Can statistical methods be used to create a trading strategy? | Page 3 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Can we interpret fundamentals into statistical data? 3 replies

idea for a statistical strategy 33 replies

Where can I find these statistical indicators? 1 reply

A simple way to create your own trading strategy 2 replies

How do do you create your trading strategy? 20 replies

Attachments: Can statistical methods be used to create a trading strategy?
Exit Attachments

Can statistical methods be used to create a trading strategy?

  • #41
  • Jun 16, 2014 8:40am Jun 16, 2014 8:40am
Quoting Fractalist
Why not just use chi-squared? That isn't unbiased at all. It assumes a normal distribution (or one of its cousins) in the data. Market data is not normally distributed; the distribution has been shown to be more like the Pareto-Levy distribution
MCRotter was talking about the return of the strategy not of the market. The returns of the strategies often follow a normal distribution with a means near zero, positive if the system is profitable. Or it may be a bimodal mixture of gaussian with a peak at the average loss and another peak at the average profit. If you don't like the X2 you can try the Kullback–Leibler divergence to check the null hypothesis of distributions similarity.
No greed. No fear. Just maths.
  • #42
  • Jun 17, 2014 5:00pm Jun 17, 2014 5:00pm
Hell yes, of course !! Watch this video and see how Derren Brown is using statistics to predict horse races

Inserted Video
Simplicity is the key to brilliance.
  • #43
  • Jun 17, 2014 7:26pm Jun 17, 2014 7:26pm
Quoting Inceptionist
Hell yes, of course !! Watch this video and see how Derren Brown is using statistics to predict horse races http://www.youtube.com/watch?v=9R5OWh7luL4
Maybe the long term winners in fx are not really consistent winners, but rather only on a long winning streak which may have lasted decades for them, but if they were to trade over 100-200 years, they could lose it all after that.
It is fascinating nontheless, because the market seems to be very random-like, and the fact that so little % of retailers win, makes you doubt that the edge is it even real or its just a lucky winning streak in a zero sum random game?

What do you think guys?
"There's a sucker born every minute" - P.T. Barnum
  • #44
  • Jun 18, 2014 2:54am Jun 18, 2014 2:54am
Quoting Proximus
{quote} Maybe the long term winners in fx are not really consistent winners, but rather only on a long winning streak which may have lasted decades for them, but if they were to trade over 100-200 years, they could lose it all after that. It is fascinating nontheless, because the market seems to be very random-like, and the fact that so little % of retailers win, makes you doubt that the edge is it even real or its just a lucky winning streak in a zero sum random game? What do you think guys?
This is Redlion's null hypothesis in his thread "why have you accepted the price-predictability assumption?"
No greed. No fear. Just maths.
  • #45
  • Jun 18, 2014 6:00am Jun 18, 2014 6:00am
Quoting PipMeUp
{quote} This is Redlion's null hypothesis in his thread "why have you accepted the price-predictability assumption?"
I wasnt refering to the price as it is predictable or not, i was refering to the traders who get to that level to be profitable are randomly "choosen".

Because either the price itself is predictable or not doesnt matter, if only a small percentage of traders get there, but by what means to they get there.

Because manual trading is so subjective, there are some core principles that are objective, but still its very subjective, and not every trader gets to be a billionaire.

It can be lack of information, discipline, or knowledge of the markets, but even with these 3 already mastered, you can still be a long term loser.

So the priviledge to be a billionaire from trading, is pretty much random, regardless of the predictability of the markets, it may sound silly but it is true.

It would be like if you knew with a 100% certainty the next weeks lottery numbers, and e-mail this info to 10.000 players, and lets assume that all of them would believe you.There would be people who would fill the wrong numbers, who lose their ticket, who forget to buy a ticket, and so on.

As if in the markets, if everyone had the total knowledge about the price, there would still be people who would f**k it up.
"There's a sucker born every minute" - P.T. Barnum
  • #46
  • Jun 19, 2014 2:49pm Jun 19, 2014 2:49pm
Quoting Proximus
{quote} Strange, my research shows that the EMA, LWMA , SSMA is very ineficcient at predicting price. From the classical moving averages, the simple moving average is the most efficient, although not efficient at all, probably your frequency filters are waaay better, but compared to the others its way more efficient. The problem with EMA is that it puts exponentially more emphasis on certain bars, which implies that after a downtrend to say, an uptrend has exponentially increasingly more and more chance of happening, so that the individual outcomes...
Let me throw you a bone guys, i redid the experiment, turns out that i was right.

