### Author Topic: Basics Of A Quantitative Roulette Strategy  (Read 4746 times)

0 Members and 1 Guest are viewing this topic.

#### MrPerfect.

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #30 on: May 20, 2019, 09:39:02 AM »
Fiben7,  it's a pleasure to meet another human. Use of brain is extra activity that many sims to escape by default.
Coming back to that case and creation of roulette systems for specific wheels...
The thing is that average conventional math available for such a task is normally based on wrong model of assumptions. Everyone thinks that wheels are perfectly round, ball has same chance to hit any number and jump after that to any other number... wich is as far from reality as it can be. It just not happening, phenomenon not observed in nature...period.
Wheels are deffective,  poket compartments are different, poket pads are not leveled and are not same level in relation to each other, balls are not perfectly round, ball tracks are uneven...ets.
Almost every part of roulette wheel can/do possess deffects/ differences that can and often do affect final number result. There is 1:37 for no one, as there are no perfect wheels.
This "perfect wheel model" where 1:37 is true is no more then imaginary abstraction in minds of pseudo scientists and serve for no purpose if not for a purpose of comparison to determine how results generated by real wheel are different from results that could be generated by its imaginary model of perfect wheel.
Yet... everyone try to develop systems to beat this imaginary model instead of focusing on real wheel. No need to mention that their results of such systems are slightly out of reality focus.
Math is just a tool... like a hammer,  it can be used for good or just harm it's own user.

The following users thanked this post: fiben7

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #31 on: May 20, 2019, 09:41:14 AM »
I see your point dobbelsteen, thanks a lot for your insight.

Your logic resembles my basic approach in quantitative trading in finance where I use non-linear analysis, mainly chaos theory and fractals, in order to model and forecast short-term "normality" in a chaotic long term series.

In other words, there can be found a robust temporal geometric structure (short sessions as you call it) that can be measured and is predictable even if the longer term geometric structure (randomness) is not.

My aim, after working on more simple systems on VOT, R/B, H/L, Dozens etc. Trend and Mean Reversion analysis, is to transfer  a simplified version of such complex fractal mathematics into something playable and executable on the roulette.

As a general remark, it is very very encouraging for my real playing and systems, that my recorded spins up to now (wish I had 100 times more of these) exhibit all the theoretical features of randomness, however there appear short sessions in which my systems overperform and others where they underperform; always though in such a way that the overall risk/return metrics are positive and relatively smooth.

Thanks again for your time on my post

The following users thanked this post: Third

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #32 on: May 20, 2019, 10:04:05 AM »
MrPerfect., totally agree. No such thing as a perfect flawless roulette wheel that can yield exactly 1/37 on each number over the long run. This is a fact and professional players should deal with it. This fact alone provides an opportunity of Edge.

Thanks for your kind words as well.

My initial approach on the roulette (I am not very experienced on the game I have to say, only recently - 1 year -started researching on it), was to look for a flat-bet system (thus easier to control and execute and more robust in risk management) that does not focus on specific numbers or tries to find complex patterns and exquisite triggers, but uses simple statistic concepts that are executable easily and have a finance driven theoretical background.

This theoretical background of course is NOT that numbers have 1/37 chance to appear or Black has 18/37 chance to appear etc.

A point to make here, I agree with your quote of how a roulette wheel and its parts could have structural deficiencies that provide an edge. The edge most probably is an area on the circle and not a point. Thus, better have a system based on areas. Then, filter these areas with other systems and statistics that simultaneously minimize the required bet (say from 8 chips to 4) and maximize the return (say from 18:1 to 36:1). This will of course lower the hit ratio of the system, as less numbers will have been covered. However, if that hit ration is even slightly better that the required bet/expected return ratio, the main payout ratio that is, then in the long run the system is profitable. It has translated the edge, the structural deficiencies, into meaningful statistical properties that of course do not always win, that have temporal drawdowns, though recovery is steep and performance positive.

I wish I had more time to test my systems in the field and produce new ones, however as per your initial post on my thread, I am busy making money in finance (and not cryptos). I am running an algorithmic trading fund in foreign exchange markets and also manage two start-ups on AI and automation tools. No show off here, I am just explaining the mechanics of my day to day engagements that keep me away from such a challenging goal of beating a roulette!!!

The following users thanked this post: MrPerfect., Third

#### dobbelsteen

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #33 on: May 20, 2019, 11:33:47 AM »

Fiben7: VdZ has 17 numbersand is a 17 number bet. Orphelins has 8 numbers and is a 8 number bet. Tier has12 numbers and is a 12 number and equal to the dozens and columns. The DTOP is larger than 250 spins This means that all your trials are short sessions. Win and loss  will flucuate very much. Results of such small sessions can not be used for conclusions. All trials of systems larger than DTOP end  on average with a loss of 2,7% !!

