S17 vs. H17 at High True Counts

[Southpaw]

Introduction:

A while back I read a post claiming that S17 was equal to H17 (in terms of advantage I suppose they meant) at high TC's. Although this is likely true for extremely high TC's, I was quite skeptical of the statement as it pertained for high TC's that one will actually see occasionally during real play. I intend to investigate this further here.

Methodology:

I decided that I would choose a base game and compare the TBA at various TC's when the game was S17 and H17, all other things held constant. I arbitrarily selected the following conditions:

Six Decks, Split to 4, No RSPA, 3:2 BJ, DOA2, LS, Pen = 1.25 cut off, DAS, Head-On.

Variable Rule: S17 / H17

Simulation Parameters:

Fixed Cut-Card, 10 Billion Rounds, 1-Burn Card, Hi-Lo with the I18 and Fab 4 (S17 indices were used for the S17 game while H17 indices were used for the H17 game), Full-Deck Resolution, TC divisor = Cards in tray, Truncated TC, Rounded Deck Estimation.

Results:

Reported are the TBA's for the simulations with the said conditions under S17 / H17 conditions at each TC (between 0-10). Also, reported are the standard errors involved in the TBA calculations at each TC (0-10).

https://docs.google.com/viewer?a=v&pid=explorer&chrome=true&srcid=0B0cCldUn36hMYzcxYzM4MGEtMzJkYi00Y2I5LWE5MzYtYzc1YjY3YjdhZjZl&hl=fr

Best,

SP

 

[creeping panther]

Sp

Introduction:

A while back I read a post claiming that S17 was equal to H17 (in terms of advantage I suppose they meant) at high TC's. Although this is likely true for extremely high TC's, I was quite skeptical of the statement as it pertained for high TC's that one will actually see occasionally during real play. I intend to investigate this further here.

Methodology:

I decided that I would choose a base game and compare the TBA at various TC's when the game was S17 and H17, all other things held constant. I arbitrarily selected the following conditions:

Six Decks, Split to 4, No RSPA, 3:2 BJ, DOA2, LS, Pen = 1.25 cut off, DAS, Head-On.

Variable Rule: S17 / H17

Simulation Parameters:

Fixed Cut-Card, 1-Billion Rounds, 1-Burn Card, Hi-Lo with the I18 and Fab 4 (S17 indices were used for the S17 game while H17 indices were used for the H17 game), Full-Deck Resolution, TC divisor = Cards in tray, Truncated TC, Rounded Deck Estimation.

Results:

Reported are the TBA's for the simulations with the said conditions under S17 / H17 conditions at each TC (between 0-10). Also, reported are the standard errors involved in the TBA calculations at each TC (0-10).

https://docs.google.com/leaf?id=0B0cCldUn36hMMmI1MGRlNWQtY2YyZi00Y2UzLWJhYjEtMzhhNTUyMGZmYWRh&hl=fr

Conclusions:

Even at TC's as high as +10, the S17 game is still better than the H17 one by ~0.10%, which is about half of what it is usually reported as (0.20%). At a TC of +5, which is when most will push their max bet out, the S17 game is still better than the H17 game by ~0.14%.

While the difference between the two games may seem small at high TC's, it must be remembered that many are willing to all sorts of crazy things to obtain a "theoretical" edge similar to this (i.e., level 2 or 3 systems and side-counting aces), although in practice the edge gained from such tactics may very well be lower than this.

Best,

SP

Good job !

Your teammate,
CP

 

[21forme]

At TC=9, the H17 game has a 0.1% edge. What this is telling me is that even with 1 billion rounds, your sample size is too small to produce valid results at these TCs.

What would also be interesting is running it for a DD game, where TCs of 9 or 10 are not that uncommon.

 

[jack.jackson]

Introduction:

A while back I read a post claiming that S17 was equal to H17 (in terms of advantage I suppose they meant) at high TC's. Although this is likely true for extremely high TC's, I was quite skeptical of the statement as it pertained for high TC's that one will actually see occasionally during real play. I intend to investigate this further here.

Methodology:

I decided that I would choose a base game and compare the TBA at various TC's when the game was S17 and H17, all other things held constant. I arbitrarily selected the following conditions:

Six Decks, Split to 4, No RSPA, 3:2 BJ, DOA2, LS, Pen = 1.25 cut off, DAS, Head-On.

