Why exactly do EMA crossover systems not work? | Page 14 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Help. EMA Crossover with different EMA cross for exit 0 replies

EMA crossover (ema 14 & 34) 4 replies

Please help code an EMA Crossover 12ema x 24 ema cross 1 reply

Systems, Systems and Systems... 1 reply

EMA Crossover Systems 1 reply

Attachments: Why exactly do EMA crossover systems not work?
Exit Attachments

Why exactly do EMA crossover systems not work?

  • #261
  • Edited 12:48am Jul 25, 2011 12:33am | Edited 12:48am
Quoting nubcake
yep.

when stuff is just randomly mashed together with random values that all SEEM to point towards making a dollar, then it's a curve-fit. what else COULD IT BE!?

if, on the other hand, someone said "here is a system you can backtest... do this this and this etc" where all of the rules and parameters and so on actually had a REASON for being chosen then that would be of interest and something bloody revolutionary and beyond the norm of the forex idiocy.

as it stands it is all just random shit mashed together that someone somewhere stumbled...

This is a good point. Backtesting to me is kinda like throwing a bunch of random parameters on the wall just to see what sticks.

Scenario 1: I want to backtest and see what happens when I only place a short position on the Euro when the 50 DMA and the 200 DMA are 11.2% apart, when the RSI is 60.23, when the CCI is 40.009, when the MACD is 80.125, when the 10 minute 200 PMA is equal to the 10 minute 50 PMA, only on the 1st day of the month, after 3pm, "only" when the dollar index is down, and "only" when the Dow is up 8.2% or greater. Now backtest those parameters for the past 10 years and let's see what we get.

Now what percentage of serious traders really trade to all of these "random" parameters? Too many people have been led to believe that a FX chart behaves just like a function (i.e. Y=mX+b) that you can just program or curve fit. It is a complete fallacy that a FX chart can be explained by a mathematical formula....yet this is what people do when they set out to backtest.
  • #262
  • Jul 25, 2011 12:33am Jul 25, 2011 12:33am
Nub is right.....

Ultimately, when you find something that 'works', its been curve fitted for that particular period. Even if you check various pairs that may or may not be related , you will find the results will end up being similar for that test period


The only true edge is money management.

Nub mentioned the heads/tails arguement - its 50/50 chance. Yet, if you were told you'd get 2x for every one that is right, and lose your wager on every one you lost, in theory you should win after x number of throws and be ahead....

BUT...the most people will screw up the money management, and lose.


The answer to the above riddle is - consistent betting, and you MUST bet 25% of your account each time to have max gains. (search google for Kelly formula to understand why that is)


However....as we all know, as random as Forex is, it doesn't account for streaks. That is where the problem lies - you cannot bet 25% each time and expect to survive in Forex (or any market for that matter).


Unfortunately....for me anyway, this presents a conundrum.

Consistent betting vs pyramiding during streak periods...which is better?
  • #263
  • Jul 25, 2011 12:43am Jul 25, 2011 12:43am
The exact parameters are meaningless if the market is trending, 13/39 and 40 days will probably be as profitable as 11/27 and 32 days or 8/17 and 15 days.
I'd say any automated trading system works as long as the market is trading and you take all signals.

On the other hand, it is extremely easy to curve fit this kind of systems, regardless of how many pairs of how large the database is. 16 years of daily candles is close to nothing.
  • #264
  • Jul 25, 2011 12:45am Jul 25, 2011 12:45am
Trading strategies have (like everything in life) also a time to live. They start working, they get followed by retail traders and their edge stops working because too many of market players have detected this edge.
  • #265
  • Jul 25, 2011 12:53am Jul 25, 2011 12:53am
Quoting 4XTerrorist
Trading strategies have (like everything in life) also a time to live. They start working, they get followed by retail traders and their edge stops working because too many of market players have detected this edge.
That suggest that the edge was there to begin with.


All trading systems have one flaw - the user.
  • #266
  • Jul 25, 2011 12:56am Jul 25, 2011 12:56am
Quoting Sauron
16 years of daily candles is close to nothing.
My friend, 16 years and half of daily bars on 21 different markets are the equivalent of 83,160 bars! ( 198 months total x 20 trading days each month x 21 markets)

In statistical terms it is an extremely large and meaningful sample to base a serious study on.
  • #267
  • Jul 25, 2011 1:07am Jul 25, 2011 1:07am
Quoting nubcake
yep.

when stuff is just randomly mashed together with random values that all SEEM to point towards making a dollar, then it's a curve-fit. what else COULD IT BE!?

if, on the other hand, someone said "here is a system you can backtest... do this this and this etc" where all of the rules and parameters and so on actually had a REASON for being chosen then that would be of interest and something bloody revolutionary and beyond the norm of the forex idiocy.

as it stands it is all just random shit mashed together that someone somewhere stumbled...
One day, someone starred at some charts for a long long time and said : "Eureka, every freely-traded market has a trend and is not completely random!"

