Why inverting trades doesn't work? | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Why do indicators not work/work? Why is Forex not truly random? 129 replies

Inverting the chart 3 replies

Inverting currency pairs in mt4? 2 replies

Overlaying/Inverting Indicators? 3 replies

Why doesn't it work more recently? 17 replies

Attachments: Why inverting trades doesn't work?
Exit Attachments

Why inverting trades doesn't work?

  • #21
  • Aug 8, 2010 9:10pm Aug 8, 2010 9:10pm
  •  Czech
  • | Joined Feb 2010 | Status: You Don't Know What You Don't Know | 111 Posts
Quoting Craig
I can't believe the amount of crap people write on this forum (the above being a prime example), Razzle nailed the answer first time, yet everybody clatters on about all sorts of irrelevant shit as if he had never posted.

Maybe it was because the OP posted this right afterwards
Quote
How can that be random when the data we are running it against also is static? What is it I fail to understand?
in response to Razzel's answer.

Maybe it was because this is a discussion board.

Some people clattered on about why it had no practical application (myself included) and then Mr J nailed the OPs question on the head.

chill out and try to pay attention. This thread was at least on topic until you posted.
  • #22
  • Aug 9, 2010 2:51am Aug 9, 2010 2:51am
Quoting FXEZ
There are only two possibilities. Either the original losing system doesn't have a large enough average loss per trade to cover the actual trade costs (spreads) - [less likely given that I don't think the MT4 backtester takes spread into account],
Mt4 dose take spreads in to account in back tests, if you run it in visual mode then you will notice that it opens orders above and below market price, and this spread is determined by the current spread in account which can be 3 to 4 times more if you do the test during the weekend.
  • #23
  • Aug 9, 2010 3:38am Aug 9, 2010 3:38am
Here's my 5 cents.
You simply cannot just invert a strategy. Here's a couple of things to succesfully (try to) revert a loosing strategy:
1. you need to make the spread irrelevant. Eg. you cannot have targets at 10 pips when the spread is 3 pips. 100 pips (TP and SL) with 3 pips spread looks better.
2. anything different than 1:1 R/R will NOT work (plasmapants has a point).

And here's a few things about MT's StrategyTester:
Do NOT have interest points during the same candle. Entry point, TP, SL, even decisions about the movement of the trailing stop, and any combination of above should NOT be in the same candle, or the next ones if they are engulfing the previous one.
The reason is simple. Due to the fact that MT's ST has only OHLCV data to work with, it creates an interpolated tick version on how actually the pair has moved, simply does not reflect reality.
So, you could not know in each order the High or Low was hit first. For example: for an bullish candle, you can presume that the Low was hit first, then the High, but does this really happen in reality? This is the fallacy of MT's ST, it doesn't work in basic form, you need to make sure that the data interpolation doesn't interfere with your strategy.

Cheers.
  • #24
  • Aug 9, 2010 4:29am Aug 9, 2010 4:29am
I agree the spread would be killer. But try other Money management techniques to lower the bleed or maybe even make it marginally profitable. The entry is random, but if you set stop loss at 40 and and take profit at 100 , you may end up with an occasional runner which may offset the losses. Try it on both the original system and the inversion, and share the results. I think it would come down to cutting losses short and letting winners run even in a random entry system
  • #25
  • Aug 9, 2010 5:12am Aug 9, 2010 5:12am
Quoting Craig
I can't believe the amount of crap people write on this forum (the above being a prime example), Razzle nailed the answer first time, yet everybody clatters on about all sorts of irrelevant shit as if he had never posted.
Well, yes, Razzle was right if you believe either trading in general is random or that the original poster's EA was based on random entry.

But since trading isn't random...
  • #26
  • Aug 9, 2010 5:30am Aug 9, 2010 5:30am
Quoting McKeen
I recently saw a BBC Planet Earth documentary where they compared the humans failure of understanding the laws of the universe with a dog's failure of understanding simple physics. In this problem I am about to present to you I feel like the dog in the example.

I have been creating EA´s for Metatrader for more than a year now with limited success as far as any results goes. Because of the failures to produce a continious positive equity-curve over any longer period of time whatsoever and the relative ease to produce a negative equity-curve...
reversing one set of unoptimized trading parameters with the opposite set of unoptimized trading parameters leads to the same type of equity curve.

regards
daytrading
Enter Signature
  • #27
  • Aug 9, 2010 5:58am Aug 9, 2010 5:58am
Quoting McKeen
I recently saw a BBC Planet Earth documentary where they compared the humans failure of understanding the laws of the universe with a dog's failure of understanding simple physics. In this problem I am about to present to you I feel like the dog in the example.

