Is the "system" part really so easy to come by? | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Broker Challenge "NDD/STP" vs. True "ECN" and "Mkt Mkr" broker talk 46 replies

Dealing with "Invalid Price", "Requote", "Server Busy" in MT4 3 replies

Price consolidation system: really as easy as counting 1-2-3? 43 replies

The Really Really Boring Forex System 97 replies

Part Time Wealth Building System - "High-Low Trending" 9 replies

Is the "system" part really so easy to come by?

  • #61
  • Mar 16, 2007 12:33am Mar 16, 2007 12:33am
Quoting FXTerminator
Really?

Then set your stop at 1000 pips and your take profit at 1 pip and see if your "very robust" trading system is STILL making money after 5000 trades...

lol of course it would be. In other words, it never has a bad trade? You meant a stop one pip away and take profit 1000 pips away right? I'll assume you did and my answer would be, no, there is no such thing a perfect system, but in a strong one, it should not make a big difference where your stops or take profits are.
  • #62
  • Mar 16, 2007 1:40am Mar 16, 2007 1:40am
Quoting Bemac
That's interesting. I wonder if you have ever re "optimized" 1,000 bars later & if you have, did the test return the same number?
Oh yes, of course I have. This is exactly what I mean by applying "valid methods" to optimizing. You see, my background is reseach in theoretical and applied Psychology. When I develop, for example, a personnel selection and placement system, it has to be not only reliable and valid, but also stable across time. If I develop a theoretical model, it has to be replicable at different times and settings, by different researchers, etc. And yes, there seem to be some stable trading systems. For example, now I started forward-testing a simple crossover system which I tested 6,300 bars back, then tested samples from different markets (trending, ranging) in 2005, 2002, and 1999. And in each sub-sample, the price has to break out of a Keltner band border by exactly 11 points, not 10, not 12, to provide the signal for opening with highest profit expectancy. Such things amaze me each time anew, but actually, they shouldn't. If there was no stable predictability, noone could be a stably profitable trader. What traders call "curve-fitting" in negative sense is exactly the right thing to do, only that the sample on which the "curve" (the model) was fitted should be an appropriate one.
sans peur et sans reproche
  • #63
  • Mar 16, 2007 1:58am Mar 16, 2007 1:58am
Quoting aicccia
lol of course it would be. In other words, it never has a bad trade? You meant a stop one pip away and take profit 1000 pips away right? I'll assume you did and my answer would be, no, there is no such thing a perfect system, but in a strong one, it should not make a big difference where your stops or take profits are.
Aicccia my friend, you are just pulling my leg or you are posting with a 6-pack of Heineken next to you and five bottles are already empty...

No seriously, you mean to tell the traders here at FF that ANY stop loss and ANY take profit will make money? That's a 1 pip stop will generate the same profits than a 1000 pip stop loss?

So where is your money management if anything will do?
  • #64
  • Mar 16, 2007 2:11am Mar 16, 2007 2:11am
Quoting ilanr
If there was no stable predictability, noone could be a stably profitable trader. What traders call "curve-fitting" in negative sense is exactly the right thing to do, only that the sample on which the "curve" (the model) was fitted should be an appropriate one.
Curve fitting is not a good idea. Basically you are shaping your system around your expectations and not how the market is behaving. If you backtest and your equity goes up nice and steady without any dips (only some flat spots) over a long period of time, then you are on to something. If you start curve fitting and you have a nice net profit, but the journey getting to the profit looks like a mess, then you have come up with a system that probably has no real edge on the market. A post a while back mentioned having "to take the randomness out of the system." I think this is the backwards thinking that gets in the way of system building. Trying to manipulate a system that does not work, and "fix it." The correct approach is to say what aspects of the market can I identify as not being random, and can I profit from it? Then once you've identified what you're after, write the system.
  • #65
  • Mar 16, 2007 2:15am Mar 16, 2007 2:15am
Quoting Keys
The correct approach is to say what aspects of the market can I identify as not being random, and can I profit from it? Then once you've identified what you're after, write the system.
Sure, sure, but once you write and test the system you will immediately notice that only a specific stop and take-profit levels seem to be making the most money, any other stop/limit will only decrease your profits gradually, until your once winning system becomes a losing system.