The next day he said "Wait a minute, if the market has a tendency to trend, then it could be possible to mathematically capture this trend with a couple of moving average, let's expirement with some values."

The day after he said "Yep, just what I thought, my moving average system do make money consistently if I trade enough markets"

At the same moment Nubcake looked at him and yelled : "No, I don't want to make money with your moving average system, I am a deep thinker, I want to understand everything, including the whole universe, I want to know WHY your 13 and 39 moving average work and I don't want to earn a single dime trading your system until I fully understand the true meaning of these numbers!!"

10 years later our trader has accumulated some much profit that he is now earning over $750.000 a year trading his simple moving average cross over system, buying and selling over 45 contracts per trade per market.

During the same period Nubcake did not earn a single dollar trading because he was still looking for some mystical explanation behind the numbers 13 and 39, on the top of the Himalayas, where, according to a few buddhist monks, he keeps repeating 13 and 39 all day, in the lotus position.
  • #268
  • Jul 25, 2011 1:09am Jul 25, 2011 1:09am
the market is not the enemy, the enemy is within ourselves. no one is there lurking trying to counter act your edge. price moves according to interest of the people that move it. Mainly the regulators of the financial systems are the central banks, and every one else always looks to profit. traders do not move the market, ask the FED whether they will let someone appreciate the USD without their permission. The US Government, banned the trade of gold, and silver, and released oil into the market in open market operations to sustain the dollar, and mask inflation.

fundamental factors are something to be watching, and technical factors should be evaluated to see if in fact the market corroborates the theories, the higher the time frame the more the prediction. understand the role of the USD in the market and correlation among the major currency pairs.

also understand how to do ongoing objective analysis, do you know the meaning of each candle? do you understand how to use a reference price point, and exploit the range? can you observe price behavior and the importance of momentum?

do you know what EMH and behavioral market theory? have you read fiscal policy? and monetary policy and the affect it has in the financial systems?

actually if you claim that EMA cross over works tell me what the difference is between, SMA AND EMA WMA and why they are useful?

MR Breakout lets start here, into a meaningful discussion of trading
AVT INVENIAM VIAM AVT FACIAM
  • #269
  • Jul 25, 2011 1:11am Jul 25, 2011 1:11am
Quoting Mr Breakout
My friend, 16 years and half of daily bars on 21 different markets are the equivalent of 83,160 bars! ( 198 months total x 20 trading days each month x 21 markets)

In statistical terms it is an extremely large and meaningful sample to base a serious study on.
ok, if we are going to mathify this....

there are 1,440 M1 bars in a day. 83,160 / 1,440 = 57.75.

therefore : 83,160 bars is similar to 57 days worth of M1 bars to the perspective of an intraday trader. rounding that up makes it basically 2 months of intraday data. doesn't seem like much data to me.

sorry
Congratulations ScalaFX for coming out of the closet!
  • #270
  • Jul 25, 2011 1:16am Jul 25, 2011 1:16am
Disclaimers:

10. "There is a high degree of risk involved in trading. XXXXX involve risks and are not suitable for all investors." A lot of people gets screwed, but we don't want to tell you. Not everyone is cut out for this game you dumb f*ck.

9. "You should obtain investment advice from a licensed investment adviser before making any investment decisions." Go ask someone else because we don't want to get sued for our mistakes.

8. "XXXXX are not licensed investment advisers and the information, services and products provided are not suitable to be acted upon as investment advice." We don't want to get a license because if we do, we still don't have a clue of what we are doing. And then we are really screwed by the authorities.

7. "XXXXX assume no responsibilities for your trading and investment results." We'll provide you the support and guidance along the way, but if you f*ck up (which there is a high chance that you will)... then you're on your own.

6. "Simulated trading examples have been used with the benefit of hindsight." We're just killing time until you hand us your money. It's a lovely powerpoint presentation though!

5. "XXXXX suggested on this XXXXX, which may change from time to time." We'll change it when it suit our needs, like when you try to sue us... so you will have no case of winning this battle.

