Gorilla Player: question for Don, et al..

[Gorilla Player]

Three terms. BC, PE, IC. All well-publicized, but I can't find one important piece of information. An example to illustrate my question:

Suppose I flat-bet for at least N0 hands. IE I don't vary ever. It would seem that my BC would be fairly high anyway because about 1/3 of the time I will be in a negative count and bet the min were I counting, 1/3 of the time I would be in a neutral count and also bet the min, so about 1/3 of the time flat-betting would be "low". My question is, therefore, what is the BC for flat-betting overall? My reason for asking is simply to have some idea of how the BC of .98 (+/1 .01 depending on where you read) for HiLo compares to flat-betting.

Same question for PE. IE what is the PE for a player that plays perfect BS? I've seen Hi-Lo PE numbers of .51 to .54, again depending on where you read, but if I weren't counting, what would my PE be with perfect BS, no indices, no counting, just the correct BS card handy for the game being played?

Finally, same question for IC as well. A BS player would never buy insurance, so what is the IC for that player.

Reason for the question is simply to try to develop some sort of idea of what a PE of .5 vs a PE of .55 really means. Obviously .55 is better, but it has never been clear to me exactly _how_ much better.

If the question makes any sense, of course. And if this is available online somewhere (I haven't found it) a simple link to it will be just fine...

Thanks...

GP

[Don Schlesinger]

> Three terms. BC, PE, IC. All
> well-publicized, but I can't find one
> important piece of information. An example
> to illustrate my question:

> Suppose I flat-bet for at least N0 hands. IE
> I don't vary ever. It would seem that my BC
> would be fairly high anyway because about
> 1/3 of the time I will be in a negative
> count and bet the min were I counting, 1/3
> of the time I would be in a neutral count
> and also bet the min, so about 1/3 of the
> time flat-betting would be "low".
> My question is, therefore, what is the BC
> for flat-betting overall? My reason for
> asking is simply to have some idea of how
> the BC of .98 (+/1 .01 depending on where
> you read) for HiLo compares to flat-betting.

You can't do anything about your BC by the way you bet! It's predetermined for a count, irrespective of the bets you make. BC simply tells you how accurate your count is in assessing positive-edge opportunities. In other words, it gives the percentage of the time that your count identifies the correct advantage. For example, Hi-Lo does a pretty good job, but it doesn't count the 7s. If many 7s came out, you would have a betting edge, but Hi-Lo would fail to recognize it. And, it will be doing that no matter what you're betting.

> Same question for PE. IE what is the PE for
> a player that plays perfect BS?

PE works the same way. It gives your count's ability to reflect when you should be making a playing departure. It doesn't know how you're going to play. So, if you play BS all the time, you're wasting your count's PE, because you're not using it to make departures. But the PE is there nonetheless.

> I've seen
> Hi-Lo PE numbers of .51 to .54, again
> depending on where you read, but if I
> weren't counting, what would my PE be with
> perfect BS, no indices, no counting, just
> the correct BS card handy for the game being
> played?

See above. You're misunderstanding the meaning of the terms.

> Finally, same question for IC as well. A BS
> player would never buy insurance, so what is
> the IC for that player.

Same comment. Counts have ICs -- their ability to recognize when taking insurance is the proper play. If you don't follow the advice, it isn't the count's fault! :-)

> Reason for the question is simply to try to
> develop some sort of idea of what a PE of .5
> vs a PE of .55 really means. Obviously .55
> is better, but it has never been clear to me
> exactly _how_ much better.

Can't answer that, as it depends entirely what games you play and how you bet. So, you really can't answer questions like those.

> If the question makes any sense, of course.
> And if this is available online somewhere (I
> haven't found it) a simple link to it will
> be just fine...

Richard Reid's bjmath.com site has some discussion of all this.

Don

[Gorilla Player]

> You can't do anything about your BC by the
> way you bet! It's predetermined for a count,
> irrespective of the bets you make. BC simply
> tells you how accurate your count is in
> assessing positive-edge opportunities. In
> other words, it gives the percentage of the
> time that your count identifies the correct
> advantage. For example, Hi-Lo does a pretty
> good job, but it doesn't count the 7s. If
> many 7s came out, you would have a betting
> edge, but Hi-Lo would fail to recognize it.
> And, it will be doing that no matter what
> you're betting.

