strategy trainer game on main page

[Kasi]

ok i had 100 hands.....
i only got 2 blackjacks
dealer got 10 blackjacks :flame:
oh and i lost a thousand bucks and had to dip in and get another so i'm down $949.00 :cry:

edit: ok i now played a total of 392 hands dealer bj's = 24 player bj's 15

Hey Wise One and Canceller - if you're going to play the Trainer here, while you're keeping track of BJ's, set up the game for the weekend warrior game, rules, pen etc and go ahead and bet that way keeping the count etc. I know, you won't have your tools tracking every darn thing, you'll never know any mistakes, etc but you never know that stuff anyway lol. In real life and all. But, hey, you'll still know hands played and money lost or won every shoe.

Keep results separate if you want. If you lose, obviously there's something wrong with the game here lmao.

Or not lol.

 

[sagefr0g]

Hey Wise One and Canceller - if you're going to play the Trainer here, while you're keeping track of BJ's, set up the game for the weekend warrior game, rules, pen etc and go ahead and bet that way keeping the count etc. I know, you won't have your tools tracking every darn thing, you'll never know any mistakes, etc but you never know that stuff anyway lol. In real life and all. But, hey, you'll still know hands played and money lost or won every shoe.

Keep results separate if you want. If you lose, obviously there's something wrong with the game here lmao.

Or not lol.

yeah the Trainer must be out of wack cause i lost my a$$. got mad and started steaming lol. just kidding :joker:
so but i quit keeping track of money lost or won and just was keeping track of bj's. one problem with the trainer is betting at the level we are you only start with $1000, so you end up maybe losing most of it before a shoe is over then you hit a high tc and don't have any ammo to fire the proper bet size or double down what ever the case may be. can't get anymore money till you lose the full $1000. so that being the case i can't figure how to play to the full extent of our $100,000 bankroll.

 

[Canceler]

You beat me to it!

one problem with the trainer is betting at the level we are you only start with $1000, so you end up maybe losing most of it before a shoe is over then you hit a high tc and don't have any ammo to fire the proper bet size or double down what ever the case may be. can't get anymore money till you lose the full $1000. so that being the case i can't figure how to play to the full extent of our $100,000 bankroll.

That's exactly what I was going to say. Even dividing everything by 5 would still leave you with a max bet of $120, which could cause the $1000 to disappear pretty quickly.

 

[Kasi]

That's exactly what I was going to say. Even dividing everything by 5 would still leave you with a max bet of $120, which could cause the $1000 to disappear pretty quickly.

Well I think you could just force a re-shuffle after 249 cards or 250 cards to get the pen level? I don't think the hands-played changes if you go broke with the $1000 starting roll anyway? I thought it just asked if you want more money and gave it to you? If so, then just keep track of amount bought for that shoe, maybe?

Or multiply by 5 with a $5 unit if you want. You're only going to get 40 or so hands per shoe anyway so losing 200 units in one shoe would be pretty rare anyway. Then start each shoe with another $1000. teams aren't gping to send you out with a huge amount of units anyway. Well, I think, like I have an actual clue lol.

Or not, if you feel it's too much of a pain lol.

 

[sagefr0g]

Well I think you could just force a re-shuffle after 249 cards or 250 cards to get the pen level? I don't think the hands-played changes if you go broke with the $1000 starting roll anyway? I thought it just asked if you want more money and gave it to you? If so, then just keep track of amount bought for that shoe, maybe?

Or multiply by 5 with a $5 unit if you want. You're only going to get 40 or so hands per shoe anyway so losing 200 units in one shoe would be pretty rare anyway. Then start each shoe with another $1000. teams aren't gping to send you out with a huge amount of units anyway. Well, I think, like I have an actual clue lol.

Or not, if you feel it's too much of a pain lol.

thing is i think you gotta lose the full $1000 before you can access any more money. that could happen any time, just when you get a good count maybe you got some money but not enough to bet properly or double or split. so you find your self losing out on good counts.
a few times when i got low on money i tryed to perpously lose it so i could get some more money and ended up winning more money lol.

