Freddie: Halves vs HI-Lo

[stainless steel rat]

> Use $1 when making a general comparison. For
> a better comparison, pump up the bankroll to
> a million.

Do I look like Donald Trump?



Also, I suppose $1 would be ok, but I think it would piss any dealer off if you walk up and buy in and say give me $500 in silver. Just a few minutes sir while I get the floor supervisor to order me a fill in silver.

[Norm Wattenberger]

I?m afraid that you?re still missing the point. To accurately compare the efficacy of two strategies in general terms; you need to calculate the SCORE. SCORE is a theoretic evaluation. Adding constraints to make it more realistic invalidates the evaluation for the general case because it artificially puts you above or below the theoretic curve in an unpredictable manner. So, if you absolutely knew exactly the situation; your method would be better. But, if you want to solve for the general case; you need the real SCORE absent any betting constraints outside of spread.

[Brick Waller]

The double deck sim is way off and single deck seems to be out of line. 6 deck looks about right. Did you push some wrong buttons?

> OK.. This time I am ready.

> Please _read_ what I wrote. Here it is for
> reference:

> quote on----------------------
> I'm using CVCX and will let it choose
> everything optimally except that I will
> force it to use normal chip sizes (bets a
> multiple of $5) since I'm not going to play
> anywehre and bet $13.29 or some such insane
> amount
> quote off----------------------

> Now do you see anywhere where I say I am
> forcing any minimum bet size? Or do you see
> where I only said "use a multiple of $5
> chips" which could easily be a starting
> bet of $10 or $100 for that matter.

> Do you also see why it often gets to be a
> bit frustrating when you say "you can't
> do this" when "I didn't do it in
> the first place anyway?"

> Had you looked at the sim data I quoted, you
> would have discovered that to get those win
> rates (I'll take just the SD game here) the
> bets go from 15 to 60. Not 5 to anything.
> Ditto for the Halves data.

> Turns out there was an error in my data,
> because I had played with "wonging
> in/out" a while back and had left that
> checked, so that the numbers I quoted for
> both systems were wonging in and out at +1.

> Here is corrected data, and this time I will
> include the actual $ bet spreads rather than
> just 1-4 or 1-8.

> HiLo Halves
> SD 130/39.98 144/40.92

> ($15-$60) ($15-$60)
> DD 79/16.30 85/33.56

> ($5-$40) ($10-$80)
> 6D 68/24.88 110/29.26

> ($5-$100) ($5-$100)

> Note that I did _exactly_ that. And the
> above data (corrected) turns off the
> "wong in / wong out" option and
> makes it a play all for both. So there was
> an error, but it had nothing to do with bet
> size or bankroll size.

> And there are 1000 ways to make totally
> wrong assumptions, it seems. Even when a
> post specifically says that the only option
> used was to make bets a multiple of $5 which
> is what I have to do in any casino I have
> visited. I believe I can play at a $100 min
> table using red chips if I choose?

> I will be doing myself a service however, by
> pointing out that your perceived error in my
> post is simply 100% incorrect. I didn't
> "fix" the minimum bet other than
> to say it must be a multiple of $5. The
> numbers in () above are the actual bet
> sizes, rather than the original spreads I
> gave, which ought to eliminate that
> particular point from the discussion.

> You know, every now and then even you can
> make a mistake. I make plenty. But not the
> one you claimed above. Sorry.

[Brick]

Why is hi-lo's win rate taking the lead as pen increases?

Even though bets have been rounded off, as pen increases it would seem like any 3 level count system such as halves would be much more efficient and have a higher win rate than simple count systems.

[stainless steel rat]

> I?m afraid that you?re still missing the
> point. To accurately compare the efficacy of
> two strategies in general terms; you need to
> calculate the SCORE. SCORE is a theoretic
> evaluation. Adding constraints to make it
> more realistic invalidates the evaluation
> for the general case because it artificially
> puts you above or below the theoretic curve
> in an unpredictable manner. So, if you
> absolutely knew exactly the situation; your
> method would be better. But, if you want to
> solve for the general case; you need the
> real SCORE absent any betting constraints
> outside of spread.