I have been creating EA´s for Metatrader for more than a year now with limited success as far as any results goes. Because of the failures to produce a continious positive equity-curve over any longer period of time whatsoever and the relative ease to produce a negative equity-curve...
both curves APPEAR to be negative...they're in fact random....because the signal triggers are random in this case...

There is only one way to turn this system into a true negative edge

....and that is

when you start trading it manually....

the reason being the human mind is ill-equipped to handle the drudgery of random events and the stress coupled with it...

so best develop a system that has a truely positive edge instead of inverting and what not!

Trade Well!
GlobalMacro
Hear the music before the song is over ...
  • #28
  • Aug 9, 2010 6:29am Aug 9, 2010 6:29am
Quoting plasmapants
heres my penny's worth, if by inverting you mean your now selling on your orignal buy signal and vice versa.

if your trading with a stop of 50 pips and a profit of 100 pips, its easier to reach your stop then it is to reach your profit.

in your orignal system, when you buy the price may head to your profit by atlest 50 pips before reversing and hitting your stop.

so when you invert the trade, and you have a 50 pip stop, that stop will still be hit.

if you are going to invert the signal you might wanna try and invert the stop and profit targets...
... Plasmapants has the winning answer. In spite of the market moving in discernable trends from time to time, market noise generally takes out stops in both directions.
"A is A"
  • #29
  • Aug 9, 2010 6:58am Aug 9, 2010 6:58am
It is not about inverting your methodology or not, its rather whether your system can CONSISTENTLY lose money or win money. If it cannot consistent lose money, inverting it is just as disastrous as not inverting it.
  • #30
  • Aug 9, 2010 7:06am Aug 9, 2010 7:06am
Quoting plasmapants
heres my penny's worth, if by inverting you mean your now selling on your orignal buy signal and vice versa.

if your trading with a stop of 50 pips and a profit of 100 pips, its easier to reach your stop then it is to reach your profit.

in your orignal system, when you buy the price may head to your profit by atlest 50 pips before reversing and hitting your stop.

so when you invert the trade, and you have a 50 pip stop, that stop will still be hit.

if you are going to invert the signal you might wanna try and invert the stop and profit targets...
Quoting dakoo
I agree the spread would be killer. But try other Money management techniques to lower the bleed or maybe even make it marginally profitable. The entry is random, but if you set stop loss at 40 and and take profit at 100 , you may end up with an occasional runner which may offset the losses. Try it on both the original system and the inversion, and share the results. I think it would come down to cutting losses short and letting winners run even in a random entry system
I'm not agree with you, this is not the reason, I tested the same inverting the trades AND the SL and TP, if in the original the SL is 10 and the TP 20, in the inverted EA the TP is 10 and the SL 20, and results were the same, the total loss.

There's another reason why inverting trades don't work (inverting SL and TP too), the reason is that odd are against us, the typical example is that you need to win a 100% to recover from a loss of 50%: if I have in my account 1,000 dollars and I lose the 50%, I have now 500, so I need to double up my account (win 100%) to recover what I've lost, you have more information in these threads:

- http://www.forexfactory.com/showthread.php?t=86622
- http://www.forexfactory.com/showthread.php?t=243690

But you can minimize this risking less than a 1% in each trade, that's why I still think that the main reason is the spread, I will prove this in the next post.
  • #31
  • Aug 9, 2010 8:35am Aug 9, 2010 8:35am
Quoting plasmapants
heres my penny's worth, if by inverting you mean your now selling on your orignal buy signal and vice versa.

if your trading with a stop of 50 pips and a profit of 100 pips, its easier to reach your stop then it is to reach your profit.

in your orignal system, when you buy the price may head to your profit by atlest 50 pips before reversing and hitting your stop.

so when you invert the trade, and you have a 50 pip stop, that stop will still be hit.

if you are going to invert the signal you might wanna try and invert the stop and profit...
The trades are mirrored in all aspects, tp, sl's, triggers + sizing. See post #5 for details.

/ McKeen
  • #32
  • Aug 9, 2010 8:42am Aug 9, 2010 8:42am
Quoting Mr J
Don't be so sure. He'd have to pay a 5 pip spread to lose $100k over 2000 trades, trading 1 lot per trade. If he did, fine, but if he didn't, there's something else at work. It would be nice to know the average spread paid.

I also question the experiment. You can't just trade the opposite signal, you have to take the same entry, and flip the target profit and stop. Go long precisely where you went short, take profit where you would have been stopped out, and put the stop where you would have taken profit. Otherwise we're just trading one bad...
Hi, spread is 2.0-2.7 pips in both tests.