 

[Kasi]

thing is i think you gotta lose the full $1000 before you can access any more money. that could happen any time, just when you get a good count maybe you got some money but not enough to bet properly or double or split. so you find your self losing out on good counts.
a few times when i got low on money i tryed to perpously lose it so i could get some more money and ended up winning more money lol.

You're probably right. i have the strategy coach turned off and it seems I can bet my entire roll and if choose to double it just gives me more money.

Maybe not - you've fooled with it alot more than I have lol.

All in all, even if maybe possible, maybe not worth the effort lol.

 

[sagefr0g]

You're probably right. i have the strategy coach turned off and it seems I can bet my entire roll and if choose to double it just gives me more money.

Maybe not - you've fooled with it alot more than I have lol.

All in all, even if maybe possible, maybe not worth the effort lol.

if you don't have enough money the game won't offer you to double is one problem.
but i found one thing out trying to get to that point. if you purpously try to lose it wont let you unless you don't want it too! :joker:
oh wait a minute thats probably not the trainer that's just the nature of blackjack lol.

 

[KenSmith]

I think I was doing it a different way than you and was wondering if there was anything wrong, or if you could comment on, my method in assuming that dealer and player BJ's would be equally likely and doing something like .5^816* the sum of combin(816,0) thru combin (816,181)?

I originally used the Chi-Square test. Read more here:
http://en.wikipedia.org/wiki/Chi-square_test

It's probably more appropriate to use a Binomial Distribution though:
http://en.wikipedia.org/wiki/Binomial_distribution

What would be the chances of the dealer getting his 635 BJ's in his 7218 hands the way you are doing it?

Using the Chi-Square Distribution
Dealer result: 1 in 10,990,139,244
Player result: 1 in 15,376,735,730

Using the Binomial Distribution:
Dealer result: 1 in 4.30 X 10^47
Player result: 1 in 2.54 X 10^22

Hmmm... the 15 billion is different than the 21 billion I got previously, and I didn't save that spreadsheet so I don't know why there is a difference. Obviously I made an error one place or the other. I triplechecked my spreadsheet this time, so I'm pretty confident the error was in the initial result.

Could only playing +counts as I think kewlajckson did perhaps be a factor? Or including splits as a "hand"? Or perhaps only counting only winning BJ's?

Positive counts only should increase the number of blackjacks overall, and they'll still be evenly split between player and dealer. Counting splits as a hand would decrease the ratio of blackjacks to hands, but not nearly enough to make the results reasonable. Counting only winning blackjacks would have a minor effect as only pushed blackjacks would be excluded.

Neither here nor there but years back did you find some obscure error in something I think in a play-for-fun BJ game here? Just wondering how senile I'm getting lol. Probably not this game anyway.

I believe the error existed in both the Javascript and Flash versions here, and was corrected in both.

 

[Canceler]

Neither here nor there but years back did you find some obscure error in something I think in a play-for-fun BJ game here?

I believe the error existed in both the Javascript and Flash versions here, and was corrected in both.

I expect you guys are talking about this, http://www.blackjackinfo.com/news/bj022.php, which I found fascinating.

Ken,

Speaking programmatically about the BS Trainer...

Do you load the shoe with the proper cards, shuffle them, and then deal them out consecutively? From the above, it seems like you do. If so, I would be confident everything is okay. I mean, how much could go wrong?

This doesn't explain kewljason's results, though. Maybe he had that 1 in 21 billion (or whatever) experience, or maybe we should question his record-keeping. I don't know.

 

[sagefr0g]

.....This doesn't explain kewljason's results, though. Maybe he had that 1 in 21 billion (or whatever) experience, or maybe we should question his record-keeping. I don't know.

this is kind of what i've been wondering. just from a sort of common sense perspective. i mean i really don't understand this 1 in 21 billion stuff, the chi square probability stuff, the binomial distribution thing or what is that Kasi is referring combinetrics ?
but anyway isn't it the idea supposed to be or the rule of thumb that anything can happen in the short run?
so but i guess all the maths that's being referred to can analyze a small sample and give you an idea if it's bogus or not.
oh and i think kewljason's and the dealer results were what? close to 11% bj's? divide that by two and you have 5.5% bj's for that sample. so a little high but doesn't seem so abnormal. just the dealer got more of them and kewljason less. so i dunno just wondering if really a larger sample isn't called for?
oh yeah and i was wondering. wouldn't bj results present having a standard deviation? maybe just as wild appearing as betting results for blackjack game samples?

