New Guy and Exits | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Advise and Guidance for a NEW FOREX guy 11 replies

New Guy Making A Fool of Himself 6 replies

Any advice for new guy? 56 replies

New Guy Here... 6 replies

New guy here.... 4 replies

Attachments: New Guy and Exits
Exit Attachments

New Guy and Exits

  • #1
  • First Post: Jun 9, 2008 4:25pm Jun 9, 2008 4:25pm
I just found this little gem of a site and wanted to introduce myself.

I've been working on an automated mechanical intraday trading system in C++ for three years now. I started with equities with no luck. Last year I modified my system to trade Forex.

In these last three years I've tested several entries. I have yet to find a pattern that is a good predictor of future direction. What I do is take an entry strategy, and test with a TP/SL of 10, 25 and 50 pips against the 21 pairs I track over 1 year of data. So far all systems have averaged 50% that the entry signal is right. Never has a single pair broken the 60% win/loss boundary.

Just yesterday I tested two entries highly recommended on this site. Both averaged 50%. I'll continue testing more from here, but at don't have much hope. If you do have an entry that "works", I'd love to test it.

As for what I'd really like from this site, is a study of exits. If my assumption is right that all entries in the short term are a coin flip as to the correct direction, then managing the exit is very important and I will contend more important than a good money management scheme.

I've never found trailing stops to work and have my doubts about a finding the right stop loss value. Currently I'm testing with a dynamic stop, one that takes into account volatility, and moves a take profit and a stop loss target in on the price, until the price crosses one or the other.

Please if you have any exit strategies, I'd like to test them. If there are yet undiscovered threads on the study of exits here, please point me in the right direction.

Thanks in advance,

Dan
.
  • #2
  • Jun 9, 2008 7:56pm Jun 9, 2008 7:56pm
If your entries are no better than a coin flip, there is no exit scheme or money management method that will make the systems profitable.

Over many trials, you will wind up with a loss, roughly, equaling the amount of your spread x number of trades.
  • #3
  • Jun 10, 2008 2:10am Jun 10, 2008 2:10am
I think the only really profitable eas are based on scalping. My trades have been 100% for the last two days with about 10 trades placed. Probably more luck than judgement but maybe not. These are based on price action. I don't think price action could be automated. Van tharp did a test based on a coin toss and using proper money management managed to profit, it is mentioned on this forum a couple of times and I started a post on it if you want to find it. However I still think determining the entry is the most important thing, (the pros will say otherwise - I am still demo trading !)

Good luck

max
Gravity goes both ways in forex...
  • #4
  • Jun 10, 2008 2:24am Jun 10, 2008 2:24am
Quoting billbss
If your entries are no better than a coin flip, there is no exit scheme or money management method that will make the systems profitable.

Over many trials, you will wind up with a loss, roughly, equaling the amount of your spread x number of trades.
I would like to find an entry that does better than a coin flip, but in my testing of several entry strategies, include a couple popular here, I've seen no better. To date the best strategy is take a regression line before the open of an exchange and follow it.

And actually with a dynamic exit, I've been able to get much higher win/loss rates, with some of the mentioned exits. Bounce in price will move most of your trades into positive territory, it's just a decision of when the trade is not moving and taking profit at one of those up points.

Dan
.
  • #5
  • Jun 10, 2008 2:30am Jun 10, 2008 2:30am
Quoting max pips
I think the only really profitable eas are based on scalping. My trades have been 100% for the last two days with about 10 trades placed. Probably more luck than judgement but maybe not. These are based on price action. I don't think price action could be automated. Van tharp did a test based on a coin toss and using proper money management managed to profit, it is mentioned on this forum a couple of times and I started a post on it if you want to find it. However I still think determining the entry is the most important thing, (the pros will say otherwise - I am still demo trading !)

Good luck

max
Your probably right automated traders. I know that I can trade better then my mechanical system, but who wants to sit in front of a screen all day.

I've been down the road of over confidence with my system too many times. A string of good trades is usually followed by one or two losses that wipe it all out. I once had a profit of 23% in one day only to lose it all the next.

