Statistically testing every indicator with my Genetic Algoritm | Page 6 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

codification of the market / Genetic Algorithms 36 replies

S/R Area Algoritm 17 replies

What Constitutes a Statistically Significant Sample for Back Testing? 4 replies

Has anyone had luck with high frequency using either neural nets or genetic Pr? 3 replies

Genetic Algorithm for Metatrader 1 reply

Attachments: Statistically testing every indicator with my Genetic Algoritm
Exit Attachments

Statistically testing every indicator with my Genetic Algoritm

  • #101
  • Jun 30, 2014 7:00am Jun 30, 2014 7:00am
Quoting FxMasterGuru
Any thoughts about finding the best parameters for the best performing indicators for the past 3 months, then repeating it monthly (for the past 3 months) and always trading with the actual "optimized", i.e. best performing settings...?
I don think that the 3 month has any significance.The optimization lenght must also be a variable, and must find the best period for each cluster upon which to optimize the indicator.

And of course in each period there is different "best settings" there is no universal "best setting" for like RSI or stuff, you must find the best one in each cluster and trade it until the cluster finishes.Then repeat.This is the core of my approach.

Its head hurting but thats how objective analysis works
"There's a sucker born every minute" - P.T. Barnum
  • #102
  • Jun 30, 2014 9:42am Jun 30, 2014 9:42am
Quoting Proximus
{quote} I don think that the 3 month has any significance.The optimization lenght must also be a variable, and must find the best period for each cluster upon which to optimize the indicator. And of course in each period there is different "best settings" there is no universal "best setting" for like RSI or stuff, you must find the best one in each cluster and trade it until the cluster finishes.Then repeat.This is the core of my approach. Its head hurting but thats how objective analysis works
Thanks! Would you have optimized parameters for the current market climate for the Top3 indicators?
  • #103
  • Jun 30, 2014 8:54pm Jun 30, 2014 8:54pm
Attached Image


Ok here are the results i finished testing all indicators on H4, so here are they.Surprizingly all of them were profitable , but i only show the top 4 because i`m lazy to tell about them in detail.

Top 4 on EUR/USD
1)Wiliam's
2)Demarker
3) 1 line Simple Smoothed crossing price
4) 1 line EMA crossing price
Top 4 on USD/JPY
1)Demarker
2) 1 line Simple Smoothed crossing price
3) 1 line EMA crossing price
4)Wiliam's
Top 4 on AUD/USD
1) 1 line EMA crossing price
2) 1 line Simple Smoothed crossing price
3)Wiliam's
4)Demarker
"There's a sucker born every minute" - P.T. Barnum
  • #104
  • Jul 1, 2014 2:48am Jul 1, 2014 2:48am
Quoting Proximus
{image} Ok here are the results i finished testing all indicators on H4, so here are they.Surprizingly all of them were profitable , but i only show the top 4 because i`m lazy to tell about them in detail. Top 4 on EUR/USD 1)Wiliam's 2)Demarker 3) 1 line Simple Smoothed crossing price 4) 1 line EMA crossing price Top 4 on USD/JPY 1)Demarker 2) 1 line Simple Smoothed crossing price 3) 1 line EMA crossing price 4)Wiliam's Top 4 on AUD/USD 1) 1 line EMA crossing price 2) 1 line Simple Smoothed crossing price 3)Wiliam's 4)Demarker
Fantastic job! Just looking at the results an idea has popped in:

What if one waits for the lining up all 4 (or 3 of the 4) indicators' signaling an entry in the same direction? It might be the 99% probability entry. Taking exits might happen when any of the 4 (or 3) signaled that the show is over...

I am aware that such an experiment is beyond the computational capacity, but a manual forward test might be an option for some.
  • #105
  • Jul 1, 2014 7:43am Jul 1, 2014 7:43am
Quoting FxMasterGuru
{quote} Fantastic job! Just looking at the results an idea has popped in: What if one waits for the lining up all 4 (or 3 of the 4) indicators' signaling an entry in the same direction? It might be the 99% probability entry. Taking exits might happen when any of the 4 (or 3) signaled that the show is over... I am aware that such an experiment is beyond the computational capacity, but a manual forward test might be an option for some.
Not that easy

If you wait for them to align, that would be a very rare case, since both signals must come at the same bar.And if so then you will have very few trades, maybe 1 in a month or so, even if the accuracy is that much, its a waste of potential, better to have some losers but with compounding the position size you can earn more money that way.