 

[sagefr0g]

I expect you guys are talking about this, http://www.blackjackinfo.com/news/bj022.php, which I found fascinating.
......

yeah facinating. i don't understand it though.
like:
"The new shuffling method is very similar, but no longer biased:
For each of the 52 spots in the deck, Swap that card with a card randomly selected between that spot and the end of the deck (n to 52).
Surprisingly, that subtle change of not allowing "backward swaps" makes the shuffle a random distribution"
what happens when you get to # 52 or the last card? or that one just retains it's identity?
and just wondering does that process just get ran once? would it serve any purpose to do it more than once?
oh and this all makes one wonder how is this differant from a real casino shuffle?

 

[Canceler]

what happens when you get to # 52 or the last card? or that one just retains it's identity?

Keep in mind that the card in position #52 could be anything at this point. At the beginning of the shuffle it had a 1/51 chance of being exchanged with the card in position #1. Then it had a 1/50 chance of being exchanged with whatever card was in position #2 (there's a small chance that card could have started out in position #1!). As the shuffle continues, it is increasingly more likely to be swapped out, until it has a 1/1 chance of being exchanged with whichever card is now in position #51. I can't prove it mathematically, but it seems likely that the card in position #52 has been swapped several times at this point.

and just wondering does that process just get ran once? would it serve any purpose to do it more than once?

Beats me. In my program, I first use the "swap anywhere" method, then run through again with the "no backward swap" method. Doing both may be overkill, but since in human terms it happens instantaneously, I figured why not?

 

[Kasi]

It's probably more appropriate to use a Binomial Distribution though:
Using the Binomial Distribution:
Dealer result: 1 in 4.30 X 10^47
Player result: 1 in 2.54 X 10^22

Thanks for your input, Ken.

Who's gonna quibble over 6 billion or so ? I'm not lol.

Yeah i didn't think any of that small stuff would change anything.

So I guess if we use the above, we are left with
1)stuff happens. In this case that's 400 trillion trillion trillion for the dealer but only 25 million trillion for the player in 7000 hands for each?
2) bad data?
3) game problem?

Just wondered what your thoughts may be or, just for argument's sake, say, had you recorded these results personally, what if anything you'd think about it.

No biggie.

Would you say if the game really is this non random in regard to BJ's, would you agree it wouldn't take alot more hands to corroborate?

I'm thinking some combination of 1 and 2 lol.

 

[KenSmith]

I vote for bad data. If these numbers are correct, something's badly broken. However, with all the other players checking it out, I just don't see that as realistic. In particular, with blackjacks that skewed, there's no way that the weekend warrior results would be as positive as they are.

As for whether the data sample has enough hands to know there's a problem, absolutely yes. Otherwise you can't generate statistical tests with these huge improbabilities.

Perhaps kewljason was confused somehow. Perhaps he was actually using a game on another website (I've gotten emails before about problems, and eventually found out they were talking about another website entirely!). Whatever the case, unless a second person sees results like that, I don't believe the numbers. Sorry kewljason.

For what it's worth, I intend to grab the shuffling portion of the source code and post it here, but it will be a few days before I have time to do so.

 

[Kasi]

this is kind of what i've been wondering. just from a sort of common sense perspective.
but anyway isn't it the idea supposed to be or the rule of thumb that anything can happen in the short run?
so but i guess all the maths that's being referred to can analyze a small sample and give you an idea if it's bogus or not.

I think you got it lol. I'm not sure I understand it all either except, like you say, do some statistical testing and see what happens. In this case, see above, and Ken thinks using the binomial distribution is more appropriate than Chi_square testing, and I'll take his opinion on that because I have no idea what's more appropriate to use, but if dealer results are 4 with 47 zeroes after it, even if only for 7000 hands, I probably wouldn't play that game again just on the off-chance even if anything can happen in the short run.

