MJ: CVCX Online TC Question

[MJ]

For the Hi-Lo I-18 Fab4 sim, the chart states that the EV at TC=1 equals .48% w/6D DAS DA2 LS 75% penetration. At TC=0 the chart states EV equals -.31%. This is a net gain of .79% in EV for only 1 unit of true count. This seems a bit on the high side. Shouldn't the EV be closer to .17% at TC=1?

It is a well known fact that each unit of true count is roughly equivalent to .50% in EV. The house advantage off the top is .33%. So, at TC=1 the EV would equal .5%-.33% = .17%. I realize the advantage per unit of true count is not linear, but it gives a good ballpark estimate of EV.

It seems as though the chart is overstating the advantage at TC = 1 by .48% - .17% = .31%. Can somebody please explain this?

Thanks,

MJ

[Norm Wattenberger]

The frequency differs substantially depending on whether you round, floor or truncate.

[Norm Wattenberger]

The first chart shows the advantage by True Count for Rounding, Truncating and Flooring. The second chart shows the gain in advantage for each TC. As you can see the "well known fact" is a great deal rougher estimate than most people think. Perhaps a lot of people would be better off if this well less well-known:-)






CVCX Online

[MJ]

> The first chart shows the advantage by True Count for
> Rounding, Truncating and Flooring. The second chart
> shows the gain in advantage for each TC. As you can
> see the "well known fact" is a great deal
> rougher estimate than most people think. Perhaps a lot
> of people would be better off if this well less
> well-known:-)

> CVCX Online

What are the rules for the charts? Why is the EV at TC=1 not .48%(for rounding)as indicated on CVCX Online?

-MJ

[Francis Salmon]

> For the Hi-Lo I-18 Fab4 sim, the chart states that the
> EV at TC=1 equals .48% w/6D DAS DA2 LS 75%
> penetration. At TC=0 the chart states EV equals -.31%.
> This is a net gain of .79% in EV for only 1 unit of
> true count. This seems a bit on the high side.
> Shouldn't the EV be closer to .17% at TC=1?

You ought to know that when Norm speaks of a TC+1 he doesn't mean exactly +1 he means the range defined by the used rounding technique.So for truncating or flooring the range is between +1 and +1.99 with an average of almost +1.5. For proper rounding the range is between +0.5 and +1.49 with an average of roughly +1.
So you're right, the advantage at +1 for rounding should indeed be around 0.2% and if it's given as 0.48% this is a clear mistake.However the figure of 0.48% makes sense for truncating or flooring because it represents a TC of +1.5.

> It is a well known fact that each unit of true count
> is roughly equivalent to .50% in EV. The house
> advantage off the top is .33%. So, at TC=1 the EV
> would equal .5%-.33% = .17%. I realize the advantage
> per unit of true count is not linear, but it gives a
> good ballpark estimate of EV.

I agree with that but you should keep in mind that the advantage increase also depends on the rules.The better the rules, the more increase you get.In your example the rules are very good and the increase per TC is probably a bit more than 0.5%. If you had rules like NDAS and ENHC the increase would be a bit less than 0.5%

Francis Salmon

[Francis Salmon]

> The first chart shows the advantage by True Count for
> Rounding, Truncating and Flooring.

It is interesting to see how the blue and the red line are practically parallel navigating at a distance of roughly 0.25%,the difference between a floored and a rounded TC being 0.5.
Claiming non-linearity by pointing to the green line is simply a joke. Using truncating in such a chart is grossly misleading because the 0-bucket is twice as large as the others and the distance between TC-1 and TC+1 is actually 3 and not 2 as the figures would suggest.

> The second chart
> shows the gain in advantage for each TC. As you can
> see the "well known fact" is a great deal
> rougher estimate than most people think. Perhaps a lot
> of people would be better off if this well less
> well-known:-)

As I said we should simply forget the green line but also the blue line is slightly biased at the 0-bucket because of the much more frequent occurence of the 0-count bringing down the average of that bucket to about 0.4 compared to 1.5 for the +1-bucket.
We all know about the insurance effect but the zig-zag line at the higher end still looks quite erratic and I think it has to do with your bucketing system.Also the higher counts will more likely occur late in the shoe which adds floating advantage.
If you want to have a neat test of linearity, you need to compare exact TCs at the same penetration level.The way you're doing it gives room to too many uncontrolled influences.

Francis Salmon

[Norm Wattenberger]

What are the rules for the charts? Why is the EV at TC=1 not .48%(for rounding)as indicated on CVCX Online?

4.5/6, S17, DAS, LS, half-deck resolution, full indexes. The CVCX Online HiLo sims are truncate, not round. The results match CVCX Online. Note: I only ran two billion rounds heads-up for these charts. CVCX Online sims are 20 billion rounds and four players.

MJ, please ignore Francis' posts. He plays differently than everyone else and his comments are simply not germane. My charts are designed to show how people actually play. He is constantly pointing out "errors" that don't exist; but we are tired of responding.