I get that point. But what is the point of calculating a SCORE for a betting amount I can't possibly use in the _real_ games I used to compare things?

IF a SCORE comparison requires unlimited bet spread, unlimited small bet, a million dollar bankroll, chip sizes below $5, etc. I'm not sure it is so useful as a comparison measure.

Theory is nice. Reality is what I have to deal with at an actual table where all this "rubber" has to meet the road...

[stainless steel rat]

> The double deck sim is way off and single
> deck seems to be out of line. 6 deck looks
> about right. Did you push some wrong
> buttons?

No, the DD number is right. CVCX simply computes that for the optimal bankroll growth vs risk of ruin, halves will start the bet ramp at 2x the bet ramp for hi-lo. That is, the halves bets are exactly 2x the hilo bet unit, for the DD game only. That is why I personally don't like an unconstrained minimum bet, because as you can see, it can greatly distort things where the typical player is probably going to use the same spread and min bet no matter which counting system he plays by...

As far as SD and 6D goes, I checked and double-checked. Anyone with CVCX can do the same. The first numbers were wrong due to the wongin-wongout option being set on, but that was the only problem I really see with the comparison, the discussion about the minimum bet amount probably will not change the numbers enough to matter overall.

[Parker]

> I currently am pretty proficient with the
> Hi-Lo system, but wish to also learn the
> Halves system. I have a two part question.
> (1) Is the efficency percentage higher for
> the Halves system than the high-low and if
> it is so whether it will make a difference
> to my profit margin.

> (2) Since the high low system usually shows
> an extreme amount of variability throughout
> my bets if I was to burn a shoe and the
> count "sky rockets" does the
> halves system allow a player to
> "disguise" his play more
> efficiently while maintaining a optimal
> betting system?

Any comparison of systems should start with Blackjack Attack, 3rd Edition, by Don Schlesinger, specifically Chapter 9. Here we find a comparison of the SCORE's for various benchmark games for most of the popular counting systems. SCORE is explained in great detail in Chapter 9, but in a nutshell it is winrate (in dollars) per 100 hands at a given spread and penetration assuming a $10K bankroll and optimal betting.

For example, a double deck game (H17, DAS, 62/104 pen), 1-8 spread give us a SCORE of $39.38 for Hi-lo and $43.31 for Halves.

A six deck shoe (S17, DAS, LS), 5/6 decks pen, 1-12 spread play-all give us $52.04 for Hi-lo and $59.67 for halves. If we back-count, entering the game at TC +1 for either count, we get $90.21 for Hi-lo and $100.89 for Halves.

These gains are not insignificant but neither are they huge. Keep in mind that they assume optimal betting and perfect play. Are they worth the added complexity of a level 3 count? Only you can make that decision.

To answer your second question, I think that the cover benefits of any system are negligible. Most pit critters don't even know proper basic strategy, and any system will have you raising your bets when the count rises.

[Brick]

Below is your sim I was referring to.

"Here is corrected data, and this time I will include the actual $ bet spreads rather than just 1-4 or 1-8."

HiLo Halves
SD 130/39.98 144/40.92

($15-$60) ($15-$60)
DD 79/16.30 85/33.56
(
$5-$40) ($10-$80)
6D 68/24.88 110/29.26

NOTE:win rate for DD is over double the rate hi-lo is and you're showing a $15-$60 bet for both.

What does the first number represent 85?/33.56

I understand the point you're trying to make but many sims are set up for optimal bets to a set risk of ruin and bet spread when we compare count systems.

> minimum bet amount probably will not change
> the numbers enough to matter overall.

This is incorrect. If you really want to understand why halves is a better count system on all accounts,...lower the minumum bet to the same minimum bet as hi-lo,then look at how the win rate increases very significantly while risk of ruin decreases. It's the best of both worlds. The only difference is you're bet spread has increased,but NOT the maximum bet. So it's quite obvious which count system is better regardless if you're sims are telling you to bet more.