And the worst of them all is the SSMA, which had about -25% less profit than every other, in each and every test.I measured the lagg resulting from a suddent breakout until it adapts again to the price, and the general adapting capability, and of course how much profit it made.
It was not a MA cross type strategy, the entry was based on other criteria.

I did 62 tests, all on EUR/USD so it may have different results on other pairs, and the sample size was almost 100.000 in each test.And unanimously in every test the SSMA performed the worst

SSMA is the 3rd moving average, the Simple Smoothed.
"There's a sucker born every minute" - P.T. Barnum
  • #47
  • Jun 19, 2014 3:53pm Jun 19, 2014 3:53pm
Sorry but I don't understand your point at all. It is all nonsense to me. The various MAs are low pass filters. The purpose of a lowpass filter is to remove the frequencies higher than the cutoff frequency in an input signal. A low pass filter is not supposed to forecast anything. What is the point of misusing something then blaming it for not working in an area it has never been designed for? It's like saying a screwdriver is the worst vehicle to fly!
No greed. No fear. Just maths.
  • #48
  • Jun 19, 2014 4:47pm Jun 19, 2014 4:47pm
Quoting PipMeUp
Sorry but I don't understand your point at all. It is all nonsense to me. The various MAs are low pass filters. The purpose of a lowpass filter is to remove the frequencies higher than the cutoff frequency in an input signal. A low pass filter is not supposed to forecast anything. What is the point of misusing something then blaming it for not working in an area it has never been designed for? It's like saying a screwdriver is the worst vehicle to fly!
I certainly dont understand why you need to filter the price if you are not going to trade from it in this sense, it is like removing noise from a song that just been recorded in a studio, to never release the song to the public after.

Anyway i dont use the MA as a filter (because i dont understand why you need to filter the price like that, i believe the price is ok as it is), i use it as a price -predictor, with mean reversion tehnique, but not in the classical manner (buy above MA sell below MA), rather in conjunction with volatility analysis and only in the direction of the master trend with proper RR settings .

And 99% of traders will use it in conjunction with the mean reversion principle.

However there are atleast 3 strategies on this forum which use Simple Smoother Moving Average, which makes me sad, because if you even want to use an MA use the other 3 but not this one, because its the worst of them.

I`m not saying that EMA,SMA and LWMA is the holy grail, but it had considerable better performance, although the LWMA is the 2nd worse, yet made 25% more profit than the SSMA.

And of course my last point is that, not the tool makes the master, but the master makes the tool.The fact that a good plummer needs a hammer to fix a pipe doesnt mean that anyone with a hammer can fix a pipe.

And what is the difference between a hammer with no skilled user behind it and a good plummer with a bad hammer is this:

Attached Image (click to enlarge)
Click to Enlarge Name: noob.gif Size: 10 KB


1000$ flushed down the toilet in less than a year with conventional MA mean reversion tactics.

Attached Image (click to enlarge)
Click to Enlarge Name: pro.gif Size: 11 KB


Made 980.000$ from 1000$ in the period of 2013 january- 2014 june with my volatility analysis mixed with proper MA trading

Which one would you choose ?

Believe it or not, i would not choose either, because i have way better tools.

The tests were made both on EUR/USD between 2013 january- 2014 june, M15, initial balance 1000$, with 2% risk/trade in both cases, with simple moving average period 50. But as you can see not the moving average makes it profitable, but the trader
"There's a sucker born every minute" - P.T. Barnum
  • #49
  • Jun 19, 2014 6:11pm Jun 19, 2014 6:11pm
"Can statistical methods be used to create a (profitable) trading strategy?"

So basically the answer to this question is yes, statistical methods can be used to create a profitable trading strategy, i`m sure it was meant to ask for profitable strategies not otherwise.