I understand that. My query was about the BC for flat-betting, since a majority of the time the minimum bet is the right bet anyway...

Perhaps the experiment: I'm going to flat bet, and for each hand, the question is asked "was that bet appropriate for the remaining cards yet to be played?" It would seem that for a flat bettor, the correlation between his bet (always a min bet) and his advantage would be fairly good since we don't have an advantage in most of the hands played...

> PE works the same way. It gives your count's
> ability to reflect when you should be making
> a playing departure. It doesn't know how
> you're going to play. So, if you play BS all
> the time, you're wasting your count's PE,
> because you're not using it to make
> departures. But the PE is there nonetheless.

Again, I understand. I suspect I didn't word my question very clearly. What I am trying to do is establish a base line PE for the basic strategy player. IE lots of places (even the qfit web site) have a table showing the BC, PE and IC for lots of counting systems. I just wanted to know how the worst of the counting systems compare with the pure BS player that doesn't modify his playing whatsoever from BS.

Another approach: When I talk about demand paging in my CS operating system courses, we look at page replacement strategies. The question that always comes up is "how good is this particular strategy (least recently used, least frequently used, FIFO, etc.) What I do to answer that question is to take what would clearly be the worst acceptable strategy possible, pure random replacement, and use that to define the worst-case replacement we'd accept, because if any replacement strategy was worse than random, we'd just throw it out and use random since that is trivial to implement. Next I try to give the best possible strategy, one that uses future events to make present decisions. Can't be done in reality, but in a lab it is easy to do this. Now I have a worst-case number, a best-case number, and I have some feel for where the particular strategy I am looking at fits into the "great scheme of things."

That's what I am looking for here. IE obviously the worst-case would be a pure BS player since anyone should be able to do that. Best case is probably a level-10 counting system that very accurately measures the card-removal effect for each individual type of card. If I had the BC, PE and IC for the best and worst, I would have some idea of where the various counting strategies fit in.

I hate the Einsteinian approach of "everything is relative". I want something to compare things to. IE I can plot things naturally, or logarithmically, or on any compressed scale I want, and you can still see which is better, but it becomes hard to know what the difference between 1 and 2 is. If it is miles, I can walk it. If it is parsecs, I have a bit of a problem.

> See above. You're misunderstanding the
> meaning of the terms.

> Same comment. Counts have ICs -- their
> ability to recognize when taking insurance
> is the proper play. If you don't follow the
> advice, it isn't the count's fault! :-)

> Can't answer that, as it depends entirely
> what games you play and how you bet. So, you
> really can't answer questions like those.

> Richard Reid's bjmath.com site has some
> discussion of all this.

> Don

Perhaps either (a) I asked the question poorly, or (b) I asked something that simply makes no sense. Let me cogitate on it a bit. It made sense when I first asked it, at least to me.

But for starters, on the BC question, it would seem that it is possible to compute the correlation between flat-betting and optimal-betting, and get a number between 0 and 1.0. Were I guessing, I would guess 70% since it seems that about 30% of the time I see a count that justifies raising my bet. That's just a guess of course, but based on my number of shoes/decks played, it seems to be fairly reasonable.

GP

[Sun Runner]

With reference to Gorilla Player's questions, I have had some of the same wonderings.

In my beginning, I judged systems strictly on BE, PE, and IC. Let's say RPC has a PE of .55 and HiLo a PE of .51. Is it fair to say that RPC is 7.8% stronger than HiLo (.04/.51)? Seems to me it can't be measured that simply, correct?

Now I try to look at SCORE first. In a certain 6D game, Cacarulo gave a c-SCORE to RPC of 47.80 and TKO/A 48.96. In this example would it be fair to say that TKO/A is 2.4% stronger (1.16/47.80) in this example, for that game?

Regarding GP's post specifically -could not simple BS play be calc'd a PE (and BE?) as it would by default be locating the correct play XX.XX% of the time?

Thanks.

