FA Dropouts
The Floating Advantage theory began with Ken Uston and Al Francesco alleging that basic strategy expectation changes with penetration; that your edge in a shoe is the same at any penetration as if you originally started with the same number of cards that you have left in that shoe. The theory was not too well received.
Gwynn and Seri started the rise of the FA theory with the data they produced that resulted in the Blackjack Forum article, How True is Your True Count?. They found extreme fluctuations in the value of each true count in their simulations. Later they appeared to find a variation in basic strategy simulations that seemed to match part of Uston�s version of the theory.
As reported in Blackjack Attack, Schlesinger developed an initial SAFE play system to match one�s betting spread to this trend. Then 4 of the more famous simulation gurus convinced him that this was a rather complex system for limited gains. SAFE was thus glossed over in BJA.
All hell broke loose when a poster on rge21.com said that he found a 5.5/6 shoe game with very good rules that might be the type of game to go ahead and try the SAFE system. When I stated there was no floating advantage, just a bow-effect and some rounding errors in most simulations for how true counts are grouped together, I was PBed by Parker, for what I believe was more failure to worship and to have questioned Don�s views than any other reason. I said that the combination of rounding errors, a real bow effect, and several other minor effects was all combined into the �Sausage Math of the Floating Advantage.�
But the other dropouts from the FA theory are part of a list that has to start with Gwynn and Seri themselves. It is a bit of an adventure to try to reconstruct their later views scattered through the incomplete mentions in BJA, the original Blackjack Forum articles BJA is based on, and the later papers of Gwynn and Seri. If it makes everyone feel better you can just consider this my opinion of their scattered statements:
When simulating the presumed effects of basic strategy and the floating advantage one is presented with 3 seemingly equally valid assumptions that produce 3 very different results:
(1) Taking a shuffled shoe and burning cards until a given penetration level is reached produces no floating advantage. It is the same as simply cutting the pack at the sample point.
(2) Doing a series of simulations and taking samples where a round begins at a given penetration point appears to create an anti cut card effect, where only beginning a round at a fixed point appears to act in the opposite manner than not beginning a round if started past a fixed cutcard point.
(3) Taking samples for a given penetration by seeking a number or rounds that on average end at the target sample point appears to similarly resemble the cutcard effect, in that this is then the inverse of not starting a round until after a given number of rounds.
All 3 assumptions appear equally valid and produce 3 varied results, and appear to prevent most simulations from resolving floating advantage questions. They also did mention a caution on true count rounding where, in deeper plays, one could group higher counts with counts that on average were lower but had the same rounding ranges.
Peter Griffin can be considered a dropout in that he discussed the original, �How True�� results and alleged floating advantage by ONLY mentioning the bow-effect. It is truly strange that despite showing no support for any sort of FA he is used as a supporting source.
I have posted proofs of my views by:
(1) A pure Bayesian approach based on the first assumption above for basic strategy simulations, and the true count rounding cautions that actually go back all the way to Alan Wilson�s work, and
(2) A modified stepwise argument to refute ML�s trying to substitute an appeal to some absolute mean effect for mean deck compositions, that appears IMHO to just be an attempt to substitute something seemingly reasonable with similar alleged results to the true count rounding error.
(3) In both of these mean expectation of any given pack is also the mean expectation of all possible packs that can be derived from that given larger pack of cards. There is even one, simplest of all proof, that is based on the entropy of a pack and measuring that pack changes imperfectly by a true count---well obvious to Chem nerd, Quark, Steephan and myself maybe. Very simply the entropy of the entire shoe is preserved in measuring samples without replacement with an imperfect true count. The increase in entropy with penetration being conserved, the entropy overall causes accumulated deviations from the precise mean packs for any given true count. In exactly the same way a simple cyclonic heat exchanger is modeled, that entropy adds to the entropy of the true count where it already exists in greatest degree, the more extreme counts, and adds player expectation to the middle true count ranges that instead exhibit a bent linearity.
There is simply no floating edge, and people claiming there is have been very aggressive in covering this up and quite disingenuous. There is however a bow-effect and it is best exploited by adding more indexes and seeking more playing index gains. Folks just have to live with the way �blackjack attacking� politics has kept many topics buried. It is just the way it is! Sorry�And sometimes the guy doing the drain work, when things have become stagnant and stinky becomes the one blamed for the smell!