Variable Rule: S17 / H17

Simulation Parameters:

Fixed Cut-Card, 1-Billion Rounds, 1-Burn Card, Hi-Lo with the I18 and Fab 4 (S17 indices were used for the S17 game while H17 indices were used for the H17 game), Full-Deck Resolution, TC divisor = Cards in tray, Truncated TC, Rounded Deck Estimation.

Results:

Reported are the TBA's for the simulations with the said conditions under S17 / H17 conditions at each TC (between 0-10). Also, reported are the standard errors involved in the TBA calculations at each TC (0-10).

https://docs.google.com/leaf?id=0B0cCldUn36hMMmI1MGRlNWQtY2YyZi00Y2UzLWJhYjEtMzhhNTUyMGZmYWRh&hl=fr

Conclusions:

Even at TC's as high as +10, the S17 game is still better than the H17 one by ~0.10%, which is about half of what it is usually reported as (0.20%). At a TC of +5, which is when most will push their max bet out, the S17 game is still better than the H17 game by ~0.14%.

While the difference between the two games may seem small at high TC's, it must be remembered that many are willing to all sorts of crazy things to obtain a "theoretical" edge similar to this (i.e., level 2 or 3 systems and side-counting aces), although in practice the edge gained from such tactics may very well be lower than this.

Best,

SP

Ya Good Job! I've I've actually heard that it averages out to about .16% for the counter verses .20% for the BS player. But what im curious about is re-splitting aces. How much more does this help the counter verses the BS player? It seems to me(all the things equal) that its possible to have a higher WR with rules that benefit the counter, verses rules that arent so counter friendly, despite having a higher HE. Is this possible regardless of the rules in effect?

 

[Renzey]

Introduction:

A while back I read a post claiming that S17 was equal to H17 (in terms of advantage I suppose they meant) at high TC's. Although this is likely true for extremely high TC's, I was quite skeptical of the statement as it pertained for high TC's that one will actually see occasionally during real play. I intend to investigate this further here.

Methodology:

I decided that I would choose a base game and compare the TBA at various TC's when the game was S17 and H17, all other things held constant. I arbitrarily selected the following conditions:

Six Decks, Split to 4, No RSPA, 3:2 BJ, DOA2, LS, Pen = 1.25 cut off, DAS, Head-On.

Variable Rule: S17 / H17

Simulation Parameters:

Fixed Cut-Card, 1-Billion Rounds, 1-Burn Card, Hi-Lo with the I18 and Fab 4 (S17 indices were used for the S17 game while H17 indices were used for the H17 game), Full-Deck Resolution, TC divisor = Cards in tray, Truncated TC, Rounded Deck Estimation.

Results:

Reported are the TBA's for the simulations with the said conditions under S17 / H17 conditions at each TC (between 0-10). Also, reported are the standard errors involved in the TBA calculations at each TC (0-10).

https://docs.google.com/leaf?id=0B0cCldUn36hMMmI1MGRlNWQtY2YyZi00Y2UzLWJhYjEtMzhhNTUyMGZmYWRh&hl=fr

Conclusions:

Even at TC's as high as +10, the S17 game is still better than the H17 one by ~0.10%, which is about half of what it is usually reported as (0.20%). At a TC of +5, which is when most will push their max bet out, the S17 game is still better than the H17 game by ~0.14%.

While the difference between the two games may seem small at high TC's, it must be remembered that many are willing to all sorts of crazy things to obtain a "theoretical" edge similar to this (i.e., level 2 or 3 systems and side-counting aces), although in practice the edge gained from such tactics may very well be lower than this.

Best,

SP

It appears that your sim was done using flat bets? Is that so?? Note that the disparity will narrow further in terms of overall earnings due to multiple units being out there when the difference is at its smallest.

 

[Southpaw]

It appears that your sim was done using flat bets? Is that so?? Note that the disparity will narrow further in terms of overall earnings due to multiple units being out there when the difference is at its smallest.

No, it actually won't. What you are betting at any given TC will not change the percent return at that TC.

