Question about proving a system wins

[N0vocane]

How many spins does it take without showing a loss on a system to prove that it actually works and beats the casino?

1000?
5000?
10,000?
20,000?
30,000?
more?

I don't have one...yet, but I need to know how many spins I need to run when I test a system to know it actually works.

Thanks

[albalaha]

There can't be a true and perfect parameter of this since every session is different from all other millions of sessions, but in my opinion if some system clears zumma tester's book(american roulette) having 15,000 spins with net profits without using very high or impractical progressions, the system is good.

[hanshuckebein]

maybe you should have a look at the so called "van keeen test".

link:://:.casinoclubmagazin.com/?p=1226

this is a german page, but I guess google could provide you with something similar written in english.

cheers

hans

[esoito]

Here's my summary translation -- enough for you to get the gist:

(My apologies to our German friends  )



The Van Keelen Test

Every serious system player is interested in an objective method of determing profitability in the long run.

The Van Keelen Test is a simple measure to determine the chances of success or failure of a system.

It was was developed decades ago and will give the player information about the value of a system.

It requires the player to check the system over a certain number of placed bets before playing.

Losses then only develop on paper and not in real time at the table.

The Van Keelen Test establishes a relationship between the net gain of a system to the number of placed bets.

The minimum number of the placed bets checked for all chances must amount to at least 1000.

The units wagered are flat bets. No progression.

This testing method has the distinct advantage that a system must be checked at least over 1000 placed bets. Then a prediction about the value of the checked system is possible.

Basically, if a system wins with flat bets then it can be adapted to a progression for leverage.

Here are some suggested indicators:

Even-Money-Bets
A system could be successful
1. if after 1000 placed bets there's a profit  of 100+ units
2. if after 8000 placed bets  there's a profit  of 60+ units
3. if after 100000 placed bets a profit better than 1000

Dozens/Columns
A system could be successful
1. if after 1000 placed bets there's a profit  of 50+ units
2. if after 8000 placed bets  there's a profit  of 50+ units
3. if after 100000 placed bets a profit better than 2000

Six-Number Lines:
A system could be successful
1. if after 1000 placed bets there's a profit  of 100+ units
2. if after 8000 placed bets  there's a profit  of 200+ units
3. if after 100000 placed bets a profit better than 600

[Blood Angel]

Thank you esoito  

[Proofreaders2000]

I remember Flat Ino saying if a system ends in profit after 400 actual bets (on a live wheel), then it's a winner.

[A3on]

I remember Flat Ino saying if a system ends in profit after 400 actual bets (on a live wheel), then it's a winner.

I agree and disagree with that.

I agree if the system is only flat bets.
I disagree if the system uses progression (it's really easy to end up in positive after 400 bets with some crazy progression)

[flukey luke]

If you like playing dozens or columns (or any combination of 12 numbers), here is a good test of a systems worth.

Expectation is 32.43 wins out of 100 spins, therefore 37.11 wins equals one standard deviation. 38 is the nearest whole number greater than one STDEV and would represent a performance of 1.2 STDEV above expectation.

The chance this would occur randomly is 1 chance out of 8.

If you were able to repeat the performance, you would have won at least 76 spins out of 200. This would, in total, be a 1.68 STDEV performance with a chance of randomness of 1 out of 22.

If you could do this for a third time and get at least 38 out of 100 correct, your total performance over the 300 spins would have been at least 114 correct.

114 correct out of 300 spins represents a STDEV of 2.06 with a chance of randomness of 1 out of 50. In other words, if you were to take this same challenge 50 times, you would only acheive 114 correct out of 300 spins once by random chance. It would even be rarer to do it in three consecutive sessions of at least 38 correct predictions per session.

So if you can achieve all that, you may be on to something.

[mr.ore]

No system can be winning forever, so what we should really look for is "quality".

According to ISO 9000, quality is defined like this:
"Degree to which a set of inherent characteristics fulfills requirements." The standard defines requirement as need or expectation.

So we should first ask, what we want from a system. Do we want to win as much as possible, or play as long as possible, or never win or lose, but be as near zero balance as possible? Those requirements are all contradictory, and gaining in one leads to missing in another.