In fact i think it is the only way to succeed as a retailer.If you are a retailer you must be a quant to succeed in long term, or a lucky trader go get enough money quick to then only hedge on long term position trades and make 2-3%/ month on a 1-2 million dollar account with small risk, before you lose all your money to try to become rich.

It is just like in the movie Margin Call

Inserted Video


"There are 3 ways to make a living in this business, be first, be smarter or cheat"

Now obviously we wont cheat: arbitrage and stuff like that is very immoral.

We can only be first or smart.

Well we cant be first, because banks and brokers front-run us , and also there are HIGH FREQUENCY TRADERS. So we cant put our orders first.

The only option left is to be smart.

Now attention by smart i dont mean, highly educated PHD in economics and econometrics and statistics.I only mean to be innovative, cunning and know-what-you-are-doing type of trader.

So either be a quant, and analyze the market all day like me, or be a fundamental trader to trade news and stuff like that.There are no nuances, if you are not innovative and not invent new trading methods, other than the defaults which newbies use for decades & lose money with it, you probably not going to retire from forex income, there are little exceptions, but dont expect the unexpected

Thats how it is!
"There's a sucker born every minute" - P.T. Barnum
  • #50
  • Jun 19, 2014 6:51pm Jun 19, 2014 6:51pm
Quoting Proximus
So either be a quant, and analyze the market all day like me, or be a fundamental trader to trade news and stuff like that.
I'm all for analyzing the markets and finding an edge. But one must not fool oneself. Do you really believe that your second backtest from above (1000 -> 1 mil) was properly done and is relevant? If you do, congratulations. Next year in December you'll be a millionaire.

If anyone wants to make money in this market, it should approach it in a professional way. One can fool themselves with MT4 backtests, but the market will show them how much those backtests are truly worth.

And this is also relevant for those who do eyeball backtests...
  • #51
  • Jun 19, 2014 7:26pm Jun 19, 2014 7:26pm
Quoting Adal
{quote} I'm all for analyzing the markets and finding an edge. But one must not fool oneself. Do you really believe that your second backtest from above (1000 -> 1 mil) was properly done and is relevant? If you do, congratulations. Next year in December you'll be a millionaire.
It was probably done correctly, although it may seem too good to be true.The historical data is very high quality tick data from 2003 until 2014.

The spread was set to the daily average spread (it may cause some problems as usually in news/breakouts the spread widens alot, and the tester doesnt simulate it).

Also there are slippages, freeze levels and gaps (hopefully no stophunts and no other bad stuff and cheating from the broker).

And of course the worst of all: taxes & comissions of all kind.

Other than that, it was done property, if the edge holds trough these obstacles then indeed my hard work came to a realization.

Yet i emphasize that this is not the system that i trade, i made a much better indicator than a classical moving average, i just showed that actually even with the worst indicators you can make money if the system is good.
"There's a sucker born every minute" - P.T. Barnum
  • #52
  • Edited Jun 20, 2014 2:10am Jun 19, 2014 7:40pm | Edited Jun 20, 2014 2:10am
However there are more ways to get-rich-quick.If one lacks completely the art of trading or any knowledge of it and just cant make his mind around it.

One can design a martingale gambling system which has the highest consecutive win rate and the lowest consecutive loss rate, and martingale himself up to a million, then put his money in a hedgefund and live from the 10-15% annual returns.

It would be a high risk gambling game, but if you statistically analyse the markets and know in which cases you achieve the highest consecutive win rate and the lowest consecutive loss rates, then it could theoretically work.

However not for everyone, the winners and losers would be random, yet if someone lacks the knowledge of trading but feels very lucky, and has some money to throw away why not?

It is silly, and gambling is a 100% losing game in long term, but if you are lucky you would get to 1.000.000$ quicker before losing it all

Just think about it
"There's a sucker born every minute" - P.T. Barnum
  • #53
  • Jun 26, 2014 1:54pm Jun 26, 2014 1:54pm
Quoting Fractalist
{quote} Glad to help. Lowpass Filters...
Apologies for the belated response, I've really appreciated your help with this. I'm still not certain on exactly how low pass filters work, and I've struggled to find any online sources that explain the calculation behind them. I've also struggled to view the script behind the file you attached, in MetaEditor. Is it protected, or am I doing something wrong?
I'm now about to start reading up on the Nyquist-Shannon Sampling theorem.

