Simple Level 2's vs. Difficult Level 2's

[Deathclutch]

I have been playing around with some sims today and have come up with some surprising (at least to me) results. I wanted to compare results between what are considered simple level 2's and what I've been using. I threw in Hi Lo just to get a comparison.

The rules for the sim for the 6 deck game were:

4.5/6 pen
H17, DAS, No Surrender
I used only the Ill 18 indices for every count.
The Hi Opt II counts DID NOT side count aces.
2 of the Hi Opt II Counts used indices I came up with, one is the standard Hi Opt II
They all had a 1-10 spread
$25 units.

What I found surprising was the fact that Ben Franklin count, even with it's simplified index chart, came out on top and FELT, with having an even simpler chart, came in a close third.

They both competed well with Zen, and outperformed Hi Lo (which was expected) and outperformed all variations of Hi Opt II which I use.

I then decided to see how they all fare in a double deck game where I expected the simpler counts to get crushed.

1.6/2
H17, DAS, No Surrender
I only used I18
Hi Opt Counts DID side count aces
They all used a 1-6 spread
$25 units

This time Hi Opt II canned barely beat out the Ben Franklin count. Hi Lo and Zen were comparable and the FELT count got left in the dust.

I guess my question is, do these sims look correct to you guys? The results aren't what I expected and I'm worried I may be running these wrong. The main thing that worries me on the last sim is that FELT and Ben Franklin are so close in how they're employed, yet the results are so far apart.

 

[snorky]

I've pondered the same questions and its been bothering me for a while as well. From what I'm told, the simulations underestimate the playing efficiency of the hi opt systems, hence the lower winrate.

I'm not too sure though, so I hope someone can answer this in details.

 

[tthree]

I've pondered the same questions and its been bothering me for a while as well. From what I'm told, the simulations underestimate the playing efficiency of the hi opt systems, hence the lower winrate.

I'm not too sure though, so I hope someone can answer this in details.

The lack of the ace side count hurts the HIOPT II results. Almost everyone who uses it side counts aces. In pitch games it does very well even without the ace side count but really crushes with it.

Ben Franklin count? I heard AM never posted anything of value. Must have been a bad source of information. Thanks AM.

 

[FLASH1296]

Ben Franklin Count:

(Dead link: http://dl.dropbox.com/u/597675/Ben-Franklin-Count.pdf)

*********************************************

 

[NAP]

Try dropping the DD pen from 1.6/2 to something like 1.25/2 or 1.33/2.

 

[jaygruden]

This time Hi Opt II canned barely beat out the Ben Franklin count. Hi Lo and Zen were comparable and the FELT count got left in the dust.

I guess my question is, do these sims look correct to you guys? The results aren't what I expected and I'm worried I may be running these wrong. The main thing that worries me on the last sim is that FELT and Ben Franklin are so close in how they're employed, yet the results are so far apart.

I've read in several books that the flaw many people make when comparing different counts is failing to use the optimal betting ramp for each count. For example, If you are using the optimal betting ramp for xxx count as your standard for all counts, then the xxx count will appear to outperform all other counts. If you run the sim again but this time using the optimal betting ramp the the yyyy count then the yyyy count will appear to out-perform the others.

I think this was proven to be the problem when the first KO book came out. He was using the optimal betting ramp for KO when comparing it to other counts and in those sims, it outperformed the others. It has since been suggested that when comparing counts you need to use the optimal spread/ramp for each count to truly tell what the differences are.

Also penetration affects the score. Using the 1.5/6 pen shows only the scores at that pen but some counts will outperform others at better pen while others will outperform at worse pen.

The other question I have is......Are you "playing all" or "wonging" in this sim? Maybe I missed it, if so I apologize, but I didn't see that stated. I ask because at least in the BFC, AM suggests a 1-20 spread if playing all, but to spread 1-8 if wonging out at TC -2 or worse.

 

[FLASH1296]

Yes, using sub-optimal bet ramps will badly skew the results; but what leaps
off the page for me, is the use of the King of Counts, Hi-Opt II, in a crippled fashion.

All Ace Neutral Counts MUST Side-Count Aces. ... Doh' !

Also, Level Two Counts with their powerful Playing Efficiency are unfairly
compared by limiting the breadth of the indices employed in the sim'

:whip:

 

[kewljason]

Also, Level Two Counts with their powerful Playing Efficiency are unfairly
compared by limiting the breadth of the indices employed in the sim'

Disagree! For a fair comparison the same number of indicies must be used for each count. Ill 18 seems like a reasonable starting point to use. But that brings me to the question, what is the "complete hi-lo"? Does that involve more than the Ill 18? If so, the hi-lo numbers are tainted upwards.