This explain why your results over-  and underperform

The following users thanked this post: fiben7

#### MrPerfect.

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #34 on: May 20, 2019, 11:39:14 AM »
That's the main reason why there are not many AP out there. If one can do it, he can do it else where as well and with much less hassle.
Math skills required to perform AP are same as to run investment fund or insurance company, become highly payed market analytic or top engineering pro.. who care about roulette if money with less risk is available elsewhere?
I myself slow but steady moving towards becoming market player. Last year alone l made more money on markets then during 7 years in roulette.

The following users thanked this post: fiben7

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #35 on: May 20, 2019, 11:59:45 AM »
great stuff MrPerfect, well done!

sure, logical rules and risk/return metrics analysis are applicable in many industries and markets, one should not constrain himself just to one.

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #36 on: May 20, 2019, 12:20:29 PM »
Dear dobbelsteen,

I see and understand your point. Please allow me for some clarifications

I am not judging my system just based on the results of each and every short session trial I get to play (which could vary from 150 spins to 300 spins on a visit, and are spread throughout the weekdays and hours). I am also -and more importantly- judging the system by its performance of the combined dataset when gluing together all my session trials, that is a dataset that becomes bigger and bigger and shows of greater volatility and stubborn patterns off randomness. Thus, there are disrupted data and not 24/7 continuous, unfortunately. I wish I had thousands of continuous data to test my system again. I will try to record one or two days continuous spins soon enough for the shake of a better dataset. What makes me a bit more confident despite the disruptive data, is something I read in this forum, that there is no real difference of very long continuous sessions versus many short disrupted sessions; I hope that is correct. Do you agree with this statement, I much value your opinion.

On the note above, I am happy to see that once this dataset gets bigger and bigger, many systems in place fall apart and depict worsening risk/return metrics. A few ones though survive and depict stability. And I am focusing on understanding and fine-tuning them.

I do not find the fluctuations of my cumulative profit-and-loss very volatile to be honest and this is why I have hopes on my system. Some drawdowns and some volatility, always with a positive performance in the long and short term, are signs of a nice behavious under a very volatile environment and market, like the roulette. A system with very low volatility and kind of fixed winning patterns etc etc could be a huge caveat for model fitting and cherry picking when building the system on a specific dataset, a very common error in quantitative modelling. I am trying to overcome this by building a system, an algorithm whose logic and parameters are not derived and based on my dataset. The dataset is just there to backtest, to test the validity of the original thought, the algorithm.

On another note, I do not bet VdZ, Tiers or Orphelins etc the classic way used at a roulette table. I do not believe that the risk/return payout ratio is worth it and I agree that in the long term the house will always win with at least 2.70%. I am trying to say that the VdZ bet for me is not a 17 number bet, it is a dynamical size bet based on my signal. Same for Tiers. System point here, as implied in my initial post here, I choose to Ignore the Orphelins on my system. This feature of my system, should affect the DTOP you mention, correct?

Thanks again

« Last Edit: May 20, 2019, 12:22:41 PM by fiben7 »

#### dobbelsteen

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #37 on: May 20, 2019, 08:05:22 PM »
My explanation is only based on my long computer research and my theory about systems strategies and bet selections. I have no idea of your methods. Gathering a lot of data is very important to broad your knowledge about the random roulette sequences.

Do you have watched some of my videos on YouTube?

The following users thanked this post: fiben7

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #38 on: May 20, 2019, 08:21:48 PM »
I will watch your videos dobbelsteen and work on them, thank you!

#### scepticus

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #39 on: May 21, 2019, 10:59:16 AM »
Fiben7

Glad to see that you think that profits could be made from betting roulette.

Dealers’ Signature ?   Ignore it. It is just “ noise “ .

RNG bias ?   Those who think that should simply avoid playing there .

Fixed Bets  ? Agreed  ! Much easier for Money Management .
Prediction Paramount ?  Agreed ! If  the winning number is not among your chosen numbers no  Progression can  make you win .

Strictly follow the rules ? Agreed  ? Why have rules if you don’t follow them ?

Odds. Unlike the financial markets the odds remain fixed.

The Zero ?  Ignore. It is only one among 37 numbers and is the least  useful .

Max Drawdown . Agreed . What we call the Table Bank.

Challenge ?  You bet  !