Good luck.

[Norm Wattenberger]

It is an interpolation to smooth the curve. Picking real games will not work as a valid comparison for the general case unless you pick every possible real game. SCORE gives you an accurate relative comparison without that exercise. The theoretic values are superior to real values when making a comparison for the general case.

> I get that point. But what is the point of
> calculating a SCORE for a betting amount I
> can't possibly use in the _real_ games I
> used to compare things?

> IF a SCORE comparison requires unlimited bet
> spread, unlimited small bet, a million
> dollar bankroll, chip sizes below $5, etc.
> I'm not sure it is so useful as a comparison
> measure.

> Theory is nice. Reality is what I have to
> deal with at an actual table where all this
> "rubber" has to meet the road...

[Norm Wattenberger]

Because the risk changes. When you are dealing with rational bets (i.e. not weird bets like $47) you end up over-betting and under-betting at certain penetratoins. This causes your win rate to go higher or lower but also increases or decreases your risk. HiLo is not performing better at those points. The EV was forced higher; but so is the risk. That is the problem with forcing the bets to reasonable numbers of chips in a comparison of two systems. It is not a valid comparison. If I had allowed any bet value to the dollar, the curves would have been smooth and Halves would always be better than HiLo.

> Why is hi-lo's win rate taking the lead as
> pen increases?

> Even though bets have been rounded off, as
> pen increases it would seem like any 3 level
> count system such as halves would be much
> more efficient and have a higher win rate
> than simple count systems.

[stainless steel rat]

> It is an interpolation to smooth the curve.
> Picking real games will not work as a valid
> comparison for the general case unless you
> pick every possible real game . SCORE gives
> you an accurate relative comparison without
> that exercise. The theoretic values are
> superior to real values when making a
> comparison for the general case.

I well understand the difference between "theory" and "reality". But do I really care if strategy A is better in the general case for all possible games, while strategy B is better in the 2 or 3 specific games I have the opportunity to play?

This is why I like CVCX. Now I don't have to deal with this "general case nonsense" I can ask for the specific rules, penetration and strategy I want to use, and get the real EV for that set of circumstances.

This is similar to DOD saying many years ago "ADA is superior for most programming applications, so we are no longer going to accept outside contracted programs written in any other language, and all of our internal programmers are going to have to use nothing but ADA for our projects."

Later they realized that there are better languages for certain applications, and that the "general case" theory is superceded by the "special case" circumstances at hand. Same here, IMHO. That's the reason I picked three different games and gave the specific rules and then win rate for each strategy. And offered to do the same for a specific game he might choose if none of those fit the bill...

[Norm Wattenberger]

> I well understand the difference between
> "theory" and "reality".
> But do I really care if strategy A is better
> in the general case for all possible games,
> while strategy B is better in the 2 or 3
> specific games I have the opportunity to
> play?

NO, you don't for your own personal use. And you are using CVCX correctly for your own personal use. But, when responding to the original post; the general case must be used. Because the poster doesn't care what strategy is better for YOU. That's the point.

[stainless steel rat]

> Below is your sim I was referring to.

> "Here is corrected data, and this time
> I will include the actual $ bet spreads
> rather than just 1-4 or 1-8."

see below...

> HiLo Halves
> SD 130/39.98 144/40.92

> ($15-$60) ($15-$60)
> DD 79/16.30 85/33.56
> (
> $5-$40) ($10-$80)

The above goes with the DD game. When I hit "submit" it was grouped two lines for SD, two lines for DD, two lines for 6D. Note that in the DD game above the hilo spread is 5-40, while the halves spread is 10-80. Because halves does enough better in that game that you can double your bets with a similar ROR (not identical but similar).

No idea how that got formatted so badly.

The data was supposed to look like this, repeated three times:

hiloscore/winrate halvesscore/winrate

(min - max bets) (min-max bets)

> 6D 68/24.88 110/29.26

> NOTE:win rate for DD is over double the rate
> hi-lo is and you're showing a $15-$60 bet
> for both.

