Why? REKO/CoBJ Bet ramps effect?

[Red Green]

The answer may be "It's the sim, stupid!", but I was wondering:

REKO-F is marginally better (SCORE) than CoBJ. If CoBJ gets its 'stuff' from the true-count-ish bet ramp and REKO-F uses the largely unadulterated KO RC model, where does the REKO advantage come from? Either its the indices (doesn't seem likely) or is it the super fast bet ramp: twice as fast as KO and three times as fast as COBJ. If the latter, is it really accurate to compare the two or does it make more sense to match the ramps and and see how they perform?

Background is that I was working on CoBJ when I decided to look at REKO. It is much easier to implement the REKO betting than CoBJ and that's a big factor for me. I try to remind myself to keep it simple. On the other hand if I were to sim REKO with sped up CoBJ multistrategy, who knows (a project for later in the week, perhaps)? A simplified ramp and squeeze out a few extra points?

[sbrugby]

Not sure if you're referencing tables somewhere or sims? If I understand what you're asking, a lot of this is covered in Modern Blackjack Vol I 2nd edition (Location 9057 58%). The kindle edition is only $1 on amazon, extremely worth it if you're planning to use any KO based count (or felt, or any of Norm's software). You can have the optimized depth-based betting with any of these KO based counts, but you'd probably probably need CVCX/CVDATA to optimize them for the exact decks/rules/pen/spread of your game. To me that's a lot of work to optimize for 1 specific set of conditions. .

[Red Green]

Thanks - the questions arose in comparing canned sims on cvcx. I then reviewed Modern Blackjack on REKO, REKO F and the relevant appendices.

I was comparing the KO Full, REKO-F and COBJ in an H17 DAS, 10K BR, 1-10 spread with $10 min units.

Perhaps I asked the question badly. Maybe the real question is: does REKO perform better than KO due to accelerated bet ramps?

Sent from my iPhone using Tapatalk

[sbrugby]

Maybe the real question is: does REKO perform better than KO due to accelerated bet ramps?

Don't know much about KO but looking at KO Full 6 deck against REKO-F 6deck it seems like the only difference is the indices. REKO-F seems to have more and I think they are Risk Averse whereas KO's are probably EV maximizing? So I'm guessing combination of more indices and the fact they are RA allows them to outperform KO. It's also possible the indices for REKO-F were generated for the depth you're looking at, whereas KO's may have been generated for some other depth.

The Risk Averse Indices would allow you to bet more than EV maximizing ones, hence the faster Ramp for REKO-F(and better SCORE) over KO-Full.

[Red Green]

Thanks sbrugby


In the KO book and CoBJ the bet ramps are as wide as noted in OP. In the sims they are all about 6 points between min and max for all three systems. An odd discrepancy. I guess there is no way to know how the books derived their ramp spread.


I'm not 100% familiar with CVCX yet, but is it possible that the CVCX count column is TC and not RC when an unbalanced system is presented? '"Count Freq." is described as "true count frequency" in the manual...Not intuitively plausible to me that it would be TC for a RC system, but I don't understand how one source can ramp 1-10 in 5 moves and another source makes the same ramp in 15 moves.

[sbrugby]

I'm not 100% familiar with CVCX yet, but is it possible that the CVCX count column is TC and not RC when an unbalanced system is presented? '"Count Freq." is described as "true count frequency" in the manual...Not intuitively plausible to me that it would be TC for a RC system, but I don't understand how one source can ramp 1-10 in 5 moves and another source makes the same ramp in 15 moves.

It would only be TC if you've changed the strategy to something other than "No True Count" under define/edit strategy for the strategy being simmed (Or created a new strategy and done the same).

Maybe you could post pictures of your sims because all the sims I've made with your parameters (standard RC based, no depth-bettting) seem to have very similar betting ramps between the KO variations( excluding COBJ). They all start increasing @ ~-3 and top out @ ~+2; +/- 1.

[Stealth]

Have not looked at these for some time but would suggest you also focus on the variance levels you get with each. EV is not always the answer.

[Red Green]

Thank you both.

Sbrugby - those ramps agree with what I have.

Stealth - I will look at those variance/SDs more carefully tomorrow. thanks for pointing in that direction.

I must say I'm still a bit puzzled how the books can publish such wide ranges while the sims make the same ramp is 5 or 6 moves. If the betting ramp is Kelly derived, maybe that is the answer? The books maybe use a 1/2 K and Norm used full? I'm not a math guru so no way for me to scribble that solution out on a napkin!

[sbrugby]

U can set the Kelly factor in CVCX but for your bankroll of 10K and betting unit of $10 changing it won't change the ramp values(except @ 2.0) . Even then, the optimal units bet at each RC won't change, only your base unit will.

The K-O blackjack book appears to have come out in 1998, I think score wasn't a thing until 1999, which is what CVCX is optimizing for.

[Red Green]

I wasn't able to see how to adjust Kelly in the sim but I did find how to manipulate the CoBJ style betting application (Norm - you have thought of everything!) via the depth betting button on sim page and tested it with REKO-F. As it turns out, you can add a point here or shave a point there, but its all for nought as the changes are not material. The truth has been said before and so now I've seen it with my own eyes:

* Find and track dealers who deal deeper because penetration has a huge impact on EV;
* Grow your BR to raise your bet range within your RoR so you can earn more per hour;
* Play games that are good based on your BR and RoR.

So this experiment is over.

Back to memorizing indices!