WARNING !  The big differences  between roulette and the markets  are ( 1 ) You don’t lose ALL of  the individual bet in the markets - you do in roulette.  ( 2 )  In the markets you bet against other opinions . Not so in roulette - you bet against  The House ..

Welcome to the chase !

The following users thanked this post: fiben7

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #40 on: May 21, 2019, 01:04:07 PM »
Dear Scepticus,

Thank you very much for your welcoming here and your feedback. I 've been reading your posts since joining and really appreciate your perception.

Happy to see that we agree on the basic concepts!

Your remarks on the differences between markets and the roulette are well noted.

Point 2) is the most difficult one to beat by definition and structure!

Point 1) please allow me to mention that in financial markets, the closer analogy to the roulette betting is the options and derivatives products where you do lose all your bet/investment (premium) if not correct in your prediction whereas in certain such products, your profit is again well defined and fixed. Thus, one could check such derivatives trading models should he like to find something working in the roulette as an analogy. Even though the maths there are way difficult and need calculations that one cannot feasibly make on his head within 60 secs of each spin. Thus, mostly unapplicable.

In any case though, the well known, defined and easy to understand rules of the roulette, the bets and the returns, along with "limited" 37 numbers space, makes the challenge even more tempting!

Best regards

#### scepticus

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #41 on: May 21, 2019, 07:33:22 PM »
Fiben7

Point taken but don’t experienced “ marketeers “ hedge their bets  on options ? I am thinking here of Ed.  Thorp  though I was greatly surprised that on one bet he actually borrowed money to increase his profits. I don’t think you should  “bet  the ranch “ on ANY  bet - no matter how confident you are.

Anyway, back to your stats. Idea.  Will you use one method or more than one. Trends or Chops ? Both but at different times ? What odds are you aiming for ? How long “  at the table “?  Just a steer as to how you propose to approach the problem .  I am interested in any NEW approach .

The following users thanked this post: fiben7

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #42 on: May 21, 2019, 08:50:36 PM »
dear scepticus,

well some derivatives trading strategies are partially hedged by other positions on the underlying asset or other products yes; something like betting on the Tiers and also covering 1,6,20,34 from Orphelins

On your questions now.

-Will you use one method or more than one.
-I have 4 systems overperforming others, of which 1 system is up to now significantly better than the rest. I have made the quantitative analysis as to what would have happened using combinations of my numerous systems and there are a few couples of systems that would be slightly better as a "portfolio" than my best system itself; though I like to stick for now for gaining experience and aptitude on my best one solely. in finance terms, the best combo has a sharpe ratio of around 0.95 whereas my best system has a sharpe ratio of 0.83 itself.

-Trends or Chops ?
-Both Mean reversion and Trend patterns are evident for me on the roulette depending on the type of bet.

-Both but at different times ?
-Nice point, I believe intersecting different patterns (trend/mean reversion) of different type of bets at the same time on a signle system is the fact that provides an edge as you minimize the chips used to bet and maximize the payout if correct.

-What odds are you aiming for ?
-I am betting on straight numbers mostly, thus aiming at a 36:1 chips gross payout. My average bet is 9 chips. My average hit ratio is 28%. And I only flat-bet.

-How long “  at the table “?
-Single sessions to play and record data last from 3 to 5 hours, i.e. from 120 to 250 spins most of the times. Then I analyze the sessions performance but most importantly the cumulative performance of the system at the whole overall dataset gathered till then.

Hope that helps, I have more info in this thread regarding some other aspects of my system as well.

Best
« Last Edit: May 21, 2019, 09:21:19 PM by fiben7 »

#### scepticus

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #43 on: May 21, 2019, 10:03:54 PM »
Thanks fiben7.
I look forward to further info on your " challenge " !
Regards

The following users thanked this post: fiben7

#### fiben7

##### Re: Basics Of A Quantitative Roulette Strategy
« Reply #44 on: May 22, 2019, 01:05:49 PM »
Forgot to say that my basic system bets on each and every spin.

A variation of it, in an effort to optimise the risk/return ratio, has a rule to stop and wait. For example, after 5 continuous losses, you wait and virtually play and only after a virtual gain, then you jump into real betting as well.

Another variation could be that given the hit ratio of the system, say 28%, if within a session the hit ratio is temporarily at 60%, then you could again stop betting - the "hit and run" as discussed in the forum - and enter again real betting once the hit ration mean reverts back to around 30%.

However, the variations above are not 100% trusted by myself for now, as I feel that I need many more spins (say 50,000) in order to verify the validity of the levels (the 5 losses or 60% hit ratio). The current optimal levels could well be overfitting effect of the current limited database.