MJ: KO Question for Norm

[MJ]

If one were to use KO Full over KO-P with optimal betting incorporated, how much of a gain in SCORE could be achieved?
I recall in the book the authors advised against using the Full matrix.

When Knockout BJ was written, I'm not sure if optimal betting theory was as advanced as it is today. Did you ever run any simulations on KO-Full? If so, how did it fair against KO-P?

Lastly, when you generated your compromise RA index of +7 for the 10 splits, what ROR did you use? Also, what was the EXACT RA index for 10s vs 5 and 6? Thank you for any answers.

MJ

[Norm Wattenberger]

> If one were to use KO Full over KO-P with optimal
> betting incorporated, how much of a gain in SCORE
> could be achieved?
> I recall in the book the authors advised against using
> the Full matrix.

Depends heavily on decks and penetration. At six decks, I get about a 3% to 9% improvement:



> Lastly, when you generated your compromise RA index of
> +7 for the 10 splits, what ROR did you use? Also, what
> was the EXACT RA index for 10s vs 5 and 6? Thank you
> for any answers.

Full Kelly for playing and betting. Can't remember the exact indexes.


Serious Blackjack Software

[MJ]

> Depends heavily on decks and penetration. At six
> decks, I get about a 3% to 9% improvement:

Thanks Norm. The game I play is exactly for the rules given in the chart. Could you possibly chart Hi-Lo I-18 Fab4 with the 2 other curves?

[Norm Wattenberger]

> Thanks Norm. The game I play is exactly for the rules
> given in the chart. Could you possibly chart Hi-Lo
> I-18 Fab4 with the 2 other curves?

Can't really show this well with one chart. Spread becomes important when comparing 6D KO-Full with HiLo Ill18. At a low spread, HiLo is a bit better. At a high spread, KO wins. At common spreads they are very close except HiLo does better with extreme penetration (e.g. 5.5/6)


CVCX Online

[MJ]

> Can't really show this well with one chart. Spread
> becomes important when comparing 6D KO-Full with HiLo
> Ill18. At a low spread, HiLo is a bit better. At a
> high spread, KO wins. At common spreads they are very
> close except HiLo does better with extreme penetration
> (e.g. 5.5/6)

Do you know whether the index values the KO authors provide for the Full strategy are RA index values? If not, then could using ALL RA indexes for the KO-Full strategy plus the addition of the two 10 splits outperform Hi-Lo I-18Fab4?

With KO-Full vs KO-P, at 75% penetration how much gain is there in SCORE? I can tell from the chart it is not much, but that is a good penetration to use as a benchmark for performance. Even if the gain is only 3%, that is not minor.

Also, why did you write on another board that you were not sure which index values to use for the KO strategy on CVBJ? Everything is right there in the book (except the 10 splits).

BTW, it would be great if you added KO Full to the canned sims archive.

MJ

[Norm Wattenberger]

> Do you know whether the index values the KO authors
> provide for the Full strategy are RA index values? If
> not, then could using ALL RA indexes for the KO-Full
> strategy plus the addition of the two 10 splits
> outperform Hi-Lo I-18Fab4?

As for RA, I never looked that closely but I don't think so. In common situations I think KO with good indexes beats HiLo - particularly when you take into account deck estimation error. Bt it still depends on spread. Amazing when you think about it. Cacarulo has probably looked at this more than I.

> With KO-Full vs KO-P, at 75% penetration how much gain
> is there in SCORE? I can tell from the chart it is not
> much, but that is a good penetration to use as a
> benchmark for performance. Even if the gain is only
> 3%, that is not minor.

The chart only looks at one spread and set of circumstances. I use the charts with many penetrations to avoid the trap of looking solely at one isolated situation. But, there are so many variables.

> Also, why did you write on another board that you were
> not sure which index values to use for the KO strategy
> on CVBJ? Everything is right there in the book (except
> the 10 splits).

