jackhigh78
Joined: Aug 30, 2013
• Posts: 4
August 30th, 2013 at 9:38:03 PM permalink
Hi,

I'm playing online EU (single zero) roulette at a local betting web site, they just open their casino. (casino is not a skin of popular world wide online casino systems. They write (code) their own casino, Flash based, AS3, and probably not tracking the results)

I tracked 720 spins. I only play 6 or 7 numbers betting 1 unit per number. Generally 7 numbers per spin.

Here are the results.

0 24 times
1 16
2 28
3 27
4 12
5 16
6 22
7 29
8 27
9 26
10 19
11 18
12 19
13 17
14 15
15 21
16 25
17 26
18 12
19 15
20 16
21 14
22 22
23 20
24 23
25 13
26 19
27 21
28 19
29 11
30 25
31 15
32 15
33 19
34 18
35 20
36 16

I am not a math wizard, can you please analyse these outcomes?

Second thing is, i know 720 spin is not enough data, but may be there is a different bias this roulette game has. The outcome of 0-9 numbers are very high than 10-36. This is online and they are using random number generator. I don't know the code and the algorithm, but they probably use milliseconds as seed to generate the new number. it looks like this algorithm generates 0-9 numbers more often than the others.

so, can someone make an analyse, chi-squared test for 0-9 and the others, Thinking 0-9 and 10-36 seperately may give us a better result for this game.

by the way , sorry about my English, but i think you got the point.

Thanks.
7craps
Joined: Jan 23, 2010
• Posts: 1977
August 31st, 2013 at 11:07:28 AM permalink
Quote: jackhigh78

I am not a math wizard, can you please analyse these outcomes?

so, can someone make an analyse, chi-squared test for 0-9 and the others,
Thinking 0-9 and 10-36 seperately may give us a better result for this game.

Thanks.

You will find what looks to be bias in every 720 spin sample.
Just the way it is with multinomial distributions

get Roulette Xtreme and plug in numbers all day long
It does the math for you

PASS

In other words, with NO bias in selecting a random number,
15% (15 out of 100) samples of 720 spins would look even
more biased than the sample you showed

I used Excel CHITEST function also
(This error is sometimes hard to see without my glasses on
=-CHITEST(B1:B37,C1:C37) returns -0.152327539)

0-9
p-value: 0.011322713 (1 in 88.3)
x2: 58.0718137
10-36
p-value: 0.594664809
x2: 3.36287605

0-9 p-value is about 1 in 88
the Roulette distribution is well known
1 in 88 ain't even kind of hosed IMO

A problem for online casino game programs does not have to be the rng
as many claim...
just the way the coder programmed the code to use the rng result.
what one can not see
(oops, forgot to comment out that test line where the casino wins 93.4% of the time,
Hey, let's just blame it on a bad rng!)

Good Luck
winsome johnny (not Win some johnny)
binary128
Joined: Nov 5, 2011
• Posts: 82
August 31st, 2013 at 11:21:42 PM permalink
I used Excel's CHITEST function, comparing against an expected value of 720 / 37 = 19.45946.

The CHITEST result against all 37 values was -0.1523. I confess I have never seen a negative result here before, but am too late into my day to do an investigation.

The result for the first 18 numbers was 0.0537. The result for the last 19 numbers was 0.4988.

It looks kind of hosed to me.

Chris
binary128
Joined: Nov 5, 2011
• Posts: 82
September 1st, 2013 at 4:23:37 PM permalink
Quote: 7craps

A problem for online casino game programs does not have to be the rng as many claim...
just the way the coder programmed the code to use the rng result.
what one can not see
(oops, forgot to comment out that test line where the casino wins 93.4% of the time,
Hey, let's just blame it on a bad rng!)

Agreed,

Awhile ago I was piling data into a thread over at thePOGG on this general subject. Our ultimate goal was to create a series of Casino 101 Questions & Answers. The first tricky bit was to "ask the right questions" such that they could be structured in a logical flow. That is, the answer(s) to Question #1 would not require knowledge that is presented in the answer(s) to Question #2. However, the answer(s) to Question #2 would, or could, build on the information presented in Question #1.

Each question would be answered in three parts. The First Part of the answer would be short, sweet, and to the point. It would not contain (unless also clearly defining) any Casino "meta-language", the use of which does NOT educate the reader but confuses, and thus isolates, them.

