### Author Topic: Statistical Question @ Palestis  (Read 601 times)

0 Members and 1 Guest are viewing this topic.

#### Third

##### Statistical Question @ Palestis
« on: May 30, 2019, 04:11:58 AM »
Hi!  Anyone can answer this question according to their liking!

I know that the "hit and run" doctrine is looking for new short term statistical opportunities but what if we are playing with long-term statistics as part of our system?  For example, I know the favored long-term statistical results but have surpassed the cumulative probability for WIN/LOSS ratio and would expect a series of losses due to this.

Should I dump all my stats (long-term) just to follow the short-term WIN/LOSS probability?

I am in a rush so if anyone needs to ask further questions, feel free!

The following users thanked this post: fiben7

#### MickyP

##### Re: Statistical Question @ Palestis
« Reply #1 on: May 30, 2019, 05:05:16 AM »

Long term stats are important to set parameters for a method. The short game is enveloped in long term stats and testing. Long term stats are important to identify the entry for a short game.

The following users thanked this post: Third, fiben7

#### Third

##### Re: Statistical Question @ Palestis
« Reply #2 on: May 30, 2019, 07:01:14 AM »
OK!  Well in this case, I played my short session and received a good number of wins, which took the cumulative probability of WINS/LOSSES to a point that is lower than 50% chances.

UPDATE: In this particular session, I ended up taking it ridiculously low to where the W/L register was at 9%, where 33% represents expectation and the total number of hits vs. spins was at 25 (instead of 37).  So, in this case the long term stats + alternative bet selection (when the main selection has failed), took me into profit, thank God!  Just for information's sake, when I first asked the question, I had around 325 spins of data collected.

Also in this session, I saw a number hit 22 times, less than expectation, in a row!  The chances of this are .9999647138 or 1:28,340 hits!!!  I only got on that train for the last 6 hits but it took me into my recovery goal!

I ended up abandoning the session with about 800 spins of data, due to fatigue.  I guess my question revolves around the value of long term data.  I currently use 2.5 methods and the data applies to them all; the reason it matters to me is because when I start a new session, it can take awhile for the statistics to assume a certain recognizable pattern; new sessions seem to flounder around a bit, more often than not and it concerns me to "throw away" such valuable work just because the session is exceeding short-term statistical expectation...?
« Last Edit: May 30, 2019, 07:11:13 AM by Third »

The following users thanked this post: fiben7

#### scepticus

##### Re: Statistical Question @ Palestis
« Reply #3 on: May 30, 2019, 08:00:31 AM »
Where statistics is misunderstood is that they cannot prove that a method is false  or valid . It only means that the maths SUPPORT the theory - or doesn't .  Even if a method does not win during your tests it does not mean that it is worthless -  only that it is unsupported .

#### Joe

##### Re: Statistical Question @ Palestis
« Reply #4 on: May 30, 2019, 01:44:41 PM »
I think the word "proof" in this context is misleading. Where stats are concerned you can certainly show that a method works or doesn't work "beyond reasonable doubt", depending on how long your tests are.

#### scepticus

##### Re: Statistical Question @ Palestis
« Reply #5 on: May 30, 2019, 04:42:23 PM »
Joe "  You must have sufficient data  " has been parroted in this forum many times  yet no one has  said how much is sufficient data .
Solaris tested the 9 block  and  his tests showed that the 9 block   beat the House Edge . That  does not mean that it will continue to do so in the future,only  that it is LIKELY to continue to do so in the future.

Variation rules - not  stats which need to be interpreted .  Stats are a guide  not the Holy Grail which some would have us believe.

#### palestis

##### Re: Statistical Question @ Palestis
« Reply #6 on: May 31, 2019, 12:32:50 AM »
@Third.
I think there is a difference between long term statistics, and long term tests.
I am not very concerned with long term statistics because the results will be equilibrium or very close to it.
However the value of long term tests yields to invaluable information.
Above all it will reveal the max. back to back losses of the system you are testing.
More specifically how often you will run into this max. Also the most frequent average number of B2B losses. Which most likely you will encounter much more frequently than the max.
Also the win/lose rate of the system.
If you know the max. possible back to back losses that can happen (based on long term tests),
you are not going to encounter any new surprises when you play in the short run. (unless you run into a new record once in a blue moon).
And it doesn't matter whether you play flat bets or progression.
The known long term test results will guide to the next course of action.