And we are back to square one.

PS: finding the right stop/take-profit level for your specific trading system has nothing to do with curve-fitting, even if it seems that way.
  • #66
  • Mar 16, 2007 2:31am Mar 16, 2007 2:31am
Quoting FXTerminator
Sure, sure, but once you write and test the system you will immediately notice that only a specific stop and take-profit levels seem to be making the most money, any other stop/limit will only decrease your profits gradually, until your once winning system is a losing system.

And we are back to square one.
Right, I figured that discussion was settled. Of course there's always the fine print, past performance is no guarantee of future results. Personally I would rather base my decisions on tons of stat data, rather than shooting from the hip. You've got to make the best decision with the resources that are available. Those of you who prefer to take a more subjective approach, could be on to something. Perhaps after staring at tons of charts for years, that neural net in your skull can indentify the best exits with ease, but most likely if you can't analyse your exits objectively, then you are probably trading irrationally (which I am often guilty of as well).
  • #67
  • Mar 16, 2007 2:44am Mar 16, 2007 2:44am
Quoting FXTerminator
Aicccia my friend, you are just pulling my leg or you are posting with a 6-pack of Heineken next to you and five bottles are already empty...

No seriously, you mean to tell the traders here at FF that ANY stop loss and ANY take profit will make money? That's a 1 pip stop will generate the same profits than a 1000 pip stop loss?

So where is your money management if anything will do?
No, this is what I'm saying.

If the optimal values of a system are SL at 43 and TP at 21, then IF this is a robust system, then SL values anywhere from around 30 to around 55 and TP values from around 7 to around 35 should still give the system healthy profit. IF it is a weak system, then the system will cease being profitable by differing just marginally from the optimal values. Your goal should be to have a system where the main values for the system can be even significantly different and you still make profit. Now of course it's impossible to have system that makes profit no matter what values the parameters have, then the parameters would be pointless.
  • #68
  • Mar 16, 2007 2:54am Mar 16, 2007 2:54am
Quoting Keys
Curve fitting is not a good idea.
It's much a matter of language use. Any model building is curve-fitting. In this sense, we all look for a good curve-fit. What you mean is curve-overfitting. This is a bad thing. Statistics is a science that allows us to discriminate between the two.

Quoting Keys
If you backtest and your equity goes up nice and steady without any dips (only some flat spots) over a long period of time, then you are on to something.
Exactly. Stable performance is one of the properties of a validly "curve-fitted" (consturcted) model based on past performance that predicts future performance.

Quoting Keys
The correct approach is to say what aspects of the market can I identify as not being random, and can I profit from it?
Couldn't agree more. And, as Fxterminator has already suggested, a "non-random" solution logically implies a unique solution, or a defined set of equivalent solutions.
sans peur et sans reproche
  • #69
  • Mar 16, 2007 3:03am Mar 16, 2007 3:03am
Quoting aicccia
No, this is what I'm saying.

If the optimal values of a system are SL at 43 and TP at 21, then IF this is a robust system, then SL values anywhere from around 30 to around 55 and TP values from around 7 to around 35 should still give the system healthy profit. IF it is a weak system, then the system will cease being profitable by differing just marginally from the optimal values. Your goal should be to have a system where the main values for the system can be even significantly different and you still make profit. Now of course it's impossible to have system that makes profit no matter what values the parameters have, then the parameters would be pointless.
Every price behavior parameter is normally distributed around a specific central value. You are absolutely in the sense that you can deviate from this value and still be profitable. However, the rational decision would be to use exactly the central value that provides for the optimal solution.
sans peur et sans reproche
  • #70
  • Mar 16, 2007 3:05am Mar 16, 2007 3:05am
Paraphrasing Fxterminator's quotation, "It's all about statistics kid, and the rest is just conversation."
sans peur et sans reproche
  • #71
  • Mar 16, 2007 3:11am Mar 16, 2007 3:11am
Quoting aicccia
If the optimal values of a system are SL at 43 and TP at 21, then IF this is a robust system, then SL values anywhere from around 30 to around 55 and TP values from around 7 to around 35 should still give the system healthy profit.
Now we are getting somewhere