Thanks again for all the help.
  • #54
  • Jun 26, 2014 3:48pm Jun 26, 2014 3:48pm
Quoting Ponzi Jr
I'm still not certain on exactly how low pass filters work, and I've struggled to find any online sources that explain the calculation behind them.
http://www.ece.rutgers.edu/~orfanidi/intro2sp/
A 795 page book free to download about DSP:nerd:
No greed. No fear. Just maths.
  • #55
  • Jun 27, 2014 5:41am Jun 27, 2014 5:41am
Quoting PipMeUp
{quote} http://www.ece.rutgers.edu/~orfanidi/intro2sp/ A 795 page book free to download about DSP:nerd:
Excellent, thank you!
  • #56
  • Jul 16, 2015 8:03am Jul 16, 2015 8:03am
I have an another idea about this time " historical data analysis " with aiming a powerful software like IBM SPSS or even with IBM statistics modeler !!

the method ,exactly, is provide closest chart extractions of previous historic data to separation (from A to Z) charts for trader better decisions

that's it !!!

I can explain more in further but want a professional software user for design it , at first

somebody can help it ? it's very attractive big project and can start to another thread

promise you can convert to a profitable topic ever

waiting you !
  • #57
  • Mar 21, 2020 6:21am Mar 21, 2020 6:21am
I use statistics for trade signal generation and risk management regularly. It does help on average. But statistical analysis has to be injected into your trading process in a robust, "minimax" fashion. Simple methods like charting or linear regression are unlikely to work consistently (unless applied to some remote segments of the markets, like specific emerging markets stocks which few people can/want to trade). On the other hand, robust methods from advanced statistics, including some of machine learning, should not be frowned upon. Their complexity does not mean self-indulgence (selection bias) if the methods offer stable, conservative solutions. Much of statistics in the last 30-40 years has dealt with two related questions:

1) How does one apply regularization to the solution?
2) How does one pick only a small subset of not completely "silly" predictors out a universe of mostly "silly" predictors.

Hence we got methods like lasso, elastic nets but there are many more. I stress: I am talking about robust methods, methods which do not go completely insane under adverse conditions but deliver miracles in a few very specific cases. So I am not advocating deep learning and such. All in all, you have to gain a coherent statistics & stochastic processes knowledge before attempting statistical arbitrage. Do not just execute somebody's post "try this" and then, after failing, do not just execute somebody else's post "try that". You may not be able to "try" as well as the person who posted.

The following probability & statistics study help (the bottom of the page) delineates a study program in terms of specific chapters / sections in a relatively few books. You will skip some of the specialized master's level references but do power through much of machine learning. In summary, you do not have to learn all of the statistics & probability but, well, a lot. Useful methods: cluster analysis, hidden Markov models, PCA, least angle regression, non-parametric association measures, ...
  • #58
  • Mar 21, 2020 9:37am Mar 21, 2020 9:37am
In the end, no matter what kind of mathematical/analisis/statistic crap you apply to the market NOTHING will give you a positive result in the long run

Why??

very simple...

EVERY SINGLE CANDLE in a time based chart its of a random size, wich change the OHLC parameters every single time, you dont have constant variables so you can start analize them in the proper way

How can you find an "edge" when the variables are random every single calculation?? think about it

Almost no one notice that, they are too hard headed into the "overcomplicate everything because must find answer" mentality or just the right enough "dumb" to not see the what truly matters to watch in a trading chart
  • #59
  • Last Post: Mar 22, 2020 6:44am Mar 22, 2020 6:44am
Quoting staassis
I use statistics for trade signal generation and risk management regularly. It does help on average.......... In summary, you do not have to learn all of the statistics & probability but, well, a lot. Useful methods: cluster analysis, hidden Markov models, PCA, least angle regression, non-parametric association measures, ...
Another technique which is helpful in risk management is GARCH. Has to be re-estimated regularly. Often provides a more accurate estimate of risk than the volatility you think you see in the markets.
0 traders viewing now