If you flipped a coin 1000 times and got 1000 heads, even though the sample is small, you probably aren't going to think it's a fair coin.

That combin stuff is just the Excel function combin(x,y). I was just assuming dealer and player BJ's would be equally likely without regard to number of hands played and trying to figure how likley getting 181 or fewer out of the 816 dealt was. So a little different than either of Ken's tests. One I could actually do lol.

I guess BJ's would have a stan deviation and, in this case, it's so many (I think) that basically the conclusion is the results, if true, are impossible in a fair game.

I also find it interesting that in total things aren't so bad lol. No idea what it means lol.

Would a real-life dealer heads-up who could deal seconds always knowing the next card end up with the right total of BJ's for the game but a huge amount of them for himself? Just a bs question but something to think about in pitch games with crazy results.

I mean, let's face it, on the face of it, if this actually were to happen to me on the internet in a real-money game, there would be no doubt whatsoever in my mind the game was non-random. And, if corroborated by game logs, I don't think anyone could possibly think otherwise. Believe me I kept track of dealer and player BJ's, winning and tied, and sure sometimes I'd be down 10 expected BJ's in a few thousand hands but not 80.

Am I exagerrating ? Ken? Sonny? QFIT?

It'd be nice to hear in more detail from kewljackson what he was doing and how he was doing it. Or maybe just confess he double-added some column or something.

So, while if true, an interesting discussion, I just don't think it's true.

 

[Kasi]

I vote for bad data. If these numbers are correct, something's badly broken. However, with all the other players checking it out, I just don't see that as realistic. In particular, with blackjacks that skewed, there's no way that the weekend warrior results would be as positive as they are.

I agree with you.

But still, to me, true or not, and I think not, it shows the value of knowing what you did and what happened and trying to relate it to something as best you can.

I don't see why you should bother posting your source code.

Maybe he was just yanking our chain thinking he could make up any results no matter how bad and just dismiss it as "really bad luck" under the sub-heading sh*t happens.

But now he knows we draw the line somewhere between 20 billion and 400 trillion trillion trillion to 1. :grin:

Who knows where the line should be drawn? I sure don't but I like to know about where it is if I can lol.

 

[Kasi]

if you don't have enough money the game won't offer you to double is one problem.

Fwiw, if you turn off the strategy coach , I think it will allow you to double anytime. Although, even so, I guess it would be a pain to manually keep track of the loss because I don't think it keeps a running balance for you. I think lol.

But, like you say, I doubled a 20 and won and then bet 3/4 ths of my money, doubled a 15 and won again and had to start over lmao.

 

[Canceler]

Fwiw, if you turn off the strategy coach , I think it will allow you to double anytime.

It'll allow you to double anytime you still have money left. But you'll be doubling for less if you don't have enough money.

 

[Canceler]

I don't see why you should bother posting your source code.

I'd like to see it, just for curiosity's sake.

Also for completeness. The question of the randomness of the BS Trainer comes up periodically, from people who think their results are either too bad or too good. We may want to refer back to this thread the next time someone feels victimized by the BS Trainer.

 

[sagefr0g]

kewl's stats....

from kewl's post:
http://www.blackjackinfo.com/bb/showpost.php?p=84803&postcount=9
"total 1500 shoes played and lost 1331 units. thank god this isnt the real world where my BR is roughly 1200 units! I have kept more detailed results over these last 300 shoes. Hands played 7218"
that's lookin to me like he on average plays just under five hands a shoe before bailing? that would be 7218hands/1500shoes. if 2.7 cards each for dealer and kewl only going 5 X 5.4 cards or 27 cards deep or are we talking maybe 7218hands/300shoes = 28 hands per shoe? if 2.7 cards each for dealer and kewl thats going 28 X 5.4 = 151, circa 151 cards deep before bailing on average. almost 3 decks in before bailing.....
lol i don't think we even know how many decks he's going up against.

as many recorded dealer & player blackjacks he's got near 11% you'd almost wanna believe it's single deck?