As stated in my previous reply (above) it is a truly inverse test in all aspects, see post #5.

/ McKeen
  • #33
  • Aug 9, 2010 8:53am Aug 9, 2010 8:53am
Quoting Platypus
Ever heard about Murphy's law?...

... One way of playing the "inversion" game would be to "trick" the hard wire on your blue print by trading out of consciousness. Thats what researches do with placebo...right? If you take somebody and program a platform so losses will show as win, and and win as losses ( off course without the guy having any idea of it), the results will amaze you.
It is an interesting idea especially since this is basically what most brokers do I hear, betting against the trader. I am going to create such a project in the future, however I seriously doubt it could be effective in practice. But if they can, why couldn't we?

/ McKeen
  • #34
  • Aug 9, 2010 8:59am Aug 9, 2010 8:59am
Quoting Dutch_trader
Mt4 dose take spreads in to account in back tests, if you run it in visual mode then you will notice that it opens orders above and below market price, and this spread is determined by the current spread in account which can be 3 to 4 times more if you do the test during the weekend.
You are absolutely right, the EA must differ from ask / bid prices and if this isn't honored in the code the OrderSend & OrderModify syntaxes ends up with errors which this EA is free from at this point in time.

/ McKeen
  • #35
  • Aug 9, 2010 9:15am Aug 9, 2010 9:15am
Quoting csebastian
And here's a few things about MT's StrategyTester:
Do NOT have interest points during the same candle. Entry point, TP, SL, even decisions about the movement of the trailing stop, and any combination of above should NOT be in the same candle, or the next ones if they are engulfing the previous one.
The reason is simple. Due to the fact that MT's ST has only OHLCV data to work with, it creates an interpolated tick version on how actually the pair has moved, simply does not reflect reality.
So, you could not know in each order the High or Low was...
I must say I am partly with you on this and partly not (but i do not know for certain)

I have also been wondering whether it really counts each tick individually when running the backtest for "every tick". This is because if I look in History Center on the downloaded data i can scroll thru every single OHLCV downloaded. By doing this I have noted that every entry in that list differs witb one minute. (I have of course downloaded all data for the period on all tf's including m1 resulting in 90% mod quality).

If this observation is correct then you are right in what you are saying about the tester not knowing if the hi or low was touched first on the candle, however that should then mean every M1 candle and not the TF of which the test is running.

If I have missed something vital about this which you know of please enlighten me.

Cheers!

McKeen
  • #36
  • Aug 9, 2010 9:23am Aug 9, 2010 9:23am
Quoting Rand Yap
It is not about inverting your methodology or not, its rather whether your system can CONSISTENTLY lose money or win money. If it cannot consistent lose money, inverting it is just as disastrous as not inverting it.
That is just what I felt was a success with this specific test curve. I created it to be consistent and i must insist that it is considering the extended time period ( 2 1/2 years ) and high amount of trades (2000+). I have seldom seen a more lined up curve in any direction.

/McKeen
  • #37
  • Aug 9, 2010 9:44am Aug 9, 2010 9:44am
why is this so hard?

if you have been coding ea's for x amount of time then you would be able to easily write an ea that eliminates the spread from the final results... thereby giving you a clean non-spread affected result.

from that you will see that with or without spread you are getting nailed by market chop.

case closed.

put the strategy tester into visual mode and watch what it is happening. you will see why the ea's are all failing.

and what is there to 'partly be with' on the aspect of mt4's strat tester generating random ticks within known candles? that IS FACT. if you have M1 data you have M1 data... and no setting is going to recreate the ticks within that data exactly the same each time because it is simply pseudo-randomly generated each time. run a backtest on an ea, then run it again, and then again. the results will always differ because of the pseudo-randomly generated ticks.

:S
Congratulations ScalaFX for coming out of the closet!
  • #38
  • Aug 9, 2010 10:47am Aug 9, 2010 10:47am
Quoting McKeen
In original curve the following statics are there:

All trades are triggered by 4 types of stem-candles. (hammers & hanging man + their respectives)
StopLoss is the High / Low of that stem-candle.
Takeprofit is when an opposite candle appears.

In the reverse curve the statics are:

Identical trigger candles but reverse trade direction. (Identical to the above's trigger)
StopLoss is when an opposite candle appears. (Identical to above´s tp)
Takeprofit is the High / Low of the stem-candle. (Identical to above´s SL)

/ McKeen
You have only reversed the entry - nothing else. If your TP and SL was EQUAL and STATIC - it will allow real reverse of the system. But in that way it is NOT reversed.
  • #39
  • Aug 9, 2010 10:55am Aug 9, 2010 10:55am
Quoting GlobalMacro
both curves APPEAR to be negative...they're in fact random....because the signal triggers are random in this case...

