1357111317
Well-Known Member
I was just wondering if anyone downloaded those BJmath EV charts for each hand before it went down. If anyone has a copy of those would they like to send me a copy of them?
BJMath Expected value charts
- Thread starter 1357111317
- Start date
1357111317
Well-Known Member
They are probably like the ones in blackjack attack. The BJmath ones show every 3 card combination and and expected value of each possible play, Hit, Stand, Double, split and I am not sure about surrender although it wouldnt matter since surrender is always -.5.
They are like Don's tables, except one decimal short lol.
I think there are some small differences in assumptions regarding splits between the 2.
I think there are different assumptions regarding frequencies between the 2 which you'd have to understand to use these tables for ES or LS.
For the most part, they are identical.
I understand Don's tables better lol. It helped alot that he actually explains what he assumed lol.
But here it is, for you or anyone.
OK - I just figured out it's an 800K sheet and exceeds the limits here.
Anyway, is it even OK I publish it anyway. All that copyright stuff? Or not?
If it's OK to publish it, tomorrow I'll zip it or break it up or something to make it work. It has a BS tab too which I can delete to make it fit.
Worst case, use Don's tables. Basically same thing.
1357111317
Well-Known Member
Now i doubt that these would exist but are there EV charts for every hand at different counts and not just off the top?
Sonny
Well-Known Member
The charts are for off the top EVs. If you want the EV for every hand vs. every dealer upcard at every count, ask Qfit.
-Sonny-
I haven't seen the type of charts your looking but in CVData you can bring up the report that indicates win rate per hand type. The results would at the very least provide you with some expectation value per hand type. It does not provide splitting/doubling/standing independant results that I can find anywhere.
BJC
Here's why that's difficult...
The underlying assumption here is that the count is a rock-solid piece of information. It's not. Here are two images that both show a HiLo TC of +2. The difference between the two overall EVs is a full 1.3%.
Attachments
QFIT
Well-Known Member
Which is exactly why I have often said that representative subsets are not useful. And why the CVData index generator provides the data assuming all possible subsets according to their frequencies.
Here they are. You can combine them into one sheet if you want.
Sorry I had to break it up into thirds due to file size limits here.
It came with a Basic Strategy for Late Surrender so that's the 4th one. I have it in a separate tab.
I do stuff in Lotus and when I convert it to Excel, if I'm lucky enough it can actually translate the formulae in the first place, it takes up twice as much space for some reason.
So I hope it makes sense to you.
What is the probability of drawing each rank at a given HiLo running count/pen combination? In your first image specifically 24 sixes, 4 fives, and 18 tens were removed yielding a +10 running count with 266 cards remaining. But what would be the probabilities if we just set the condition to be a HiLo running count of +10 with 266 cards remaining and nothing specifically removed? In that case there would be many different possible HiLo subsets, each with a different probability of occurring. When each subset is weighted according to its probability, the probability of drawing each rank is a floating point number as shown in the attached image. So the theoretical HiLo comp for a +10/266 card slug dealt from 6 decks is:
Code:
2-6 266 * .07309 = 19.44 each, 97.21 total
7-9 266 * .07716 = 20.52 each, 61.57 total
10 266 * .32245 = 85.77 each, 85.77 total
A 266 * .08061 = 21.44 each, 21.44 totalObviously these are not integral type numbers so they need to be rounded before being input into a CA but they should serve as a good basis for analysis. I rounded the comp to (2 through ace) 19-19-20-20-19-21-21-20-86-21 and get an EV of +.2749 using basic strategy for 6 decks, S17, NDAS, no resplit, no hit split aces. The comp could be rounded slightly differently and probably some sort of rounding average algorithm should be used, but +.2749 should be in the ballpark.
The point is a CA can compute probabilities for both likely and unlikely compositions but it's an equal opportunity employer for both likely and unlikely scenarios.