Edit: I believe that I had mistaken what you were trying to say before making my initial response. To be frank, I'm still not quite sure what it is you're trying to suggest. What I'm saying is that even if you were ONLY betting at TC's between +5 and +10 in a H17 game (that would be nice, ehh?), your return would still be "a bit" lower than it would be in the S17 game; IOW, S17 does not equal H17 at these TC's. So, by using a more liberal ramp you can make the H17 rule less damaging, but you cannot make it any less damaging than say .10% or, assuming that we are only examining TC's that are sufficiently high.

 

[creeping panther]

Sp

No, it actually won't. What you are betting at any given TC will not change the percent return at that TC.

Now that makes alot of sense......this could get very interesting.

CP

 

[Southpaw]

At TC=9, the H17 game has a 0.1% edge. What this is telling me is that even with 1 billion rounds, your sample size is too small to produce valid results at these TCs.

What would also be interesting is running it for a DD game, where TCs of 9 or 10 are not that uncommon.

Yes, the sampling size is too small. In fact, if you look at the Standard Error (S.E.) at TC of +10, then you'll find that the S.E. is about as large as the difference between the S17 and H17. However, the S.E. gets smaller as the TC examined gets smaller, and to be frank I thought that what happened at TC = +9 was a fluke (but that I'd report it as is anyways), especially since S17 was consistently better at all other TC's.

SP

 

[Southpaw]

I am going to rerun this with 10 Billion rounds per trial.

SP

 

[zengrifter]

I am going to rerun this with 10 Billion rounds per trial.

Spaw - I have an interesting concept for a set of sims, for you...

If a decent counter can play (or see) 100 hands per hour, 30 hours per week, 50 weeks per year, for 20 years... equals 3 million hands in a lifetime of play, yes?

So use a strong strategy like AO2 w/ASC against a decent game like H17 65% pene DAS 1-2x4 spread.

Run sims of 3M each - each is a lifetime - run 100 separate lifetimes of 3M and let us see the distribution of results among them.

Start a new thread for this assignment. Be interesting to see what a best and worst lifetime of counting toil can produce. Perhaps you can graph it as well. zg

 

[Southpaw]

Spaw - I have an interesting concept for a set of sims, for you...

If a decent counter can play (or see) 100 hands per hour, 30 hours per week, 50 weeks per year, for 20 years... equals 3 million hands in a lifetime of play, yes?

So use a strong strategy like AO2 w/ASC against a decent game like H17 65% pene DAS 1-2x4 spread.

Run sims of 3M each - each is a lifetime - run 100 separate lifetimes of 3M and let us see the distribution of results among them.

Start a new thread for this assignment. Be interesting to see what a best and worst lifetime of counting toil can produce. Perhaps you can graph it as well. zg

I like the idea. I will start on it in the morning. I am assuming that you want this for a DD game, yes?

Edit: I have a simulation running right now that may not be done for another half-hour. I'm not sure CVData will let me run a simulation for that few hands. If you look at what is in my signature, this is directly quoted from the program. But, when the current sim is finished, I will see if it will let me do this.

SP

 

[Blue Efficacy]

I would like to know how much difference there is for the counter in a H17, RSA game vs a S17 nRSA game. In a shoe game especially maybe it could almost make the two even depending on the spread!

 

[Southpaw]

I would like to know how much difference there is for the counter in a H17, RSA game vs a S17 nRSA game. In a shoe game especially maybe it could almost make the two even depending on the spread!

Yeah, hmm. Interesting thought. Splitting Aces gets better at high TC's, while H17 matters less and this is where your big bets would be. If you specified parameters (you could use my original post as a guide), and a spread, (optimal spread capping out at 12, 15 or 20 units and Wong at TC = -2 sound all right?) I'd be willing to check it out.

SP

 

[psyduck]

The frequency of TC10 (hilo) is too low (0.022% in a typical 6deck shoe) to be practically useful. You will need a huge number of rounds to get statistically meaningful data for this high count.

 

[Southpaw]

I have redone the simulation, this time with 10 billion rounds. The S17 provides a better return at every TC by at least .07% and now the Standard error is at its max of .03% at TC = 10, but is much lower for the other TC's.

