could establish some group values ala KO. The cherry on the top would be to get some wonging and -out points with the TC version.:cool:
 
Last edited:

boneuphtoner

Well-Known Member
#20
A different approach

To NightSpirit and Mimosine,

You guys have done some great work. I simmed UBZ-OS-Full 6 decks with CVCX and was able to reproduce your results. Which surpassed KO and and the canned UBZ sims by a noticeable margin! Great job.

I've taken a slightly different approach. My goal was to come up with a single set of indices that would best the canned UBZ, KO, and Hi-Lo sims for SD, DD, six and eight decks. One set of indices for everything. I have sims to prove that I've done just that for SD-eight decks.

The UBZ-OS-Full six deck strategy bested my strategy for a six deck sim, which isn't suprising since my strategy was designed as a compromise for all numbers of decks. When I tried your six deck UBZ-OS-Full strategy with a two deck sim, my compromised strategy came out ahead. I have no doubt that your 2 deck strategy would beat my compromised strategy.

I'm using the same IRCs and insurance indexes as you guys...here are my SBA risk averse generated compromise indices:

16 v. 10 -5
15 v. 10 5
16 v. 9 10
12 v. 2 0
12 v. 3 0
12 v. 4 -5
12 v. 5 IRC
12 v. 6 IRC
13 v. 2 IRC
13 v. 3 IRC
9 v. 2 0
9 v. 7 10
8 v. 5 10
8 v. 6 0
10 v. 10 15
10 v. A 10
11 v. A 0
A,8 v. 5 0
A,8 v. 6 0
A,9 v. 5 10
A,9 v. 6 10
10,10 v. 5 10
10, 10 v. 6 10
9,9 v. A 10
9,9 v. 7 10
Surrender
15 v. 9 0
15 v. A 0
14 v. 10 0
14 v. 9 10
14 v. A 10
13 v. 10 15
16 v. 8 10

Could someone verify my results? For single deck, the only difference is all of the +10 indices become +5. That's it. It sims consistently better than the canned UBZ sims across the board. This also sims consistently better than the RPC and Mentor counts (with sweet 16, fab 3 indices) across the board.

Just opening yet another wrinkle in the discussion. The thing I like about this strategy is that there is only one set of indices to remember.
 
Top