[Don Schlesinger]

> I understand that. My query was about the BC
> for flat-betting, since a majority of the
> time the minimum bet is the right bet
> anyway...

Not too interesting your way!

> Perhaps the experiment: I'm going to flat
> bet, and for each hand, the question is
> asked "was that bet appropriate for the
> remaining cards yet to be played?" It
> would seem that for a flat bettor, the
> correlation between his bet (always a min
> bet) and his advantage would be fairly good
> since we don't have an advantage in most of
> the hands played...

I really don't get your point. What you're really asking is: For a given count, such as Hi-Lo, what percentage of the time will the counter who spreads, rather than flat bets, be betting more than one unit? That information is easily attainable from the Chapter 10 charts and varies for every conceivable type of game that is catalogued there.

And, if one were back-counting, the counter would always be betting more than the flat-betting BS player, so the percentage you're looking for would be zero, if I understand your question correctly.

> Again, I understand. I suspect I didn't word
> my question very clearly. What I am trying
> to do is establish a base line PE for the
> basic strategy player. IE lots of places
> (even the qfit web site) have a table
> showing the BC, PE and IC for lots of
> counting systems. I just wanted to know how
> the worst of the counting systems compare
> with the pure BS player that doesn't modify
> his playing whatsoever from BS.

I must not be understanding. A system's PE tells us how well it detects those situations where we are to depart from BS. If you never depart from BS, what would you want to call the PE of such a system?

> Another approach: When I talk about demand
> paging in my CS operating system courses, we
> look at page replacement strategies. The
> question that always comes up is "how
> good is this particular strategy (least
> recently used, least frequently used, FIFO,
> etc.) What I do to answer that question is
> to take what would clearly be the worst
> acceptable strategy possible, pure random
> replacement, and use that to define the
> worst-case replacement we'd accept, because
> if any replacement strategy was worse than
> random, we'd just throw it out and use
> random since that is trivial to implement.
> Next I try to give the best possible
> strategy, one that uses future events to
> make present decisions. Can't be done in
> reality, but in a lab it is easy to do this.
> Now I have a worst-case number, a best-case
> number, and I have some feel for where the
> particular strategy I am looking at fits
> into the "great scheme of things."

> That's what I am looking for here. IE
> obviously the worst-case would be a pure BS
> player since anyone should be able to do
> that. Best case is probably a level-10
> counting system that very accurately
> measures the card-removal effect for each
> individual type of card.

Griffin showed that, for single-parameter counting systems, beyond level 4, you're just wasting your time. Level-10 won't help.

> If I had the BC, PE
> and IC for the best and worst, I would have
> some idea of where the various counting
> strategies fit in.

See Griffin, pp. 45-47.

> I hate the Einsteinian approach of
> "everything is relative". I want
> something to compare things to. IE I can
> plot things naturally, or logarithmically,
> or on any compressed scale I want, and you
> can still see which is better, but it
> becomes hard to know what the difference
> between 1 and 2 is. If it is miles, I can
> walk it. If it is parsecs, I have a bit of a
> problem.

Frankly, none of this is as effective as the SCORE methodology. When you will have plotted your PEs and your BCs, to compare, and you will attempt to form some kinds of ratios to measure strengths of systems, you will never find that those ratios are the same as the ratios of the SCOREs that they produce when apples-to-apples comparisons are made. So, one would have to ask what you intend to do with the information you will have.

> Perhaps either (a) I asked the question
> poorly, or (b) I asked something that simply
> makes no sense. Let me cogitate on it a bit.
> It made sense when I first asked it, at
> least to me.

It doesn't make a great deal of sense to me, but maybe I'm missing something.

> But for starters, on the BC question, it
> would seem that it is possible to compute
> the correlation between flat-betting and
> optimal-betting, and get a number between 0
> and 1.0. Were I guessing, I would guess 70%
> since it seems that about 30% of the time I
> see a count that justifies raising my bet.

See above.

> That's just a guess of course, but based on
> my number of shoes/decks played, it seems to
> be fairly reasonable.

Different for every game. And, not terribly interesting, I'm afraid.