All books present a challenge when it comes to replicating their tables since they have tables and then extending paragraphs and then footnotes and then appendicies. The KO book has a "generic basic" and then a footnote on DAS mods and no mention of ENHC. My discussions with Olaf indicated a disinterest in NHC. But, we are still discussing.

> BTW, it would be great if you added KO Full to the
> canned sims archive.

I need to redo every sim for the next release as the file format will change again. But, that is not due until the end of the year.


Serious Blackjack Software

[Don Schlesinger]

> Do you know whether the index values the KO authors
> provide for the Full strategy are RA index values? If
> not, then could using ALL RA indexes for the KO-Full
> strategy plus the addition of the two 10 splits
> outperform Hi-Lo I-18Fab4?

Strange question. It might outperform, but then, if you used RA indices for Hi-Lo, too, to make a valid comparison, then Hi-Lo might outperform. Why use RA indices for one system and then compare that to another system that is deprived of those same indices?

In any event, RA indices (see BJA3) add very little to SCORE.

Don

[Cacarulo]

> Strange question. It might outperform, but then, if
> you used RA indices for Hi-Lo, too, to make a valid
> comparison, then Hi-Lo might outperform. Why use RA
> indices for one system and then compare that to
> another system that is deprived of those same indices?

Good question! I think it should happen the same as with EM-indices (good spreads: KO beats Hi-Lo) although I haven't run sims to confirm that for sure.

Sincerely,
Cac

[Cacarulo]

> Do you know whether the index values the KO authors
> provide for the Full strategy are RA index values? If
> not, then could using ALL RA indexes for the KO-Full
> strategy plus the addition of the two 10 splits
> outperform Hi-Lo I-18Fab4?

Look, if you simply use Catch-22 (EV-maximizing indices) for both systems KO does outperform Hi-Lo provided that a good spread is used.

Sincerely,
Cac

[MJ]

> Look, if you simply use Catch-22 (EV-maximizing
> indices) for both systems KO does outperform Hi-Lo
> provided that a good spread is used.

When you mention EV-maximizing indices are you referring to KO-Full or KO-P? Would the above comment be true for all penetration levels provided a good spread is used?

MJ

[Cacarulo]

> When you mention EV-maximizing indices are you
> referring to KO-Full or KO-P? Would the above comment
> be true for all penetration levels provided a good
> spread is used?

Neither KO-Full nor KO-P, I used the Catch-22 indices. I haven't checked all penetration levels but with a good spread and normal levels of pen it should remain true.

Sincerely,
Cac

[MJ]

> Neither KO-Full nor KO-P, I used the Catch-22 indices.
> I haven't checked all penetration levels but with a
> good spread and normal levels of pen it should remain
> true.

Two quick questions...

1) Is Catch-22 the I-18Fab4? If so, then did you just generate each index for yourself via simulation? Most of the I-18Fab4 indices are given in the KO Book anyhow so I guess you only had to figure out the unlisted ones.

2) When you generated the compromise RA indices for the 10 splits, why did you use 5/6 for the penetration level? 4.5/6 is far more common and realistic. Maybe that is why Norm got +7 and you got +8 (pretty close to each other).

MJ

[Cacarulo]

> 1) Is Catch-22 the I-18Fab4? If so, then did you just
> generate each index for yourself via simulation? Most
> of the I-18Fab4 indices are given in the KO Book
> anyhow so I guess you only had to figure out the
> unlisted ones.

No, the C22 are the I18 + 4 additional plays: 8v5, 8v6, A8v5 and A8v6. The indices were generated by simulation for the conditions I mentioned before: 6D,S17,DOA,DAS,SPA1,SPL3,NS,5/6.

> 2) When you generated the compromise RA indices for
> the 10 splits, why did you use 5/6 for the penetration
> level? 4.5/6 is far more common and realistic. Maybe
> that is why Norm got +7 and you got +8 (pretty close
> to each other).

Yes, that could explain the difference.

Sincerely,
Cac