There are a couple of books written too on the exit and how to improve the results, from entries that were a coin toss.

Dan
.
  • #6
  • Jun 10, 2008 2:52am Jun 10, 2008 2:52am
Quote
If your entries are no better than a coin flip, there is no exit scheme or money management method that will make the systems profitable.

That's not really true, tho. I'm profitable overall and my win/loss rate is only around 40% (less than (50/50, some days much less... lol). It's just that my losses are tiny (15 to 20 pips) and my gains are much bigger (50 to 150 pips).

I have never ran into a system that, over time, generates better than 50% odds. Not to say one doesn't exist, but in the studies I've seen win/loss percentage is inversely proportional to returns... ie most people that have higher win/loss ratios also, over time, have far lower returns. A lot of people chase perfection, forgetting that risk is commensurate with reward.

To that extent I think money management is crucial. I am able to get higher pips by watching for retraces and trailing my stop appropriately. I almost never set a fixed TP, altho if I do it's way the heck out there and unlikely to get hit. I trade for infinite expectancy, and have had intra-day trades turn into trades that last all week (not often tho, but it really rocks when it happens).

Quote
Van tharp did a test based on a coin toss and using proper money management managed to profit, it is mentioned on this forum a couple of times and I started a post on it if you want to find it.

I actually wrote an EA to simulate this theory... the basic coin toss entry system did horribly. It only pulled about 25% expectancy over 10 years data. The way I improved it was by tweaking exits, letting "profits run" and cutting losses very short. Basically that meant starting with a tiny stop loss, resetting lost trades, trailing winners over time and reversing around key points.

After that experiment I realized my profitability is less about entries and more about listening to the market. I redesigned my system around that theory. I've done much better since then.
  • #7
  • Jun 10, 2008 5:32am Jun 10, 2008 5:32am
Dan,

I think you are doing some really interesting work here. I do think that your testing methodology has some flaws and I wanted to offer some advice.

If you want to test the accuracy of the entry signal then you can't include a subjective exit. It is not fair to a system to impose stop losses. You didn't specify the time frames you tested but to be fair to the developer, you should test it on the time frame they suggested, not your own. Even on a 15 minute chart, stops of 5,10,15, or 20 pips is largely inappropriate since those values generally lie well within the "fog of randomness" of about 3ATR.

To be totally accurate you should test with a timed exit only. You can try 5 candles, 10 candles, 15 candles, and 20 candles. This takes out all other variables. At the exit the trade is either a winner or loser. While I think this is much more objective, it is likely that you still won't find an accuracy much over 50%. It's interesting to read about all the hot systems on here that in actuallity, are not as accurate as a coin flip. Until you can develop and edge with an entry you shouldn't bother with adding an exit. To be fair, there are profitable systems with entry accuaracies less than 50%. For example, moving average crossover systems rarely exceed 35%.

By the way, there are systems that have an entry accuracy that exceeds 60%.
  • #8
  • Jun 10, 2008 7:57am Jun 10, 2008 7:57am
I think it would have to be ATR based.

max
Gravity goes both ways in forex...
  • #9
  • Jun 10, 2008 8:17am Jun 10, 2008 8:17am
Quoting max pips
I think it would have to be ATR based.

max
What would have to be ATR based?
  • #10
  • Jun 10, 2008 3:52pm Jun 10, 2008 3:52pm
Quoting philmcgrew
Dan,

I think you are doing some really interesting work here. I do think that your testing methodology has some flaws and I wanted to offer some advice.

If you want to test the accuracy of the entry signal then you can't include a subjective exit. It is not fair to a system to impose stop losses. You didn't specify the time frames you tested but to be fair to the developer, you should test it on the time frame they suggested, not your own. Even on a 15 minute chart, stops of 5,10,15, or 20 pips is largely inappropriate since those values generally lie well within the "fog of randomness" of about 3ATR.

To be totally accurate you should test with a timed exit only. You can try 5 candles, 10 candles, 15 candles, and 20 candles. This takes out all other variables. At the exit the trade is either a winner or loser. While I think this is much more objective, it is likely that you still won't find an accuracy much over 50%. It's interesting to read about all the hot systems on here that in actuallity, are not as accurate as a coin flip. Until you can develop and edge with an entry you shouldn't bother with adding an exit. To be fair, there are profitable systems with entry accuaracies less than 50%. For example, moving average crossover systems rarely exceed 35%.