But if you want to do it fine.However here is a bit of logical correction, the logical opposite of AND is the OR, so you have to take the entry when INDICATOR 1 AND INDICATOR 2 perfectly align, and exit when INDICATOR 1 OR INDICATOR 2 signals the opposite.
"There's a sucker born every minute" - P.T. Barnum
  • #106
  • Jul 1, 2014 8:29am Jul 1, 2014 8:29am
Yes, you are correct.

As an alternative one may want wait for at least 2 of the 4 best indicators to enter into the "green zone" and trigger a trade with the 3rd one (a 4th may come or not as a confirmation); the exit might be when only 2 green "lights" left.

If one could achieve a 90% probability setup with a good R:R ratio, it may be worthwile to wait for that one trade per month.
  • #107
  • Edited 8:36pm Jul 1, 2014 8:24pm | Edited 8:36pm
Thanks for this Proximus. Any chance in showing us how the RSI performed? It seems to be one of the ones which many people, including myself use. Howver, the DeMarker indicator appears to be calculated the same as the RSI but with SMAs rather than the EMAs of the RSI.

Also, if they are all profitable, it begs the question why do so many people lose out in trading? It really sounds like it's more of a psychological game.
  • #108
  • Jul 1, 2014 10:31pm Jul 1, 2014 10:31pm
Quoting Macdon
Thanks for this Proximus. Any chance in showing us how the RSI performed? It seems to be one of the ones which many people, including myself use. Howver, the DeMarker indicator appears to be calculated the same as the RSI but with SMAs rather than the EMAs of the RSI. Also, if they are all profitable, it begs the question why do so many people lose out in trading? It really sounds like it's more of a psychological game.
Well not quite, yes all of them were profitable but the majority had like 100-1000 net pips after a 10 year trading period.Which is pretty lame, and no sane person would stick to that indicator which performs this bad.

Of course its the game of changing the system every month or so after it loses too much, but if you find a decent one then there is no point in changing it, yet some of us do.It is a psychological game, but it also needs good tools to begin with, so i think both the edge and the psycho is equally important, you cannot trade profitable if one of them is missing.

I will test the RSI soon but atm i`m busy doing something else.
"There's a sucker born every minute" - P.T. Barnum
  • #109
  • Jul 1, 2014 11:48pm Jul 1, 2014 11:48pm
Quoting Proximus
Well not quite, yes all of them were profitable but the majority had like 100-1000 net pips after a 10 year trading period. Which is pretty lame, and no sane person would stick to that indicator which performs this bad.
Yes, this is true, however it still is better than them not being profitable. Perhaps these low performing indicators would perform very well when combined with some other indicator? From your results, 1SMMA SMOOTH or 1EMA comes to mind.
Quoting Proximus
Of course it’s the game of changing the system every month or so after it loses too much, but if you find a decent one then there is no point in changing it, yet some of us do. It is a psychological game, but it also needs good tools to begin with, so i think both the edge and the psycho is equally important, you cannot trade profitable if one of them is missing.
That’s very true even of the world’s greatest system.
Quoting Proximus
I will test the RSI soon but I am busy doing something else.
Thanks for looking into it Proximus. I find it very intriguing.
  • #110
  • Jul 2, 2014 12:16am Jul 2, 2014 12:16am
Quoting Proximus
{quote} Well not quite, yes all of them were profitable but the majority had like 100-1000 net pips after a 10 year trading period.Which is pretty lame, and no sane person would stick to that indicator which performs this bad. Of course its the game of changing the system every month or so after it loses too much, but if you find a decent one then there is no point in changing it, yet some of us do.It is a psychological game, but it also needs good tools to begin with, so i think both the edge and the psycho is equally important, you cannot trade...
You know Proxy.....

At first I smirked at this thread. Trying to back test for a profitable indicator and all. Back testing is worse than useless, and any indicator put in the hands of a newbie/bad trader will lose money.

But what you are actually showing is that no mechanical system can ever work over the long term, and that every system, EA, or what have you has to have a human component. The RSI and the Stochastics have proven for many many years to be one of the most rubust indicators out there. And yet, fixed parameters will never work over the long term. You cannot just trade like a blind monkey. Even HFT algos have humans watching over them.