4. "The past performance of any trade whether actual or hypothetical is not an indication of future results. Past results are not indicative of future returns." F*ck that!!!... If we got it all figured out, do you really think we need you?

3. "Information for XXXXX observations are obtained from sources believed to be reliable, but we do not warrant its completeness or accuracy, or warrant any results from the use of the information." We'll provide you with unreliable materials, but we'll justify it by proclaiming we believe in it.

2. "The indicators, strategies, columns, articles and all other features are for educational purposes only and should not be construed as investment advice." We have no solid proof that are educational materials are of any benefits and will do you any good. Although we provide you these materials, we don't want any responsibility when the sh*t hits the fan.

1. "XXXXX make absolutely no representation that profits or losses demonstrated in these examples is likely or achievable." We sell dreams, we do not sell tangible goods. Again, we don't want to be liable for the unreliable materials/presentations we'll provide to you.
  • #271
  • Jul 25, 2011 1:21am Jul 25, 2011 1:21am
Quoting MackS10
You can apply close, open, median, low, you name it. Heck change the colour too, you can even make rainbows. Unfortunately newbies will always find unique ways to fail while veterans do the opposite.

Cheers~
I know i can apply any type, even colors and also how thick and how thin too.

But, is that what I was asking? I was simply asking if there was any particular USE in applying open/close as opposed to high/low....
  • #272
  • Jul 25, 2011 1:22am Jul 25, 2011 1:22am
Quoting nubcake
ok, if we are going to mathify this....

there are 1,440 M1 bars in a day. 83,160 / 1,440 = 57.75.

therefore : 83,160 bars is similar to 57 days worth of M1 bars to the perspective of an intraday trader. rounding that up makes it basically 2 months of intraday data. doesn't seem like much data to me.

sorry
mubcake the problem with back testing EMA and SMA, WMA and all sorts of combinations is that it is bogus on its face.

if MR breakout knew that a 13 SMA (there really is not much difference between all MAs exept the hocus pocus, curve fitting giving more mathematical weight to the newer value crap)
is basically an average of the last 13 closing periods (assuming he is using the closing price to derive the MA) so by definition as new data becomes available the MA will drop the first value and replace it with the newer value, to always give a constant averaged price.

the problem comes because it gets plotted in the chart as a line and some one decides to put a pretty color on it and give it meaning.

the meaning should be the actual average price as its attempting to calculate the mean. and by looking at the line, you can have a visual read of what the mean has been, giving a smoothing effect to signal direction.

now add more MA's to that and you get pretty lines crossing each other post facto, giving the illusion of an infallible system of moving average crosses. this in fact has no edge real time but will always give you a great system in backtesting
AVT INVENIAM VIAM AVT FACIAM
  • #273
  • Jul 25, 2011 1:24am Jul 25, 2011 1:24am
Quoting nubcake
ok, if we are going to mathify this....

there are 1,440 M1 bars in a day. 83,160 / 1,440 = 57.75.

therefore : 83,160 bars is similar to 57 days worth of M1 bars to the perspective of an intraday trader. rounding that up makes it basically 2 months of intraday data. doesn't seem like much data to me.

sorry
So let me get this straight, you need 57 consecutive days of 1 minute bar (almost 3 trading months !) to day trade the Forex on the 1 min charts?

Do you have any idea how hugely insane this sounds?
  • #274
  • Jul 25, 2011 1:29am Jul 25, 2011 1:29am
Quoting the redlion
mubcake the problem with back testing EMA and SMA, WMA and all sorts of combinations is that it is bogus on its face.

if MR breakout knew that a 13 SMA (there really is not much difference between all MAs exept the hocus pocus, curve fitting giving more mathematical weight to the newer value crap)
is basically an average of the last 13 closing periods (assuming he is using the closing price to derive the MA) so by definition as new data becomes available the MA will drop the first value and replace it with the newer value, to always give a constant...
i'm not sure if you realize this.... but i do understand what an MA is.

your post actually had zero to do with the post of mine you quoted. so yeah, not sure where you were trying to go there dude.
Congratulations ScalaFX for coming out of the closet!
  • #275
  • Jul 25, 2011 1:33am Jul 25, 2011 1:33am
Quoting Mr Breakout
So let me get this straight, you need 57 consecutive days of 1 minute bar (almost 3 trading months !) to day trade the Forex on the 1 min charts?