Here are the new results:

https://docs.google.com/viewer?a=v&pid=explorer&chrome=true&srcid=0B0cCldUn36hMYWRmNDFlMmItNmJlZC00ZDA1LWJkZGYtOTk3ZjQ4OTZhNjg3&hl=fr

I will edit these into the original post.

SP

 

[Southpaw]

Zg,

CVData will allow me to run sims of only 3M rounds. I will start on this in the morning.

SP

 

[zengrifter]

Zg,

CVData will allow me to run sims of only 3M rounds. I will start on this in the morning.

SP

This will show a range of likely lifetime outcomes, I think. zg

 

[Southpaw]

Yes, I'm curious to see how many (if any) poor souls lose money or make little to nothing after 3M rounds, even when playing a great game with a powerful system.

SP

This will show a range of likely lifetime outcomes, I think. zg

 

[neversplit5s]

On the subject of how various rule variations affect BS players vs. card counters, I got to thinking about some others. Obviously any that add options benefits everyone more or less (and the opposite for taking away options). Here's what I think I found (correct me if I'm wrong):

Affects counters more:

Reduced BJ payout - since one of the main reasons counters keep track of 10s and Aces is to know when a BJ is more likely, shortchanging the payout hurts everyone obviously but hurts counters even more.

Surrender - since the EV when the dealer is showing a 10 (and to some extent an 8 or 9) goes down as the count goes up, surrender helps counters more for this reason (since it's EV is fixed at -0.5). Early surrender would be a counter's dream since the chance of dealer BJ goes up with the count.

ENHC - for the same reason mentioned in the last sentence above.

Restricted double down - since doubling down generally becomes more profitable as the count goes up (due to better chances of catching a 10 or Ace), allowing DD on 9-11 or 10-11 only hurts counters more.

Double After Split - same reason as when double down is restricted to certain totals.

Restricted resplitting - since being able to split multiple times generally makes splitting more profitable, this hurts more at high counts when the dealer is weak. Being able to resplit Aces is especially helpful, since both Aces and 10-point cards (making 21s after splitting Aces) are more prevalent at high counts.

Affects BS players more:

Dealer hit/stand on soft 17 - as described above.

Drawing to split Aces - since being able to take additional cards after splitting Aces is more helpful when low ones are plentiful, this benefits BS players more (on the other hand, since pairs of Aces come up more at high counts this might neutralize the effect). However, being able to double down after splitting Aces would benefit counters more for the reasons mentioned earlier for doubling.

 

[jack.jackson]

On the subject of how various rule variations affect BS players vs. card counters, I got to thinking about some others. Obviously any that add options benefits everyone more or less (and the opposite for taking away options). Here's what I think I found (correct me if I'm wrong):

Affects counters more:

Reduced BJ payout - since one of the main reasons counters keep track of 10s and Aces is to know when a BJ is more likely, shortchanging the payout hurts everyone obviously but hurts counters even more.

Surrender - since the EV when the dealer is showing a 10 (and to some extent an 8 or 9) goes down as the count goes up, surrender helps counters more for this reason (since it's EV is fixed at -0.5). Early surrender would be a counter's dream since the chance of dealer BJ goes up with the count.

ENHC - for the same reason mentioned in the last sentence above.

Restricted double down - since doubling down generally becomes more profitable as the count goes up (due to better chances of catching a 10 or Ace), allowing DD on 9-11 or 10-11 only hurts counters more.

Double After Split - same reason as when double down is restricted to certain totals.

Restricted resplitting - since being able to split multiple times generally makes splitting more profitable, this hurts more at high counts when the dealer is weak. Being able to resplit Aces is especially helpful, since both Aces and 10-point cards (making 21s after splitting Aces) are more prevalent at high counts.

Affects BS players more:

Dealer hit/stand on soft 17 - as described above.

Drawing to split Aces - since being able to take additional cards after splitting Aces is more helpful when low ones are plentiful, this benefits BS players more (on the other hand, since pairs of Aces come up more at high counts this might neutralize the effect). However, being able to double down after splitting Aces would benefit counters more for the reasons mentioned earlier for doubling.

And just think what a players BJ beats dealers BJ and after splitting tens or aces and get A,X is considered a BJ would do for the counter. Those are some rules to look for. Those would greatly be enhanced from counting. I guess it is possible to have a higher WR verses another game with a lower HE depending on the rules.