Your thread has actually proved invaluable for people who know what they are looking at.
If you are good at something, never do it for free--Joker
  • #111
  • Jul 2, 2014 12:49am Jul 2, 2014 12:49am
Quoting jmn5611
{quote} You know Proxy..... At first I smirked at this thread. Trying to back test for a profitable indicator and all. Back testing is worse than useless, and any indicator put in the hands of a newbie/bad trader will lose money. But what you are actually showing is that no mechanical system can ever work over the long term, and that every system, EA, or what have you has to have a human component. The RSI and the Stochastics have proven for many many years to be one of the most rubust indicators out there. And yet, fixed parameters will never work...
Maybe, the human component is important however you can enhance a little bit your odds if you do an analysis like this: when to trade, which pairs, with which indicator, what time of the day, etc

Answering correctly to these questions can enhance your odds significantly, however human discipline will still be needed.
"There's a sucker born every minute" - P.T. Barnum
  • #112
  • Edited 6:26am Jul 2, 2014 6:14am | Edited 6:26am
Quoting Proximus
{quote} Maybe, the human component is important however you can enhance a little bit your odds if you do an analysis like this: when to trade, which pairs, with which indicator, what time of the day, etc Answering correctly to these questions can enhance your odds significantly, however human discipline will still be needed.
"...when to trade, which pairs, with which indicator, what time of the day, etc" - Very well summarized! That is the art of trading.

I wish an "algo" could that... But then, it would be the "holy grail of trading" which - as they say - does not exits...

So personally I stick with a high-tech newstrading strategy with an IT-grounded team. I know, it is not a "real strategy" as it does not need indicators or special market knowledge, just some very fast VPS's and a couple of fancy software to do the job. But hey, whatever works...

62% over the past two months is not bad... (http://www.myfxbook.com/members/Para...9KjKMWXQFVDKMg) although it requires regular account rotation when the "parasite" plug-ins are placed on the accounts... http://www.forexfactory.com/showthread.php?t=70582 This account has just just got the "plug-ins", so the funds will be moved soon...
  • #113
  • Edited 6:59am Jul 2, 2014 6:53am | Edited 6:59am
Quoting FxMasterGuru
{quote} "...when to trade, which pairs, with which indicator, what time of the day, etc" - Very well summarized! That is the art of trading. I wish an "algo" could that... But then, it would be the "holy grail of trading" which - as they say - does not exits... So personally I stick with a high-tech newstrading strategy with an IT-grounded team. I know, it is not a "real strategy" as it does not need indicators or special market knowledge, just some very fast VPS's and a couple of fancy software to do the job. But hey, whatever works... 62% over...
Hehe news trading nice, i`ve also developed a news trading strategy which has atleast 60% winrate with a 1:1 RR however i cannot use it, because slippage and broker freeze is too high around those periods.Maybe when i`ll get to the Currenex level i will use it, until then i keep it in my pocket

I`m curious how do you combat this on MT4 ? Or does your broker have lightning fast execution with no requote, slippage, spread widening around news time ?
"There's a sucker born every minute" - P.T. Barnum
  • #114
  • Jul 2, 2014 7:00am Jul 2, 2014 7:00am
Quoting Proximus
{quote} Hehe news trading nice, i`ve also developed a news trading strategy which has atleast 60% winrate with a 1:1 RR however i cannot use it, because slippage and broker freeze is too high around those periods.Maybe when i`ll get to the Currenex level i will use it, until then i keep it in my pocket I`m curious how do you combat this on MT4 ? Or does your broker have lightning fast execution with no requote, slippage, spread widening around news time ?
I will respond in PM. It is not very public.
  • #115
  • Jul 7, 2014 7:28am Jul 7, 2014 7:28am
Quoting nonor
Ok, thank you. It's a fact that RSI is well used in many manual and automated systems and your results goes against the common belief. is it possible to test the two following variation of RSI ? Thank you {file} {file}
Sorry for late response, i didnt forgot your request, but i was too busy to test it

So after a few weeks passes i tested your 2 indicators, however i no longer use the entire 14 year EUR/USD database only after 2010, since after the flash crash of 2010 the market was changed drastically, the volatility had dropped since than by about 10% each year.This year almost 30% lower than in 2013, so i dont think anything in 2004 to say its relevant today.