Do you have any idea how hugely insane this sounds?
wow. you just don't get it do you, and you are serious too! let me paint a picture just for you.

i get a wicked idea that i think will make me millions. i backtest the idea on M1 data over 57 days of data. EUREKA I MADE A PROFIT!!!! Quick, let me load-up my account with my life savings so i can make huge profits with my clearly proven and backtested system!!!!!!

MASSIVE PROFITS FALL FROM THE SKY! Oh wait, no they don't. 57 days is stuff all time to backtest an idea on M1. ever heard of a curve-fit? it's not like it's ever been brought up on the forum before, or in this thread even.

wall, meet my head. head, meet the wall.
Congratulations ScalaFX for coming out of the closet!
  • #276
  • Jul 25, 2011 1:47am Jul 25, 2011 1:47am
Quoting Mr Breakout
My friend, 16 years and half of daily bars on 21 different markets are the equivalent of 83,160 bars! ( 198 months total x 20 trading days each month x 21 markets)

In statistical terms it is an extremely large and meaningful sample to base a serious study on.
I didn't understand correctly, I thought you mean that it is impossible to curve fit due to the size of data and not that the data is big enough to prove that your system works.

@redlion:
It doesn't matter which kind of moving average you use or if you enter after a retracement or at breakout, as long as the market is trending you make money and this is an edge in forex. The biggest problem of these systems is that you can spend years in drawdown during ranging periods until the trend of trends develops and makes for all your losses plus some handsome profit. I find it psychologically impossible but some are able to use it.
  • #277
  • Jul 25, 2011 1:52am Jul 25, 2011 1:52am
Quoting Sauron
I didn't understand correctly, I thought you mean that it is impossible to curve fit due to the size of data and not that the data is big enough to prove that your system works.

@redlion:
It doesn't matter which kind of moving average you use or if you enter after a retracement or at breakout, as long as the market is trending you make money and this is an edge in forex. The biggest problem of these systems is that you can spend years in drawdown during ranging periods until the trend of trends develops and makes for all your losses plus some...

yeah , but if instead of working on a system, you work on YOURSELF and your understanding of the market. you would always make a profit

in a range, bull or bear market and will be able to troubleshoot if needed
AVT INVENIAM VIAM AVT FACIAM
  • #278
  • Jul 25, 2011 1:59am Jul 25, 2011 1:59am
Quoting Troikaone1
This is a good point. Backtesting to me is kinda like throwing a bunch of random parameters on the wall just to see what sticks.
Absolutely not!

A serious backtest ALWAYS begins with an idea, a theory of some kind. Example : The market trends. Or the market quickly reverses when it touches a support/resistance level

And THEN we try to find a way to capture that trend using trend following indicator like the moving average, or we try to verify the support and resistance theory by finding a way to define a support/resistance level objectively .

Then there is the garbage backtest, where you simply combine a few indicators (example RSI + MACD + Bollinger Bands + pivot points), give them all the possible values and say : "Hey, look, these particular values work, let's open the bottle of Champagne".

Of course the garbage backtest is a total waste of time and has no real value, it's just a highly optimized junk with no real logic behind it, even if the profits are "amazing".
  • #279
  • Jul 25, 2011 2:00am Jul 25, 2011 2:00am
Quoting Mr Breakout
Absolutely not!

A serious backtest ALWAYS begins with an idea, a theory of some kind. Example : The market trends. Or the market quickly reverses when it touches a support/resistance level

And THEN we try to find a way to capture that trend using trend following indicator like the moving average, or we try to verify the support and resistance theory by finding a way to define a support/resistance level objectively .

Then there is the garbage backtest, where you simply combine a few indicators (example RSI + MACD + Bollinger Bands + pivot...

back to my original post:

"b/s"
AVT INVENIAM VIAM AVT FACIAM
  • #280
  • Jul 25, 2011 2:19am Jul 25, 2011 2:19am
yes, curve-fitting is a problem when backtesting. The problem comes when a certain system only works with a couple of parameters.

What I try to find are systems that work with a lot of different parameter settings, so that I can see if the system is worth my while or only curve-fitted parameters work.

Obviously a system also needs to make sense. I rarely use indicators for my systems, most are directly based on price. For me ma's are good to define trends, e.g. if they all point down we are very likely in a downtrend.

Anyways, the curve-fitting problem was the reason why I was looking into ways to define robustness. Well, and the solution was that if I can use a couple of hundred parameters and achieve profits in most scenarios and most instruments, then it must have an edge wihin itself. Then for me that is a viable system.
0 traders viewing now