The following users thanked this post: kav, Third, Astutillo, fiben7, Jamndy

#### Joe

##### Re: Statistical Question @ Palestis
« Reply #7 on: May 31, 2019, 09:49:25 AM »
Joe "  You must have sufficient data  " has been parroted in this forum many times  yet no one has  said how much is sufficient data .
Solaris tested the 9 block  and  his tests showed that the 9 block   beat the House Edge . That  does not mean that it will continue to do so in the future,only  that it is LIKELY to continue to do so in the future.
Scep, if you don't know how much data is sufficient, how can you assert that your 9 block beat the house edge? maybe the tests didn't use sufficient data!

It depends on various factors like whether you use a progression and how many numbers you bet as to what constitutes "sufficient". There is no one size fits all answer.

#### Joe

##### Re: Statistical Question @ Palestis
« Reply #8 on: May 31, 2019, 09:55:40 AM »
I think there is a difference between long term statistics, and long term tests.
I am not very concerned with long term statistics because the results will be equilibrium or very close to it.
palestis you seem to have a very narrow view of what statistics are (I've noticed this about other forum members too). You can get statistics on anything including back to back losses or the results of any tests. There does seem to be a lack of understanding and imagination about what stats are and how they can be used. Statistics are the key to success, in my view, if used creatively.

#### scepticus

##### Re: Statistical Question @ Palestis
« Reply #9 on: May 31, 2019, 12:17:10 PM »
Joe
I didn't have any data .  Solaris tested it and " proved" it - not me.  Makro said he would test it but didn't .

#### palestis

##### Re: Statistical Question @ Palestis
« Reply #10 on: May 31, 2019, 09:04:25 PM »
I think there is a difference between long term statistics, and long term tests.
I am not very concerned with long term statistics because the results will be equilibrium or very close to it.
palestis you seem to have a very narrow view of what statistics are (I've noticed this about other forum members too). You can get statistics on anything including back to back losses or the results of any tests. There does seem to be a lack of understanding and imagination about what stats are and how they can be used. Statistics are the key to success, in my view, if used creatively.
I don't know what you mean about narrow view. Statistics is the key to success if you can filter out information that you can use to win. Unlike other types of statistics, long term roulette statistics will always lead to equilibrium. Between EC's, dozens ,columns, DS's, streets and among the 37 numbers.
How can you use this information creatively?
If you see a great imbalance in the short run, you can apply this equilibrium fact, not for a total equilibrium, but for at least one successful hit per trigger involving a great imbalance.
But it is the back to back losses that destroys the bank roll.

The following users thanked this post: Third

#### Mako

##### Re: Statistical Question @ Palestis
« Reply #11 on: May 31, 2019, 09:09:34 PM »
Joe
I didn't have any data .  Solaris tested it and " proved" it - not me.  Mako said he would test it but didn't .

Yeah sorry scep, I've been testing two methods that have taken up most of my time this year and I haven't had a chance to break out the lego blocks yet.  I will though, it intrigues me.  When was the last time you were in contact with Solaris?

#### Third

##### Re: Statistical Question @ Palestis
« Reply #12 on: June 01, 2019, 02:49:06 AM »
I saw you edit that quote.

#### Third

##### Re: Statistical Question @ Palestis
« Reply #13 on: June 01, 2019, 10:55:54 AM »
Btw, just an update.  I am currently using the hit and run model because it is what has brought me success.  When the Win/Loss registry shows <50% chances, I simply transfer any remaining debt to a new session.  As far as losing the data, I am counting on standard probability for my wagering plan (which is just over 80%) against the fresh data (this has worked in the past).

The only thing I am doing differently is taking debt to the new session, instead of closing all sessions with no debt.
« Last Edit: June 01, 2019, 10:59:43 AM by Third »

The following users thanked this post: Astutillo, fiben7

#### Joe

##### Re: Statistical Question @ Palestis
« Reply #14 on: June 01, 2019, 01:33:46 PM »
.
« Last Edit: June 01, 2019, 01:36:20 PM by Joe »