Author Topic: Testing a strategy  (Read 9021 times)

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1600
  • Thanked: 652 times
  • Gender: Male
Testing a strategy
« on: July 02, 2015, 06:33:38 PM »
I hear and read people trying to test strategies.
They even buy programs to do it.
I'm not sure how they try to test roulette strategies and I'm sure each one does it differently.
Some use excel, some rx, some the free online applications.
And I guess they can use different numbers of spins. Some may test with pen and paper for a few hundred spins, some other may test a few thousand spins, some test millions of spins and some may even test billions.

I believe testing a system is a very delicate thing. A very difficult thing. And it is most often done wrong. That's why, although we have posted a very detailed article about system testing , I believe it is much more important to deeply understand a strategy than test it. Understanding doesn't mean just knowing the rules, but identify the main concept, the weaknesses and strengths, and the kind of sequences that can make you or break you.

As to those who believe that if they run a system for millions of spins is a good test of its effectiveness, I want to say this:
There is no point in testing millions of spins. It can tell you nothing about the system. It only tells you about the house edge.
It is like trying to test your health by asking if you will be alive in 300 years. This can tell you nothing about your health. It only tells you you are mortal.

"In the long run we are all dead"
 John Maynard Keynes
« Last Edit: July 02, 2015, 07:01:19 PM by kav »


 

Marigny

  • New
  • **
  • Posts: 12
Re: Testing a strategy
« Reply #1 on: July 02, 2015, 07:12:52 PM »
Kavouras is absolutely right. If you had a method that would win over a million spins, it might be a consistent loser at a hundred spins.
Years ago when the Apple was the computer of the day I was determined to beat Horse Racing. I fed in information to a multiple regression program until the computer memory was full. I bought a newer computer with twice the memory and continued. When I tried to get it to make a selection it picked the most likely horse in every race. Every one was either the first or second favorite. I had given the computer so much information it had smoothed out all the bumps where the longshots won.
 

BlueAngel

  • I always express my opinion
  • Hero Member
  • ******
  • Posts: 1437
  • Thanked: 196 times
  • Gender: Male
  • Do you want truth? You cannot handle the truth!
Re: Testing a strategy
« Reply #2 on: July 02, 2015, 07:34:26 PM »
I strongly agree, understanding the principle which is being translated as the pros and cons of a strategy is essential!
I cannot stretch enough how much important it is, it could save you much time,money,effort (emotional,mental,physical)!
I test my methods manually because I don't know how to program, nor I can trust easily a programmer.
I've a full guide which explains step by step how to code, from the simplest to the most complicate strategy for the RX software.
I have to admit that I'm a bit lazy to study it and use it, for those who would find useful such instructions for scripting on RX, I have attached here.
Enjoy!
 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1600
  • Thanked: 652 times
  • Gender: Male
Re: Testing a strategy
« Reply #3 on: July 02, 2015, 08:05:58 PM »
Here are some free testers:
http://www.loothog.com/
 

BlueAngel

  • I always express my opinion
  • Hero Member
  • ******
  • Posts: 1437
  • Thanked: 196 times
  • Gender: Male
  • Do you want truth? You cannot handle the truth!
Re: Testing a strategy
« Reply #4 on: July 02, 2015, 08:34:09 PM »
I had a problem to upload the file due to size restrictions,but I've reduced it now it's ok.
 

dobbelsteen

  • Hero Member
  • ******
  • Posts: 1218
  • Thanked: 286 times
Re: Testing a strategy
« Reply #5 on: July 03, 2015, 08:12:05 AM »
Kav I fully agree  that long run test can not give us the answers how we can beat the roulette.

Long run is relative. For a EC test 200 spins are enough, for a number test 1M is not sufficient.

There are different methods and programs to test ideas.For a fair and honest judgment we need exactly the input of the programs.
Is the input rubbish, the cumputer output is also rubbish!!!
 

Harryj

  • Veteran Member
  • ****
  • Posts: 353
  • Thanked: 156 times
  • Gender: Male
Re: Testing a strategy
« Reply #6 on: July 03, 2015, 07:08:47 PM »
   I too test all my ideas by hand, using sessions drawn at random from my previous records The fact that the recods I use are spread across a number of casinos and time tends to make the results more valid.

     Like Dobbelsteen and Kav  I do not believe computer tests of massive amounts of spins has any basis in reality. We play in relatively short sessions. Each session comprises many games or betting series. I have often suggested that those who do these marathon tests should break them down ito more realistic segments. 100 spins for the sessions and 5-10 spins for each game. unfortunately that wouldn't give then the regulated results they are looking for.

     This cannot be said strongly enough......

      THE SHORT TERM PROBABILITY AND THE LONG TERM PROBABILITY ARE NOT THE SAME. MATHEMATICS CAN'T SEPARATE THE TWO.

   Harry

   
 

dobbelsteen

  • Hero Member
  • ******
  • Posts: 1218
  • Thanked: 286 times
Re: Testing a strategy
« Reply #7 on: July 04, 2015, 09:27:38 AM »
Harry I completely agree with your statement about short run and long run samples. The problem is we have a very few followers,
Most visitors on the fora are gold-diggers lookingf for the HG of a system or strategy The lessons from experence players are for deaf ears.
 

Mike

  • Veteran Member
  • ****
  • Posts: 557
  • Thanked: 23 times
Re: Testing a strategy
« Reply #8 on: July 04, 2015, 09:57:24 AM »
Kavouras is absolutely right. If you had a method that would win over a million spins, it might be a consistent loser at a hundred spins.