[Derail] A fundamental flaw in online casino forums - the natural tendency of the human species to form tribes, or cliques, and then reinforce tribal membership through the use of "meta-language". Newbees typically find themselves in the rather nasty position of having to pretend they understand, or trying to find some resource that will explain things to them. If they actually ask a member of "the forum tribe" for an explanation, some of those tribesmen will respond with ridicule and abuse. (I shit you not - I've actually seen the response - "If you don't know the answer to that question, then you shouldn't even be playing that game.") [/Derail]

Getting back on track - The First Part of the answer would be written such that it would be generally understood by the majority of thePOGG's visitors. The Second Part of the answer would contain a little more meat. The Third Part would contain the entire meal.

As I said, I was piling data into a POGG thread (linked above), but neither thePOGG nor I had the time necessary to do all of the formatting mentioned above - short answer, medium answer, long answer. (If you read the linked thread you'll undoubtedly conclude that there is a LOT of editing required here.)

Indeed, we were still going back and forth on the question for which all of this would be the answer. What is the hierarchy of questions? On what branch in this "Christmas tree" hierarchy would, or should, the "What is an RNG?" question be hung? Obviously, it would be briefly mentioned in various areas. However, to break analogies, what chapter in the book would be dedicated to the RNG?

Still, there is an impressive pile of stuff up there (if I do say so myself). Maybe when thePOGG comes up for air, or when I come up for air, one of us can revisit it and start the long process of trying to pull it all into a nice, neat, tidy presentation. Yes, the later material gets quite dense, but all of this would be parked under the Long Answer section and would only be read if the reader felt like giving it a go.

Chris
jackhigh78
Joined: Aug 30, 2013
• Posts: 4
September 3rd, 2013 at 11:08:32 AM permalink
Hi,

First thanks a lot for the answers.

I played 861 spins yesterday. It looks like my 0-9 theory fails or not favourite anymore according to first 720 spins. I found an excel sheet and enter the results.

R01=My first 720
R02= Testerday
R06= overall

and the chisquared results are
R01: 15,23%
R02:24,40%

R06:10,92%

I want to ask what number is the borderline for a chi squared test? 1% or 5%?

I want to try more spins to get better results, does it worth? need a suggestion.

Thanks.

EDIT:

NEW total 1990 spins and chi-squared test: 6,15%

beachbumbabs
Joined: May 21, 2013
• Posts: 14232
September 3rd, 2013 at 1:42:04 PM permalink
Quote: binary128

Agreed,

Awhile ago I was piling data into a thread over at thePOGG on this general subject. Our ultimate goal was to create a series of Casino 101 Questions & Answers. The first tricky bit was to "ask the right questions" such that they could be structured in a logical flow. That is, the answer(s) to Question #1 would not require knowledge that is presented in the answer(s) to Question #2. However, the answer(s) to Question #2 would, or could, build on the information presented in Question #1.

Each question would be answered in three parts. The First Part of the answer would be short, sweet, and to the point. It would not contain (unless also clearly defining) any Casino "meta-language", the use of which does NOT educate the reader but confuses, and thus isolates, them.

[Derail] A fundamental flaw in online casino forums - the natural tendency of the human species to form tribes, or cliques, and then reinforce tribal membership through the use of "meta-language". Newbees typically find themselves in the rather nasty position of having to pretend they understand, or trying to find some resource that will explain things to them. If they actually ask a member of "the forum tribe" for an explanation, some of those tribesmen will respond with ridicule and abuse. (I shit you not - I've actually seen the response - "If you don't know the answer to that question, then you shouldn't even be playing that game.") [/Derail]

Getting back on track - The First Part of the answer would be written such that it would be generally understood by the majority of thePOGG's visitors. The Second Part of the answer would contain a little more meat. The Third Part would contain the entire meal.

As I said, I was piling data into a POGG thread (linked above), but neither thePOGG nor I had the time necessary to do all of the formatting mentioned above - short answer, medium answer, long answer. (If you read the linked thread you'll undoubtedly conclude that there is a LOT of editing required here.)

Indeed, we were still going back and forth on the question for which all of this would be the answer. What is the hierarchy of questions? On what branch in this "Christmas tree" hierarchy would, or should, the "What is an RNG?" question be hung? Obviously, it would be briefly mentioned in various areas. However, to break analogies, what chapter in the book would be dedicated to the RNG?

Still, there is an impressive pile of stuff up there (if I do say so myself). Maybe when thePOGG comes up for air, or when I come up for air, one of us can revisit it and start the long process of trying to pull it all into a nice, neat, tidy presentation. Yes, the later material gets quite dense, but all of this would be parked under the Long Answer section and would only be read if the reader felt like giving it a go.

Chris

Chris,

If the House lost every hand, they wouldn't deal the game.
jackhigh78
Joined: Aug 30, 2013