But here is my point, I will use the same numbers from your example:
SL = stop loss, TP = take profits, we are backesting a system and here are the hypothetical results after 5000 trades:

SL = 43 pips, TP = 7 pips ---> total profit: +180 pips
SL = 43 pips, TP = 21 pips---> total profit: +1500 pips
SL = 43 pips, TP = 35 pips---> total profit: +7200 pips

So even if all TP levels are making money, the 35 pip TP is the one to keep, it is by far the optimum TP, NO question asked, it is just plain math and statistical edge.

We could do the same thing with stoploss.

Contrary to popular belief, an extremely robust trading system does not make roughly the same amount of money ("healthy profit") as you move the stop or the take profit, on the contrary the profits will decrease rapidly because the system is losing its statistical edge very quickly, while so-so system seem to work fine with a fairly larger range of optimal values.

PS: 4:21 EST, my $1.9400 sell order on the GBP/USD has been triggered, now let's see...
  • #72
  • Mar 16, 2007 9:19am Mar 16, 2007 9:19am
Quoting FXLogic
There are a million profitable systems. All of them are very hard to find.
i remember reading this exact line. it stuck in my head too, wierd. i am trying to remember exactly who said this, its on the tip of my tounge....
Relax and be happy.
  • #73
  • Edited Nov 3, 2007 7:51am Nov 2, 2007 11:17am | Edited Nov 3, 2007 7:51am
Quoting mathematician
Early last summer I programmed a coinflip strategy and tested it with TradeStation. There rules were simple:

1. Enter at 3:00am EST (8:00am GMT, i.e. London's open). I chose the London open because it's generally a very volatile time where big moves can be expected.
2. The direction (long/short) is chosen randomly by a coinflip. For my program I had the computer generate a random number between 0 and 1, if it was above 0.5 then I went long, if it was below 0.5 then I went short. So essentially this is a coinflip.
3. There was a 25-pip stop.
4. Hold the position until 12:00pm EST (5:00pm GMT, London's close). Then exit at market.

Oh...and I tested this with GBP/USD.

Results? Well I ran the simulation over 5-years of history (I think it was hourly data) and I did this test 10 times...it was basically a Monte Carlo simulation. I should have run many many more of course, but I was at work. Now in the simulation I started with a $5000 account.

Guess what...surprise surprise! Exactly 50% of the time it worked. lol That's right 5 out of the 10 times the account grew over the five years to something bigger. Some times the account doubled but more often it was up like 8-fold. In some cases the $5,000 grew to almost $80,000. The other five times the account was shot, lost all the money eventually at some point in the five years of simulation.

What does this mean? Well if you think about it that still means it's profitable...I mean let's say you did this 10 times with 10 separate accounts of $5,000. Total losses = 5*$5000 = $25,000. But total gains were much higher...something like $200,000. So you would net about $175,000. This was all without compounding/scaling the lot sizes.

In short: Yes. It can work...well some times. But I still want to use that system. Hehe

-mathematician
So instead of the 10 accounts with $5000 each, why not do two accounts of $25,000 each, since it works half the time, you still lose the 25,000, and the other account grows, right?

Wrong, because you don't have a large enough sample for it to be statistically valid. Even though it works half the time, it might fail to make money twice in a row.

Then this would mean that it would be better to spread the money out into 20 or 30 accounts. Half would blow out, you lose half your money, the other accounts grow.

So, then, would it come out the same if you just did it all in one account, but did 30 small trades? No.

The problem is, if it were a large enough sample size (great number of tiny trades), half the trades would be long and half short.

Which is the same as if you had 30 accounts (or 10, or 20) - over time, half would be long, the other half short, on any given trade.

Without spreads, you'd break even over time, no matter how you did it.

Which leads us back to what Merlin said. It's negative expectancy.

