I have done rather a large number of sims combining different tweaks. I see no reason to believe that.
I get this but don't understand why you would think the percentages should be added. I have also done a lot of sims combining tweaks and while sometimes the interact a little the multiple is always much closer than the added to the actual result. And the reason it is different is that the gain is no longer the same as what it was without the interaction. In other words for our example when one tweak was added the gain wasn't actually 15% anymore because of the interaction as I stated.
It is silly batting this back and forth unless you are denying that gains compound on the new whole so the sum of gains is more than the sum of their parts. You worked in finance and should know this. You can go to the thread were I showed this concept for actual sim results:
https://www.blackjacktheforum.com/sh...eling-an-onion
The sum of the parts was for a 1-12 spread was 135.36% but the sims showed actual gains for the combined tweaks to be 205.01%. While the multiple of the parts is a 203.67% gain, which shows either rounding error for combining the 4 tweaks or a small interaction. Multiplication is so much more accurate a predictor being less than 1% off the sim results compared to adding where the actual sim results are 51.46% higher than the added prediction.
I didn't say they should be added. Let me try again, they shouldn't be added, multiplied, subtracted or divided. In many cases, if not most, they would be less than added. Blackjack is all about interrelationships. Everything impacts everything. You can't use simple algebra to combine the effects of multiple strategy changes.
"I don't think outside the box; I think of what I can do with the box." - Henri Matisse
You need to make shorter posts. I can't read all of that. The point is that you must sim results, since blackjack card counting is an intractable math problem. You cannot take a bunch of different variations and try to combine their results using algebra or simple arithmetic. There simply is no proof that this would work, and 20 years of sims tell me that it doesn't work. There is no mathematically shown reason this should work. This is not curve-fitting. This is not linear algebra. This is not regression theory. There is simply no reason to assume that those concepts fit to blackjack card counting.
"I don't think outside the box; I think of what I can do with the box." - Henri Matisse
Incidentally, there is a study that will likely be published in the near future based on the concept that blackjack card counting can be improved using regression theory. Sounds very weak to me.
"I don't think outside the box; I think of what I can do with the box." - Henri Matisse
Can you reveal whose notion this is ? It sounds rather fanciful to me.
How does one define Explanatory and Dependent variables for BJ.
In the 1970's I used regression analysis to develop a highly successful
"system" for "beating" the races. In 1977 and 1978 I won a considerable
amount with my system of flat-bets. Win % about 22% Mean winning
odds were about 4.6 to 1. Wagers placed on approx. 33% of the races.
Horse racing has a nice assortment dependent variables. Does BJ ?
Thanks Norm. I have been saying the same thing about sims and use the sim surprises when compared to the math to try to identify areas where nonlinear approaches may gain some on the linear approaches. I am well aware of what you are saying. In my humble opinion the multiple approach to several tweaks is the best approximation for things that are worth doing. I find surprises from time to time when I sim to confirm math but more often things are within the range of expectation. People should sim to see how things are for themselves but in general improvements compound each others gains. To what degree a sim will tell but I haven't been surprised very often. Most of the time sims outperform but often I am looking at things I nobody has discussed to my knowledge after looking into them so expectation was an educated guess. The interesting thing about blazing new trails is you really don't know what you will find. You have what you expect to find but since nobody has talked or written about it before you really don't know what you will find until you run the sims. I have some sims cooking now that I have an expectation of what my plan will be with the results but if the results are surprising the plan must be amended or scrapped.
moses,
I'm curious. Why didn't you stick with hit? Horse handicapping offers the horses last 10 races at best and some of those are considered irrelevant. Blackjack requires a 200m sim and anything less is considered meaningless. Can you imagine going back to review results of the NFL games in the 1930's to look for trends to this weekends games?
In the 70's there were no personal computers; just simple calculators.
Daily handicapping of one (1) track took me several hours with a pencil.
There was no O.T.B. (yet) or Race Books where I was residing back then.
The commute was boring and the hours that I put in ~ led to my burnout.
This predates, by several years, my learning to play a silly card game that
has something to do with 21, Aces & Faces, etc. etc.
Trends ? Do not make me laugh. Bunkum. Regression Analysis permitted
me to correlate dependent and independent variables - with the CRUCIAL
"weighing" of the various factors ... to create, what was, in effect, a power
rating. THEN ... those ratings were matched up with the ODDS (dependent
upon the SIZE of the field). For example, my bet had to be one of the top 4
rated nags AND had to have odds between 3-1 and 6-1 in a field of 9 to 12.
In an 8 horse field I would accept betting at odds between 5-2 and 5-1.
"offers the horses last 10 races at best"
L.O.L. Nowadays, (on-line) it is easy to get life-long results in detail.
Last edited by ZenMaster_Flash; 01-14-2016 at 09:25 AM.
Sorry T-three..........explain your math in more detail........increasing the benefit by 15% each time, isn't cumulative (45%) but wouldn't be the way you multiplied it, would it? Aren't you in your above example adding the gains? When you increase it by 15% by 15% by 15%, is this not instead an additional 15% OF 15% of the ORIGINAL 15% or just an additional overall 0.003375 gain added? Therefore 15.3375%? Rather than a gain of 52.0875%?
"Women and cats will do as they please, and Men and dogs should just relax and get used to the idea" --- Robert A. Heinlein
Each tweak gains 15% on the whole, not the original whole but the new whole after all previous tweaks. So:
Original: 1
First 15% tweak: 1*1.15 = 1.15 as the new whole.
Second 15% tweak: 1.15*1.15 = 1.3225 as the new whole.
Third 15% tweak: 1.3225*1.15 = 1.520875 as the final whole.
Subtract the original whole and divide by the original whole to get the gain as a decimal for all 3 tweaks:
(1.520875 -1)/1 = .520875 or 52.0875%
Does that help clear it up?
Bookmarks