CVCX Online

[MJ]

> What are the rules for the charts? Why is the EV at
> TC=1 not .48%(for rounding)as indicated on CVCX
> Online? 4.5/6, S17, DAS, LS, half-deck resolution,
> full indexes. The CVCX Online HiLo sims are truncate,
> not round. The results match CVCX Online. Note: I only
> ran two billion rounds heads-up for these charts. CVCX
> Online sims are 20 billion rounds and four players.

Earlier in the month I asked you to clarify what method of TC calculation CVCX Online uses (Hi-Lo I-18 Fab4) and you said it was rounding. See 12/6/05 Computing for Counters. Now you are telling me it uses truncating. Did you simply make a mistake the first time I asked? It is ok if you did we all make mistakes.

Ok this makes more sense! So the EV at TC=1 represents an average TC of 1.5, which is for truncating. That would explain why the EV is a bit on the high side. According to the chart, the EV for rounding at TC=1 is right about where I thought it should be.

-MJ

[Norm Wattenberger]

> Earlier in the month I asked you to clarify what
> method of TC calculation CVCX Online uses (Hi-Lo I-18
> Fab4) and you said it was rounding. See 12/6/05
> Computing for Counters. Now you are telling me it uses
> truncating. Did you simply make a mistake the first
> time I asked? It is ok if you did we all make
> mistakes.

Yes I was caught in that post by your assumption that PBJ rounded. My attempt was to run every sim as they were conceived in their respective books. Unfortunately, the books rarely tell you. They all give obvious examples where rounding, truncating and flooring would yield the same results. And PBJ depends on the version of the book. Oddly, PBJ switched from Flooring in the original to Truncating. So it went from the new method to an old method. But, looking at the numbers it is truncating as in the newer version of PBJ.

>Ok this makes more sense! So the EV at TC=1
> represents an average TC of 1.5, which is for truncating.

Well, I wouldn't say that. First TC is what it is according to how it is calculated. Second, if you calculated TCs without rounding, flooring or truncating, then the group that encompasses truncated +1 would not average 1.5. It would have a median of 1.5 but an average of less than 1.5. Not that any of this matters What matters is how people actually count.

[Brick]

Norm,how difficult is it to simulate what my optimal bets would be at 1.5,2,2.5,3,3.5,etc.when rounding is used? In other words I would like to fine tune my betting strategy.

How much of a drop in EV would have appeared on your charts if you did not use full index numbers,but the basic Il 18 or 22?

In my opinion, if you would like simulations to be performed for practical purposes of real world conditions, I would simply use the basic index numbers. Is there any particular reason why you chose full indexes?

thanks,
Brick

[Norm Wattenberger]

> Norm,how difficult is it to simulate what my optimal
> bets would be at 1.5,2,2.5,3,3.5,etc.when rounding is
> used? In other words I would like to fine tune my
> betting strategy.

This could be done by doubling the tags and count values. Or more easily, probably by using the custom TC calculation feature to divide by double decks. The gain is not much. There is a large drop in gain by using half-deck divisors in level 1 strategies or quarter-deck divisors in level 2 strategies. But, the gain in switching from half-deck to full-deck divisors is tiny in a level 2 strategy. This has the same effect as what you are proposing - refining the bets to half TCs.

> How much of a drop in EV would have appeared on your
> charts if you did not use full index numbers,but the
> basic Il 18 or 22?

> In my opinion, if you would like simulations to be
> performed for practical purposes of real world
> conditions, I would simply use the basic index
> numbers. Is there any particular reason why you chose
> full indexes?

The charts would have looked about the same. You can use CVCX Online to see the differences.


CVCX Online

[Francis Salmon]

It's sad to see how Norm fails to answer to the point and dismisses any logical argument.
It seems that he doesn't like people to understand his charts and their fallacies .That's why he tells you not to listen to me.But it's enough if you keep listening to your sense of logic.

Francis Salmon

[Francis Salmon]

> MJ, please ignore Francis' posts. He plays differently
> than everyone else and his comments are simply not
> germane. My charts are designed to show how people
> actually play. He is constantly pointing out
> "errors" that don't exist; but we are tired
> of responding.

The way I play is not the subject here nor is it other people's way of playing that you pretend to know so well.
We are talking about your truncating chart that's allegedly supporting your claim of non-linearity especially in the range between TC-1 and TC+1.This is wrong and here is why:
The 0-bucket contains all TCs >-1 and <+1 with an average of 0.Right?The +1-bucket contains all TCs>=1 and <2.The average here is 1.5 or slightly less if you like.That means that the distance between the bucket labelled 0 and the bucket labelled +1 is not only 1 as one would assume but almost 1.5.So the increase in advantage here of 0.81% has to be multiplied by 2/3 which brings us back to 0.54% per TC. This is in line with the rest and looks quite reasonnable for these excellent rules.Voil?! You may continue to ignore this blatant truth but I'm sure the readers won't.

Francis Salmon