Don

[Don Schlesinger]

> With reference to Gorilla Player's
> questions, I have had some of the same
> wonderings.

> In my beginning, I judged systems strictly
> on BE, PE, and IC. Let's say RPC has a PE of
> .55 and HiLo a PE of .51. Is it fair to say
> that RPC is 7.8% stronger than HiLo
> (.04/.51)?

No.

>Seems to me it can't be measured that simply, correct?

Correct.

> Now I try to look at SCORE first. In a
> certain 6D game, Cacarulo gave a c-SCORE to
> RPC of 47.80 and TKO/A 48.96. In this
> example would it be fair to say that TKO/A
> is 2.4% stronger (1.16/47.80) in this
> example, for that game?

Yes.

> Regarding GP's post specifically -could not
> simple BS play be calc'd a PE (and BE?) as
> it would by default be locating the correct
> play XX.XX% of the time?

Possibly for BC, but not for PE. For betting, using a flat bet is correct part of the time (by default) in much the same manner that a stopped clock is "right" twice a day. But for PE, by definition, we're examining what percentage of the time our system recognizes when it is correct to depart from BS. That being the case, by your definition, the PE of the basic strategist would be zero.

Don

[Sun Runner]

[gorilla player]

> Not too interesting your way!

> I really don't get your point.

Sorry. Even being a faculty member doesn't always mean I can communicate properly.

Here is my quandry:

I can compare two things, by giving you two numbers, and telling you "bigger is better".

But if I tell you A is .51 and B is .55, you would conclude A is better. But how _much_ better? I want some known point on the number scale that I can "grasp". IE it would be pretty gross of me to come up with a new temperature measure, and when you ask me whether the coffee or the soup is hotter, and I just say "coffee = 1450 degrees GP scale, soup = 1477 degrees GP scale." The "burning" question I would want answered is "OK, just how the hell hot is 1450 degrees GP scale?" If I give some index points on the "GP scale" for freezing water, boiling water, etc" then one could conclude whether to drink the soup or the water. Given the choice of soup at 5 degrees C or coffee at 6 degrees C, and I am hunting for something to warm me up, I can tell that the coffee is hotter, but how _much_ hotter.

Hence my question about doing an _accurate_ measure of a player's advantage based on perfect knowledge of card removal, and then seeing what percent of the time the player's advantage is low enough that the flat-bettor is doing the right thing... IE if I simply were to flat-bet and not count at all, serendipity would have me betting the right amount simply because I always bet the min and sometimes the min is correct...

Ditto for a BS player's "playing efficiency" as opposed to a card counter's doing BS deviations based on the count...

And finally for insurance, a BS player never takes insurance, so how well does that correlate to whether he should take it or not based again on the "perfect information" the BS player doesn't have nor use. IE by always not talking it, he makes the right decision every time the TC is < 3, so his "always reject" has some correlation to the correct decision...

> What you're
> really asking is: For a given count, such as
> Hi-Lo, what percentage of the time will the
> counter who spreads, rather than flat bets,
> be betting more than one unit?

Not quite. IE a flat bettor will also play pure BS. No indices either. So his BC will be lower than most any counter's of course, and his PE will be lower as well. What I'm looking for is _how much_ lower is the pure BS player, compared to players that use a real counting system, and use it correctly.

Again, the pure BS flat-bettor provides a reference point against which the others can be compared...

That
> information is easily attainable from the
> Chapter 10 charts and varies for every
> conceivable type of game that is catalogued
> there.

> And, if one were back-counting, the counter
> would always be betting more than the
> flat-betting BS player, so the percentage
> you're looking for would be zero, if I
> understand your question correctly.

> I must not be understanding. A system's PE
> tells us how well it detects those
> situations where we are to depart from BS.
> If you never depart from BS, what would you
> want to call the PE of such a system?

OK. Maybe we are getting to the crux. I (again, I am a 4-year counter that can count cards well and follow the indices and do the betting right, but I've not spent a lot of time dealing with the probability and statistical part of this very much at all) was simply thinking about answering this question, with a perhaps flawed definition of PE.

first, PE I was taking as how often does my count tell me the playing decision that matches the best possible play based on the remaining cards. IE for any hand and remaining deck, there is a single play that is correct most of the time, this can be derived obviously by simulation. So if Hi-Lo + I18 + F4 is ".xxx PE" then what is the "PE" for the pure BS player?"