So because of this it cant be compared to the other RSI tests, however atleast this gives a perspective how well it performed in the january 2010 - 1st of june 2014 interval

Rapid RSI:

Profit: +44092
Trades: 32
Period: average profitable period above 60
Winrate: 46.12%
Timeframe: H1 (to be more accurate i dont use M30 anymore)

Figurelli RSI:


Profit: +77142
Trades: 1252
Period: average profitable period above 70 and the gain ~ 26
Winrate: 56.891%
Timeframe: H1 (to be more accurate i dont use M30 anymore)

Obviously both were tested with the 30/70 level strategy, so it seems like the Figurelly beats all RSI's so far.Good choice bro, i will also investigate it further as it looks like a promising indicator
"There's a sucker born every minute" - P.T. Barnum
  • #116
  • Jul 7, 2014 6:10pm Jul 7, 2014 6:10pm
Proximus,

I donīt know who you are. I donīt know for sure if this work has mistakes. Actually I canīt say anything about your research once I have just started to study it. But even now I can say for sure it has a lot of interesting and important information. And it must have taken a lot of work (yours and from your PC). So now I just want to say: thank you very much for sharing this beautiful work. Also, thank you for all the information and explanations.
  • #117
  • Jul 7, 2014 6:33pm Jul 7, 2014 6:33pm
Good Job Gekko, this thread will be referenced for years to come. The information in it is just gold, be well and may the pip be with you.
Being a trader is lonely, but being a great trader is lonelier still
  • #118
  • Jul 8, 2014 2:15am Jul 8, 2014 2:15am
Once again, thanks very much for all this Proximus. It's a real eye opener.

There's only 1 problem at my end, I've looked up the Rapid RSI and so know what that is now (basically RSI without smoothing), but for the life of me, I don't know what the Figurelli RSI is and there's not much information on the internet. Is that Rogerio Figurelli's propriety indicator? They talk about a "gain" variable as 1 of the parameters, but i have no idea what it is.

Can someone please post the formula or a link with meaningful information? Thanks.
  • #119
  • Jul 8, 2014 4:11am Jul 8, 2014 4:11am
Quoting Reis
Proximus, I donīt know who you are. I donīt know for sure if this work has mistakes. Actually I canīt say anything about your research once I have just started to study it. But even now I can say for sure it has a lot of interesting and important information. And it must have taken a lot of work (yours and from your PC). So now I just want to say: thank you very much for sharing this beautiful work. Also, thank you for all the information and explanations.
Not that much work, i have a very fast PC, besides i was also curious about their performance

Quoting Macdon
Once again, thanks very much for all this Proximus. It's a real eye opener. There's only 1 problem at my end, I've looked up the Rapid RSI and so know what that is now (basically RSI without smoothing), but for the life of me, I don't know what the Figurelli RSI is and there's not much information on the internet. Is that Rogerio Figurelli's propriety indicator? They talk about a "gain" variable as 1 of the parameters, but i have no idea what it is. Can someone please post the formula or a link with meaningful information? Thanks.
This is the only stuff that i found about him

http://championship.mql5.com/2012/en/news/167

If you need that formula look inside the code, it may seem messy but i dont think there is a mathematical formula which describes it, its all in MQL4 code.
"There's a sucker born every minute" - P.T. Barnum
  • #120
  • Jul 8, 2014 6:22am Jul 8, 2014 6:22am
Quoting Proximus
This is the only stuff that i found about him http://championship.mql5.com/2012/en/news/167 If you need that formula look inside the code, it may seem messy but i dont think there is a mathematical formula which describes it, its all in MQL4 code.
Thank you once again Proximus. That's just about all I found on it too. I got the code from nonor's post on this thread http://www.forexfactory.com/showthre...23#post7568523 I'll have to read it all and digest it.

Are you going to do any more tests? These 2 RSIs look very good - lots of PIPS and high strike rates of 46% and 57%.

Since you did the original tests with 12 years data on the M30 charts, would it be worthwhile to test the indicators you originally tested on the data from 1st Jan 2010 on 1HR charts? Just a suggestion if you're not too busy.

Cheers and thanks again.
0 traders viewing now