No, somehow the post process seems to group them that way but if you notice, there are three sets of min/max bets, they appear to be above the score/win rate, but they are actually below, so the 15-60 numbers you are looking at go with the SD game... and 5-100 goes with the 6d game.

> What does the first number represent
> 85?/33.56

SCORE as computed by CVCX, which is a way of comparing the two counting methodologies.

> I understand the point you're trying to make
> but many sims are set up for optimal bets to
> a set risk of ruin and bet spread when we
> compare count systems.

That's what I did also. I told CVCX "use a BR of $3000, optimize for BR growth" and let it choose everything, bet min, bet max, bet ramp, and so forth." Only constraint I had set was "all bets in multiple of $5" since that is the way you have to play in a real casino, unless you know of one where you can get pennies, nickles, dimes, quarters in addition to the fifty-cent pieces most have already. And assuming you would really bet in those oddball amounts if you could.

Your idea makes perfect sense. Fix the ROR and BR, and see what the bet ramp looks like. Fix the BR and ramp and see what the ROR looks like. For those that simply want to compare SCORE, that is ok too. But SCORE is not the "complete story" because I can't look at it and discover anything about the reality of playing that game, while I can look at the BR, the bet spread, the ROR and use all of those to produce a mental picture of what is going on. SCORE reduces all of that to one number, which is fine for comparison at a gross level. But you are going to date a girl, the sum of her bust, waist, hips and age is 124. Are you going to date her? Oh, did I fail to mention she is 60 years old? That _is_ an important detail after all. Composites are good, specifics are sometimes better. I happen to like to pick a real game I play, and compare things for that game. Or I'll pick a real game you play and compare for you that same way. It tells me a lot more. SCORE may tell others more. Just not me...

> This is incorrect. If you really want to
> understand why halves is a better count
> system on all accounts,...lower the minumum
> bet to the same minimum bet as hi-lo,then
> look at how the win rate increases very
> significantly while risk of ruin decreases.

Don't do that. You'll get Don's attention. Been there, done that, got the T-shirt. He will immediately tell you "setting the min bet voids everything". I happen to believe that is the way _most_ players (not pros) bet with their current system, and it is interesting to try a new system using the exact same min bet to see how it would do. For the SD and 6D games, optimal min bet was the same for both systems. For halves, optimal min bet was 2x HiLo min bet. I let CVCX make the choices for those to avoid yet another long "that is the wrong way to compare" threads. Seems as though I have failed once again however.

But for fun, since I'm a "big boy" and can make up my own mind, here's the data.

I'll take the DD game above. I set the min chip size to $1 rather than $5. I forced the min bet to $5 for both.

hilo has an hourly win rate of $15.92 and a ROR of 5.1% with that $3K bankroll. Halves has an hourly win rate of $16.40 with a ROR of 4.4%. Clearly halves is better in that game. But from that comparison it is not _much_ better. Now back to the "CVCX chooses the min bet" (still using $1 min chip now so this won't match the previous results exactly, as min bet will only be some multiple of $1). Now halves has an hourly win rate of $26.37 with a ROR of 14.3%. Hilo has an hourly win rate of $22.25 with a ROR of 11.8%. Now you get to choose. $4.12 increase in hourly win rate (nearly 20% up), with a ROR increase of 2.5% (15% increase). min bets are different, max bets are therefore different, etc.

> It's the best of both worlds. The only
> difference is you're bet spread has
> increased,but NOT the maximum bet. So it's
> quite obvious which count system is better
> regardless if you're sims are telling you to
> bet more.

No disagreement. But read what you wrote, read what I wrote again, read what Don has written, and read what Norm has written. There are _many_ ways to compare two strategies. General case (theory). Specific game (reality). Specific min bet or optimal min bet. same bets, look at ROR, etc. You'd get the idea from reading here that most are invalid. Or that one way is the "right way". I happen to disagree. Each way of comparison sheds some light on the situation.

> Good luck.

Don't need good luck in BJ. Counting takes care of that. Need more luck posting here however.