Sorry, but this is absurd. If a system was a CONSISTENT loser at a hundred spins it couldn't possibly be a winner over a million.
Quote
THE SHORT TERM PROBABILITY AND THE LONG TERM PROBABILITY ARE NOT THE SAME. MATHEMATICS CAN'T SEPARATE THE TWO.

So short sessions don't add up?

I'm starting to feel like this forum is something out of Alice in Wonderland.
 

Mike

  • Veteran Member
  • ****
  • Posts: 557
  • Thanked: 23 times
Re: Testing a strategy
« Reply #9 on: July 04, 2015, 10:12:32 AM »
I believe it is much more important to deeply understand a strategy than test it. Understanding doesn't mean just knowing the rules, but identify the main concept, the weaknesses and strengths, and the kind of sequences that can make you or break you.

As to those who believe that if they run a system for millions of spins is a good test of its effectiveness, I want to say this:
There is no point in testing millions of spins. It can tell you nothing about the system. It only tells you about the house edge.
It is like trying to test your health by asking if you will be alive in 300 years. This can tell you nothing about your health. It only tells you you are mortal.

"In the long run we are all dead"
 John Maynard Keynes

Kav,

While I agree that it's important to understand a strategy, it seems to me that there is no conflict between understanding and testing over millions of spins. If you understand what it is that makes a strategy lose or win, then surely testing it over a million spins will confirm it - IF your understanding is correct. The value of long term testing is to find out whether your assumptions are true and/or whether your understanding  is only partial, or flawed.

Lots of member here assume, for example, that "virtual bets" give them an advantage, but long term testing (and it needn't be MILLIONS of spins, a few thousand is usually enough) shows that their assumptions are false.

It can be difficult to see the big picture with a system, especially if it's complicated, and intuition very often leads you astray.
« Last Edit: July 04, 2015, 10:14:23 AM by Mike »
 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1600
  • Thanked: 652 times
  • Gender: Male
Re: Testing a strategy
« Reply #10 on: July 04, 2015, 10:18:39 AM »
I'm starting to feel like this forum is something out of Alice in Wonderland.
That comment was a giveaway "Mike".
Derogatory comments about the forum are just plain rude and say more about you than the forum. Enough!
This forum is the most thought provoking forum there is about roulette.
Create yours where you teach each other things you already know.

Quote
The value of long term testing is to find out whether your assumptions are true and/or whether your understanding  is only partial, or flawed.
I thought to you all system were the same and all conformed to the Average expectation over the long run. so what's the piint of your "testing" if every system produces the same result.

Quote
Lots of member here assume, for example, that "virtual bets" give them an advantage, but long term testing (and it needn't be MILLIONS of spins, a few thousand is usually enough) shows that their assumptions are false.
I guess in your opinion NOTHING produces an advantage, right? So why bother with testing systems in the first place?

 
« Last Edit: July 04, 2015, 10:24:45 AM by kav »
 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1600
  • Thanked: 652 times
  • Gender: Male
Re: Testing a strategy
« Reply #11 on: July 04, 2015, 10:30:02 AM »
Kavouras is absolutely right. If you had a method that would win over a million spins, it might be a consistent loser at a hundred spins.

Sorry, but this is absurd. If a system was a CONSISTENT loser at a hundred spins it couldn't possibly be a winner over a million.
Quote
THE SHORT TERM PROBABILITY AND THE LONG TERM PROBABILITY ARE NOT THE SAME. MATHEMATICS CAN'T SEPARATE THE TWO.
So short sessions don't add up?
I'm starting to feel like this forum is something out of Alice in Wonderland.
Let me give you a very rough example why this point is valid.
Imagine a 18 step Paroli, where you need 18 consecutive wins to win it.
This can be a consistent loser for years, yet a few wins can turn the overall balance into profit.
 

december

  • Veteran Member
  • ****
  • Posts: 375
  • Thanked: 39 times
  • Gender: Male
Re: Testing a strategy
« Reply #12 on: July 04, 2015, 10:37:36 AM »
@Mike

If you know "what it takes to" compete with Roulette, you can do following (regarding this forum):
- keep your knowledge to yourself
- share it and help others
...
Just don't be wiseguy here, please. And talk so wise and from above - you will not impress anyone here with that kind of smartness!
As someone somewhere already said:
"It is like coming to a church on Sunday mass to teach all the religious people that there's no God!"
 

Mike

  • Veteran Member
  • ****
  • Posts: 557
  • Thanked: 23 times
Re: Testing a strategy
« Reply #13 on: July 04, 2015, 10:38:45 AM »
I guess in your opinion NOTHING produces an advantage, right? So why bother with testing systems in the first place?

I've just told you why - so you can understand them and/or check that your understanding conforms to reality. Testing is the ultimate reality test, numbers don't lie.

Quote
Let me give you a very rough example why this point is valid.
Imagine a 18 step Paroli, where you need 18 consecutive wins to win it.
This can be a consistent loser for years, yet a few wins can turn the overall balance into profit.

But that sounds like an argument FOR long-term testing, not against it, which is the point I thought you were trying to make.

BTW, what's with the "Mike" in quotes? do you think I'm Real? LOL
 

kav

  • www.Roulette30.com
  • Administrator
  • Hero Member
  • *****
  • Posts: 1600
  • Thanked: 652 times
  • Gender: Male
Re: Testing a strategy
« Reply #14 on: July 04, 2015, 10:44:53 AM »
But that sounds like an argument FOR long-term testing, not against it, which is the point I thought you were trying to make.
I wanted to explain to you how it is possible this
 
Quote
If you had a method that would win over a million spins, it might be a consistent loser at a hundred spins.