By the way, there are systems that have an entry accuracy that exceeds 60%.
That's fair point and another good way to approach it.

I've been testing on the time frame suggested by the author. I'm not trying to invalidate any system, I'm just looking at what the percentage of correct predictions are. I also want to know how to maximize taking profit.

For my own system I currently use 15 minute candles, with a timed exit of four hours. Depending on volatility, the stop loss and take profit start around 100 pips away and close in on the price based upon movement.

My trend line entry has a win/loss of 57%, but I cheat by cherry picking the best performers and using really restrictive entry rules. When I add my dynamic stop to the system, it bumps up to 67%. I've taken this system from backtesting to demo testing and I'm up 8% this week. Even with those results, I think a lot more can be squeezed out from an intelligent exit.

My point is not to discorage any choice of entry, but to understand how to make an exit that works well with most entries. How do authors of these systems produce 80% winners(their claim) from systems which if using an equal distant TP/SL, score 50%. They obviously know how to take profit? What is it they do? I'm making a big assumption, but I expect an exit that makes money with a coin flip entry will do much better with 60% win/loss entry.

BTW, if you do know of a system that has a 60% accuracy, could you please point me in the right direction?

Dan
.
  • #11
  • Jun 10, 2008 4:37pm Jun 10, 2008 4:37pm
Dan, a stop loss is an exit and therefore influences the entry. The only true way to test the accuracy of an entry system is to have a timed exit.

If you want a high accuracy system you only have to widen the stops and reduce the targets. System sellers do it all the time to lure in people who think that you can buy groceries with accuracy rather than pips.
  • #12
  • Jun 10, 2008 4:44pm Jun 10, 2008 4:44pm
Quoting philmcgrew
Dan, a stop loss is an exit and therefore influences the entry. The only true way to test the accuracy of an entry system is to have a timed exit.

If you want a high accuracy system you only have to widen the stops and reduce the targets. System sellers do it all the time to lure in people who think that you can buy groceries with accuracy rather than pips.
I think a test based on the average true range would also be accurate. Have a read through this because I can't explain it

http://www.investopedia.com/articles/trading/08/ATR.asp

Max
Gravity goes both ways in forex...
  • #13
  • Edited 6:15pm Jun 10, 2008 5:12pm | Edited 6:15pm
Quoting philmcgrew
Dan, a stop loss is an exit and therefore influences the entry. The only true way to test the accuracy of an entry system is to have a timed exit.

If you want a high accuracy system you only have to widen the stops and reduce the targets. System sellers do it all the time to lure in people who think that you can buy groceries with accuracy rather than pips.
Phil,

I'm just running simulations and exiting on an equal distant point. But since you're so insistent here are the results.

The tests are run with 21 pairs, approximately one years worth of data. This was against a 9/20 EMA entry system.

1. Time exit 5 bars.
Total Trades: 5,665
Winning Trades: 2,925
Win/Loss: 52%
Best Pair AUD/NZD - 64%
Worst Pair USD/CAD - 40%
Standard Deviation: .0488

2. Time exit 10 bars.
Total Trades: 4,703
Winning Trades: 2,504
Win/Loss: 53%
Best Pair AUD/NZD - 80%
Worst Pair USD/CAD - 43%
Standard Deviation: .080947

3. Time exit 15 bars.
Total Trades: 4,349
Winning Trades: 2,342
Win/Loss: 54%
Best Pair AUD/NZD - 81%
Worst Pair USD/CAD - 44%
Standard Deviation: .093702

4. Time exit 20 bars.
Total Trades: 4,186
Winning Trades: 2,253
Win/Loss: 54%
Best Pair AUD/NZD - 81%
Worst Pair USD/CAD - 42%
Standard Deviation: .093724

Not sure how to interput these results. I've noticed in all my testing AUD/NZD is a big winner. Also, notice that as the accuracy goes up, so does the variablity.