However, that being said, the numbers look reasonable to me. With the exception of the disparity between felt and Ben Franklin which as noted are similar, both using RPC tags. Something seems out of whack there.

I have always argued that level 2 improvement over hi-lo is in the 5-10% range, closer to the lower end of that range. Usually during these discussions, proponents of the level 2 counts will cherry pick condition which will reflect the higher end of that range. 6 decks, 75% pen seems pretty standard and fair for comparison sake.

 

[QFIT]

I've pondered the same questions and its been bothering me for a while as well. From what I'm told, the simulations underestimate the playing efficiency of the hi opt systems, hence the lower winrate.

I'm not too sure though, so I hope someone can answer this in details.

No, sims are more accurate. PE is an estimate.

 

[Deathclutch]

I've read in several books that the flaw many people make when comparing different counts is failing to use the optimal betting ramp for each count. For example, If you are using the optimal betting ramp for xxx count as your standard for all counts, then the xxx count will appear to outperform all other counts. If you run the sim again but this time using the optimal betting ramp the the yyyy count then the yyyy count will appear to out-perform the others.

I think this was proven to be the problem when the first KO book came out. He was using the optimal betting ramp for KO when comparing it to other counts and in those sims, it outperformed the others. It has since been suggested that when comparing counts you need to use the optimal spread/ramp for each count to truly tell what the differences are.

Also penetration affects the score. Using the 1.5/6 pen shows only the scores at that pen but some counts will outperform others at better pen while others will outperform at worse pen.

The other question I have is......Are you "playing all" or "wonging" in this sim? Maybe I missed it, if so I apologize, but I didn't see that stated. I ask because at least in the BFC, AM suggests a 1-20 spread if playing all, but to spread 1-8 if wonging out at TC -2 or worse.

I did not set up an optimal ramp for each count.

Every level 2 is using a 1-10 spread through +1 through +6.
The level 1 is using a 1-10 spread through +1 to +3.

I did not set a Wong out point.

Even with an optimal bet ramp wouldn't the BFC and FELT have a nearly identical ramp since they use the same tags?

 

[Deathclutch]

Try dropping the DD pen from 1.6/2 to something like 1.25/2 or 1.33/2.

I will try that one when I get home. I was simming for games I play.

 

[Deathclutch]

Yes, using sub-optimal bet ramps will badly skew the results; but what leaps
off the page for me, is the use of the King of Counts, Hi-Opt II, in a crippled fashion.

All Ace Neutral Counts MUST Side-Count Aces. ... Doh' !

Also, Level Two Counts with their powerful Playing Efficiency are unfairly
compared by limiting the breadth of the indices employed in the sim'

:whip:

Sorry Flash, I was simming how I play. I can side count in pitch games, but feel my accuracy drops in 6 deck. I had always heard Hi Opt II outperforms HiLo even without the side count so I felt okay using it this way.

As for the I'll 18, I wanted to make all counts comparable to the simpler counts to see if switching to something so easy would hurt my win rate and allow me to concentrate on other things.

 

[Deathclutch]

Disagree! For a fair comparison the same number of indicies must be used for each count. Ill 18 seems like a reasonable starting point to use. But that brings me to the question, what is the "complete hi-lo"? Does that involve more than the Ill 18? If so, the hi-lo numbers are tainted upwards.

However, that being said, the numbers look reasonable to me. With the exception of the disparity between felt and Ben Franklin which as noted are similar, both using RPC tags. Something seems out of whack there.

I have always argued that level 2 improvement over hi-lo is in the 5-10% range, closer to the lower end of that range. Usually during these discussions, proponents of the level 2 counts will cherry pick condition which will reflect the higher end of that range. 6 decks, 75% pen seems pretty standard and fair for comparison sake.

Even though it says hi lo complete I set it to only use Ill18 so the results should be comparable.

 

[QFIT]

I did not set up an optimal ramp for each count.

Every level 2 is using a 1-10 spread through +1 through +6.
The level 1 is using a 1-10 spread through +1 to +3.

I did not set a Wong out point.

Even with an optimal bet ramp wouldn't the BFC and FELT have a nearly identical ramp since they use the same tags?

I'm afraid 300 million rounds are nowhere near enough when comparing simalar strategies. Five billion will give much better results. I always run the sims separately, instead of at the same table. Much faster, and players can affect each other. If you have the latest version, CVData has a Call CVCX button that will give you the optimal bet results.

 

[Deathclutch]

I'm afraid 300 million rounds are nowhere near enough when comparing simalar strategies. Five billion will give much better results. I always run the sims separately, instead of at the same table. Much faster, and players can affect each other. If you have the latest version, CVData has a Call CVCX button that will give you the optimal bet results.

Thanks for the tip. I'll try again with more rounds and a custom bet ramp for each one.