Sorry but I am probably mangling the terminology, since I had always assumed PE was a measure of how close the TC BS index variations + normal BS when there is no index was approximating "perfect play" based on a large number of simulations... IE for Hi-Lo I often see .51... But I'd like to know what my PE drops to if I just play pure BS with _no_ indices, not that I would ever do that of course. But I want to know how hot the damned soup is before I drink it and either fry my tongue, or find it unpalatably cold...

> Griffin showed that, for single-parameter
> counting systems, beyond level 4, you're
> just wasting your time. Level-10 won't help.

> See Griffin, pp. 45-47.

> Frankly, none of this is as effective as the
> SCORE methodology. When you will have
> plotted your PEs and your BCs, to compare,
> and you will attempt to form some kinds of
> ratios to measure strengths of systems, you
> will never find that those ratios are the
> same as the ratios of the SCOREs that they
> produce when apples-to-apples comparisons
> are made. So, one would have to ask what you
> intend to do with the information you will
> have.

> It doesn't make a great deal of sense to me,
> but maybe I'm missing something.

> See above.

> Different for every game. And, not terribly
> interesting, I'm afraid.

> Don

[gorilla player]

> No.

> Correct.

> Yes.

> Possibly for BC, but not for PE. For
> betting, using a flat bet is correct part of
> the time (by default) in much the same
> manner that a stopped clock is
> "right" twice a day. But for PE,
> by definition, we're examining what
> percentage of the time our system recognizes
> when it is correct to depart from BS. That
> being the case, by your definition, the PE
> of the basic strategist would be zero.

> Don

OK. I think that is part of the problem, in that I was using a term wrongly. IE a pure BS player has no "PE" based on your clearly written last statement. I was assuming that PE was a measure of _overall_ playing efficiency, but in thinking, that obviously was wrong since 80% of the time (roughly) indices probably don't apply making a PE of .5 tough to get.

So forget the PE for BS... Then do I interpret PE as that in 51 percent of the cases where a BS deviation is correct, HiLo will get 'em?

So many terms, so many ways to interpret them wrongly without knowing.

[Don Schlesinger]

> OK. I think that is part of the problem, in
> that I was using a term wrongly.

I tried patiently to point that out. :-)

> IE a pure
> BS player has no "PE" based on
> your clearly written last statement.

Correct.

> I was
> assuming that PE was a measure of _overall_
> playing efficiency, but in thinking, that
> obviously was wrong since 80% of the time
> (roughly) indices probably don't apply
> making a PE of .5 tough to get.

Right. It's a measure of how often, when you are supposed to depart from BS, your count system detects that departure and tells you the correct thing to do.

> So forget the PE for BS... Then do I
> interpret PE as that in 51 percent of the
> cases where a BS deviation is correct, HiLo
> will get 'em?

Yes, exactly.

> So many terms, so many ways to interpret
> them wrongly without knowing.

You're forgiven.

Don

[gorilla player]

> I tried patiently to point that out. :-)

> Correct.

> Right. It's a measure of how often, when you
> are supposed to depart from BS, your count
> system detects that departure and tells you
> the correct thing to do.

> Yes, exactly.

> You're forgiven.

> Don

I'm also a bit wiser now. As I now see that pure BS is 0.00 on the "chart", which gives context to the .51 or .54 or whatever PE numbers

[Sun Runner]

It is my understanding that at best PE will never reach past .70.

What is it that causes the limitation?

[BTW, to Gorilla Player, Canfield's Master system carries an approx .67 .. maybe worth crowing about .. at the time .. ya think? (All in good fun.) ]

[Don Schlesinger]

> It is my understanding that at best PE will
> never reach past .70.

> What is it that causes the limitation?

The complexity of the interaction of various rank depletions on strategic decisions (wow, that was a mouthful!). single-parameter count systems just aren't equipped to handle this; they need multiple side counts to be able to improve PE.

Don