So how should we measure system's quality? I think, that one such a measure might be a MTBF(mean time between failure), that would mean how many spins/placed bets/units on average are needed for the system to get to a state, where it would not recover any more.

link:://en.wikipedia.org/wiki/MTBF

To measure that, many thousands of simulated sessions are needed, and then compute average. What should be also simulated is starting bankroll, target, and table limits. If comparing two systems, those things should be the same for each session. Bankroll should be always lost, if session failed, and every unit should be used to recover. Sessions could have subsessions(stoplosses, recover subsystems and other stuff), of course, but that's the problem of that particular system, measurement should be alway taken when target reached, or system tanked. What should be really measured is an ability to reach target, since a lot of good systems might be losing most of the spins, but end up winning target more often.


Now thought about requirments:

If you want a system with max. probability of winning, just make a parachute that will win in one hit, so that it win at least the target, or maybe that average win is the target. The hole is often deep, but it win most of time. Use as few units as possible to win the target bankroll, so that house edge eats as few units as possible. Or just play max. unit allowed on a single, stop maybe -400 or +350 units. MTBF should be best.

Some people want to play as long as possible, but are afraid of losing, for them a quality would be howering around zero with parachute, waiting for a streak of luck, and with that use a positive progression(parachute again), increasing both units and covered area, ending with a new level to hower on.

Most of players dream of a winning system where there are many small wins, like winning one unit every few spins, so that their bankroll grows lineary. It looks nice and is rewarding and addictive. But those systems are the worst, because too many units are exposed to house edge to win just one or few units. Those systems have very high probability of winning that one unit, and players are trying to find "bet selection" which should prevent too long losing streak. Those systems are the biggest illusion. Why are those systems so prefered? Because if the game behaves, and that is most of the time, those systems works pretty well, better than those "math" based parachutes. That's because there is fake "advatage" caused by random. Played with big units, it is nice to win maybe +20 - +30. But MTBF is the worst.

Another question is a bankroll. How big it really is, and is it ok to consider it infinite? Having given table limits and allowed unit increment, what is the greatest MTBF system can have with an infinite bankroll? How much is it even realistically possible to win on a given table? Table limits defines maximal bankroll size, that if you lose so much, recover would be almost impossible.

Good tip - compute how many units can you win with particular's table limit with probability ~50% (you have to find a method where units won ~ units loss) and treat that as an even chance. That particular amount of money is a bankroll size for a session. Now remember from experience, how many units you can usually win flat betting even chance. We all know, that there are holes, that cannot be recovered with flat betting it. Progression is not possible now, because of limits. That leads to a simple fact MTBF depends mainly on the table's limits. Table limits also says, how much can be realistically won.

Practically - code, simulate, measure MTBF.

[mr.ore]

Now reading my post again, I little messed it up. MTBF idea is not that simple, it should be "normalized". System that tanks once 1000 spins on average but on average NET is 100 units is wore than a system with MTBF=500 spins, but average NET is 60 units, because 100/1000=0.1, 60/500=0.12 . Hmmm.

[Skakus]

If you can profit on the even chances after about 50,000 placed bets then you probably have a decent system.

[mr.ore]

The Van Keelen Test

it seems like a good measurement for bet selections. I cannot imagine how could such a selection work, though. There is no reason why should it be possible. Well, if you try, you can have that selection showing superiority over zero, but it is because of luck. If you trace all ECs and jump from one to another when their graphs are crossing, it can happen. Maybe also jump if too much high in positive  SD side and it starts losing. Then a great illusion can be created, I once had a graph which was losing half the house edge speed. Another 100000 spins, and "advantage" has gone.

Maybe if there were some long term effect from dealer's signature, or casinos attempts to screw winning streaks of a players, that could be exploited. Changing wheels or dealers should not have any effect, but why are they doing it? Can it really distract winning streak? Today's wheels should not never be too biased because of random, if they could(?) tamper the random results...

Actually, bet selection for progression should be measured differently from flat betting system...

[Calypso]

hmm .  i guess the best way to know if a system is a winner .  u need to calculate the z-score

just google z score

[GARNabby]

Like happiness, if you have to ask yourself about it, or analyze it...  you're not.

Learn to feel it, just by staying in the game as long as possible by not "going for broke" (by going for big, consistent wins.)


G. (I don't win 10 units a day either, lol.)

[A3on]

Like happiness, if you have to ask yourself about it, or analyze it...  you're not.

Learn to feel it, just by staying in the game as long as possible by not "going for broke" (by going for big, consistent wins.)


G. (I don't win 10 units a day either, LoL.)

I would be more than glad to make 10 units a day *.*
If it was by flat bets only each unit can be a BIG €
:p