There is only one way to turn this system into a true negative edge

....and that is

when you start trading it manually....

the reason being the human mind is ill-equipped to handle the drudgery of random events and the stress coupled with it...

so best develop a system that has a truely positive edge instead of inverting and what not!

Trade Well!
GlobalMacro
This post is for you and for Craig and Razzle, here you have a simple EA coded by miself (you can edit it and fix mistakes, if they are) that is not random at all, a system with fixed entries, and the reversal method, it is a daily breakout EA:

hour1 = the hour to begin to open the trades, if they have to be opened.
hour2 = the hour to close all previous trades, if they are still open.

I'm using hour1 = 1 (the beginning of the asian session + 1 minute) and hour2 = 22 (the closure of the WS session + 1 minute) in GMT, you can change it to your timezone.

threshold = if the high is too near to the low of the day, don't open any trade, threshold must to be > than SL+TP.

breakout = pips (in 5 digits, 1 pip = 10...) when the price break the H with H+breakout, then buy, when the price break L with L-breakout, then sell.

divisor1 and divisor2 is for calculate the volume, they are in 100 and 10 to test it in a broker with micro lots of minimal lot size, you can change it too, it is a rudimentary volume calculation, but works for testing purposes. divisor1 is about the leverage (you can reduce divisor1 as much as leverage increases) and divisor2 is about the minimal lot size...

Initial balance: 10000, leverage: 1:500.

bought = true, there is a buy already open, you can't buy again, the same with sold...

Tested for all the year 2009.

These EAs work with sell stop and buy stop orders (buy limit and sell limit in the reversed EA)... That's because the Closeall() function of the end to close all previous pending orders after the closure of the WS session.

Well, here you have the EA and the reversed EA (if you want to test the reverse EA with a SL not equal to TP, reverse too the SL and the TP, I'm using here the same pips for both: 20 pips, and 1 pip of breakout) and their backtests for the same period:

Daily Breakout results:
Attached Image (click to enlarge)
Click to Enlarge Name: Daily Breakout.gif Size: 6 KB


Reversed Daily Breakout results:
Attached Image (click to enlarge)
Click to Enlarge Name: Reversed Daily Breakout.gif Size: 7 KB


You can see in the backtest reports how sell limit and buy limit orders of the reversed EA are every day at EXACTLY the same price than the buy stop and sell stop orders of the original EA respectively, but because the spread, orders are hit at different times, and some orders are not hit at all, and this makes that in the end both EAs end up losing money, in the test the total amount of orders are different, but if you look order by order you can see the effect of the spread.

You have to take in consideration also my previous post (http://www.forexfactory.com/showpost...postcount=30): odds are against us, to recover from a drawdown is harder than to lose what you have won, you can see this also in the backtests.

When the daily breakout fails, at April 1th, with 800 dollars, the balance of the reversed EA is 9,000, when according to your theory it would be 19,200.

The spread when I have made tests is 1.1 pip in both versions, and this is a generous spread, imagine the results with a 2 or 3 pips of spread...
Attached File(s)
File Type: zip Backtests.zip   33 KB | 370 downloads
File Type: mq4 Daily Breakout.mq4   4 KB | 380 downloads
File Type: mq4 Reverse Daily Breakout.mq4   4 KB | 347 downloads
  • #40
  • Aug 9, 2010 10:55am Aug 9, 2010 10:55am
Many good answers have already been given, starting with Razzle.

Whether or not your signals are completely random remains questionable. Over your sample, ignoring transaction costs, your method had a positive edge, and consequently the inverse a negative edge. This is confirmed by the fact that your method busted after 2400+ trades, and the inverse busted after 1800+ trades.

Whether the apparent edge in your backtest is due to coincidence or because your system does have value - for example under certain market conditions - remains to be researched.

Unfortunately, your method's edge is not able to overcome the transaction costs, and thus both methods appear unprofitable in your backtest. Like others stated, the reason for this is that your method still provides too many random signals, and then its just a matter of time before the spread eats up your account

You see a clear negative correlation between both methods at some point in time, where your method was making profits, while the inverse was losing significantly. This got me thinking: If market conditions are "sticky", and your method performs well under certain conditions and badly under others, you could devise a money management strategy that increases risk when there appears a strong negative correlation between the method and its inverse. This is something that I will hypothesize about this summer, and will attempt to research further later on.
The nail that sticks out gets hammered back in
0 traders viewing now