Which also means there was something wrong with your model.
  • #74
  • Edited 2:51pm Nov 2, 2007 1:49pm | Edited 2:51pm
Gotta agree with the thread starter first post completely.
  • #75
  • Nov 3, 2007 5:30am Nov 3, 2007 5:30am
Hi.
I can't seem to sleep tonight and sometimes come here to read the posts. I've been an intraday trader for a decade or so and after reading so much on here, I just feel the need to speak my mind. I don't think I know it all and what works for me may not work for others, and so. As far as systems go, I can't help but think that all the searching and backtesting and optimazations and such, are in a way, people looking for a free lunch. In my mind, the very nature or the markets, and I won't long wind this point, are not going to allow the systems to work with any long term success. God knows thousands of good men have tried an failed. My conclusion, and it wasn't made yesterday, is that the money is to be made consistently with naked price action. It works the same way today as it did fifty years ago. It does take alot of experience (work) but it's not brain surgery either. I do believe there is alot of pychology involved and that can take time, depending on the person. It takes experience to know what to trade and when to trade and of course, how to read price action with sufficent accuracy, money management, proper brokers, WHEN TO STAY ON THE SIDELINES, etc. It takes years and I think most new traders just want to hook up with a system that will make them a million dollars in six months. IT AINT GONNA HAPPEN. While I'm in rant mode I'd like to make a point about time frames. Whenever I come here there always seems to be a thread urging people to use longer time frames. I use 1 min bars. I read a thread on here not long ago where someone stated that they use longer time frames because they would rather set their orders and go play golf. Now maybe that works for some of you, I doubt it but maybe, but to me there it is, that free lunch mentality. Try pulling up a chart of the last 24 hours of AUD/USD. Use 1 min bars and compress them a bit. I'm not going to get into how to trade price action but you don't trade 1min bars like you do 1 day bars. You really new traders, stare at that chart for a while, print it out. Note the highs and the lows and the retracements caused by reactionary buying or selling from prices being hit and the absorbtion of that buying and selling. Stare at that one chart for a long time. My losing trades range from 1 to 7 pips and my winners range from 1 to 80 pips. As for the argument that trading like this requires sitting in front of a screen. Your damn right it does, it's a job. On the other side, I don't spend one minute doing technical or fundamental analysis. I don't have to give one minutes thought or worry to what the market is doing when I'm not in a trading session. I'm not saying that longer time frames cannot be successfuly traded, but what I am saying is I think alot of new traders really overlook the benefits and profits of short time frames. While these shorter time frames do require discipine, I do firmly believe that these shorter time frames offer the greatest statistical edge. In other words, under certain price action conditions, I don't think I have a statistical edge, I know I do. On longer time frames I'm sure some of you have an edge as well. For myself, all the fundamental and technical analysis in the world to determine what something will do over the next week will not come close to the edge I had on that chart I mentioned earlier. Is there noise and unpredictable crap, yea and that's what stop losses are for. So for some of you new traders that are searching thru systems and indicators and methods, it may be worth your time to at least consider the points I've made, simple, old school. Sorry for being so long winded. Best of luck to all.
  • #76
  • Nov 3, 2007 8:07am Nov 3, 2007 8:07am
Quoting SoyYoConBonete
Let's take for example james16 system thread (please anyone feel free to correct me if I'm misquoting or misunderstanding anything about it). The basic idea of his system is to enter after certain candle formations which happen at "confluence" points. The truth is that if I want to enter any trade, almost always I will be able to rationalize it with the system. There are hundreds of points from where to draw the fib, hundreds of different trendlines, several types of pivots, several types of support/resistance lines. I could probably find almost any post in the chart to be a "confluence" point. I think the reason why james16 (and others) is successful trading it is because he has learned which trendlines are valid and which ones are not, which fibs are valid, etc.
I do get your idea, and you are right. Let me clear up a few things, though.

James 16 is not a system. It is a body of data that is, IMO, geared to help train you into that market "feel" -- how to see the charts, know what's more important, what's less important. If several signals align, that's confluence. If you have confluence of several signals that you know are strong, that's an A trade - high probability.

There are usually only a few points from where to draw fibs (where -- that's subjective, BUT, there will be many others drawing them at the same place you do, and if you draw 2 or 3 fibs, then confluence points will agree with a greater number of traders' fibs) or trendlines (there is a right way to draw them, and many wrong ways.)

