Is this software fixed (SD calculation needed)

[scrollock]

ive been contacted by someone who has a bad run on baccarat, ive seen the logs and can verify the figures are accurate.

so whats the S.D. of a -44 units over 127 hands

 

[caruso]

I make it a little over 4.

 

[scrollock]

I make it a little over 4.

thanks caruso

i make that a 0.003 % chance or a 1 in 33,000 chance

have some new figures, dont know of they make it worse or better.

but are -60 over 212 hands of baccarat

if anyone is interested the software is chartwell

 

[paul1]

of a -44 units over 127 hands

Gosh, that's a pretty bad run of guessing on a coin flip. Bad guesser, I guess.

So what does this mean and what Caruso said? Scrollock, maybe you should drop a note to The Wizard of Odds.

 

[caruso]

Scrollock, maybe you should drop a note to The Wizard of Odds.

As usual, a statistical test is only valid if the parametres are established first, as in the Casino Bar experiment. In this case, assuming 33,000 to one is correct, you'd expect a result like this over 127 hands once every 33,126 hands if my recollection of the process is correct - you need to "reset" the test on each subsequent hand when you do it retrospectively like this.

If the player had set out to play 127 hands with intention of establishing proof of a fix, he'd have landed on a four million to one event. Pretty good evidence. As it is, this is just painful, not conclusive.

All of which is to say: don't go bothering Mike with this, 'cos it'll lead nowhere.

I've never played Chartwell baccarat, but I've played plenty of blackjack and have not the slightest doubt that the game is not randomly dealt. Can't prove it, nor am I much inclined to try. Just horrible software.

 

[paul1]

...he'd have landed on a four million to one event.

Sh*t happens.

 

[BBKPoker]

As usual, a statistical test is only valid if the parametres are established first, as in the Casino Bar experiment. In this case, assuming 33,000 to one is correct, you'd expect a result like this over 127 hands once every 33,126 hands if my recollection of the process is correct - you need to "reset" the test on each subsequent hand when you do it retrospectively like this.

You are close.

You'd expect -44 units *or worse* over 127 hands....etc etc rest of your post.

This is alarming but not entirely impossible. If this is a casino that uses independent software rather than one of the major software providers (Microgaming, Playtech, RTG, Boss, Wagerlogic, etc) I'd personally chose to avoid any further play there.

 

[Dirk Diggler]

I had a run at Wagerworks BJ where I was 81 units up after 400 hands/units.

Works out to be over 3.5 SD.

Just goes to show these 'bad sessions' can go both ways.

 

[crovax4444]

SD doesn't mean anything if you can't give me a confidence interval. Who cares if you get a SD of 50 if your CI is only 2%. Give me something that is 99.5% CI and tell me the SD and then tell me Charterwell is cheating.

Crovax

 

[GrandMaster]

What do you mean?

 

[crovax4444]

I'm basically asking for a larger, overall sample size. Anything small doesn't cut it. It's like flipping a coin 4 times, getting tails all 4 and claiming the coin is rigged.

Crovax

 

[tencardcharlie]

The thing is that you don't get 4 standard deviations from your expected result by flipping a coin four times - ever. When you already know the SD, you don't really need the sample size. The number of events is already taken into consideration when calculationg the standard deviation.

However, as someone posted before, you need to set the parameters before you make the test. Either that or get a ridicoulusly improbable result, like 50 SD down.

I don't think chartwell is cheating either. But, man is that guy unlucky?...