Quote
My point here is that the difficulty with discretioary system lies in really LEARNING it, rather than just memorizing the steps required to trade it. You also have to learn when not to trust your system, and that comes from a subjective market feel; which takes (for most) a long time to learn and, as fiji says, is very hard to separate from emotions.

I also believe this to be completely true. You've summed up the difficulty of learning to trade. You can't USE a system, you BECOME the system. The system is the intetgration of everything you've learned and seen, and no two traders will ever trade exactly the same way. It's a developed intuition.

A profitable trader can trade any "system" because when they evaluate the signals, they're really trading their own system.

They decide which trades to take, and when to exit. They don't follow the rules mechanically, no matter what they say they do.

Nobody does. Do you take a signal right before NFP? Of course not.

That's using discretion. Therefore, all traders are discretionary.
  • #77
  • Nov 3, 2007 8:10am Nov 3, 2007 8:10am
Quoting Rock
Hi.

....

. Best of luck to all.
Wow! Awesome post, Rock!



It's taken me a year to face up to what you've said. It's so much easier to spend 20 minutes learning someone's system than it is to really learn to trade. Hey, they took 6 months to develop it, so they've done all the work, and all I have to do now is trade it!

Then when it fails, hey, there's another one that really looks good!

I guess it never occurred to me that during that 6 months of sweat, that person learned a few other things about trading that are part of his system, even though he can't write them all down or maybe isn't even aware of them.
  • #78
  • Nov 3, 2007 9:43am Nov 3, 2007 9:43am
1) Mechanical Trading.
I think there's no working mechanical trading systems out there. Never seen one system working stable over long period of time. They all fail at some point, often quite badly. The system should include everything from trading signals to the money management, so even the computer could trade it.

2) Discretionary Trading.
Discretionary trading or "discretionary system trading" (if it's discretionary, then it isn't really a system longer, just a gut feeling trading).

3) Conclusion.
It's easy to prove whether a mechanical trading is possible or not simply by backtesting the data on long period of time.
It's impssible to prove anything about discretionary trading, as it's "discretionary".
  • #79
  • Feb 9, 2011 4:57am Feb 9, 2011 4:57am
I know this is a reply to a very old thread, but as i was reading the discussion on pages 4 and 5 on this thread about how SL positionning was affecting overall profitability, position sizing and reward/risk ratio, it makes me think of a freely readable book from Charlie Wright: "Trading as a business". It is quite short and you can read it at http://www.elitetrader.com/tr/index.cfm?s=17

In the chapters from "Profile of a Winning Strategy", "The Art of Strategy Design", ... and following, after taking as exemple a ridiculously naive MA crossover strategy, goes into doing a complete series of backtests to determine which distance of SL is the most profitable, for this strategy and currency pair.

And as he mentions, the best setups are not always the most evident ones, and he also shows that the same strategy can have many ways to be profitable (choice of different setup, entries, risk, MM, exits, ...), but most importantly (and this is also the subject of this thread) can be modified and fitted to your personality. As he says, there is no point in a very profitable strategy given by someone, but that your personality can not trade.

The good part about the book is it is driving reflection by driving an experiment and self building and adjusting the wheels and parts of whatever strategy so that the result is your system you can trade successfully if you have discipline to believe in what you saw.
  • #80
  • Apr 9, 2013 9:33am Apr 9, 2013 9:33am
Quoting Rock
Hi. I can't seem to sleep tonight and sometimes come here to read the posts. I've been an intraday trader for a decade or so and after reading so much on here, I just feel the need to speak my mind. I don't think I know it all and what works for me may not work for others, and so. As far as systems go, I can't help but think that all the searching and backtesting and optimazations and such, are in a way, people looking for a free lunch. In my mind, the very nature or the markets, and I won't long wind this point, are not going to allow the systems...
wow! I'm so lucky to have come across this,it changed my view completely....only one post in FF....this Guy is the real deal,its true system trading is ultimately curve fitted ....bless you whoever you are.

this is one of the greatest post in FF.
wo-yoy! wo-yoy! wo-yoy! wo-yoi! wo-yoy-yoy-yoy!
0 traders viewing now