Good idea! Something else to think about.

Dan

P.S. I think I just attached the detail results.
Attached File(s)
File Type: zip CurrentSettingsScored.timeExit.zip   32 KB | 457 downloads
.
  • #14
  • Jun 10, 2008 5:41pm Jun 10, 2008 5:41pm
Quoting max pips
I think a test based on the average true range would also be accurate. Have a read through this because I can't explain it

http://www.investopedia.com/articles/trading/08/ATR.asp

Max

I've actually tested using equal distance from 1 ATR, 2 ATR and lots of ATRs, with no better results, however I do use ATR in my dynamic exit to caluculate the starting place for my staring SL/TP levels. The idea is then to ratchet the levels in on the price based upon its movement in price from opening the position.

Dan
.
  • #15
  • Jun 13, 2008 5:39am Jun 13, 2008 5:39am
Quoting philmcgrew
To be fair, there are profitable systems with entry accuaracies less than 50%. For example, moving average crossover systems rarely exceed 35%.

By the way, there are systems that have an entry accuracy that exceeds 60%.
but moving average systems can be probitable?

and what about this system with 60% accuracy, this is public available? Or is in any book?

sorry for my bad English.

best regards.
  • #16
  • Jun 13, 2008 7:08am Jun 13, 2008 7:08am
Quoting giraia_br
but moving average systems can be probitable?

and what about this system with 60% accuracy, this is public available? Or is in any book?

sorry for my bad English.

best regards.
If you can handle only being right 1/3 of the time, moving average crossover systems can be profitable.

If you want 60% accuracy, trade with the trend. The point of this thread is to point out that you should be focusing on exits, not entries.
  • #17
  • Jun 13, 2008 11:48am Jun 13, 2008 11:48am
thanks for answer.

best regards.
  • #18
  • Jun 14, 2008 8:22am Jun 14, 2008 8:22am
This weeks testing, where my emphasis is on the exit and money management.


Total Trades: 41
Winning Trades: 21
Losing Trade: 18
Total Lots Trade: 149
Per Win Trade: $115.85
Per Losing trade: $64.45
Starting Account Balance: $7,504.00
Ending Account Balance: $8,542.10
Total Profit: $1,038.10

What's your opinion, good bad or indifferent?

Dan
.
  • #19
  • Edited 1:31pm Jun 14, 2008 1:16pm | Edited 1:31pm
Quoting redpine
This weeks testing, where my emphasis is on the exit and money management.


Total Trades: 41
Winning Trades: 21
Losing Trade: 18
Total Lots Trade: 149
Per Win Trade: $115.85
Per Losing trade: $64.45
Starting Account Balance: $7,504.00
Ending Account Balance: $8,542.10
Total Profit: $1,038.10

What's your opinion, good bad or indifferent?

Dan
Hi Dan,

Obviously the result is outstanding (win rate > 50 and RR = 1.8), but IMHO the sample size (41 trades) is too small to have any real statistical meaning.

Phil (philmcgrew) has done more testing than I have. I have reached similar conclusions to his.

To test entries objectively, IMHO you must use timed exits. I plot a frequency distribution of where price was at the close of M1 candles, up to X candles away from entry, using different values for X.

To answer your question about exits, I see at least four basic approaches:

1. Trailing stoploss. The more the maxim "let profits run" improves profitability, the wider the trailing stop can afford to be, to allow maximum room. However, obviously it can’t be infinite, hence there has to be some kind of "sweet spot" area. Using ATR multiples is probably as good a method as any, since it adjusts itself according to volatility, although you might also like to try something similar to Parabolic SAR, which tends to "close in" on price the further it moves, as a possible foil.

2. Profit targets at places of probable reversal (e.g. areas of prior S/R, Fibo retracement levels, floor trader pivots, round numbers). However, some significant problems exist: (1) these can be difficult to establish objectively "codeable" rules for (e.g. from which H/L should one extend the Fibos?), and (2) the coding of "visual" concepts like S/R can become very complex.

3. Profit targets based on a frequency distribution of "significant" moves.

4. Using a fixed SL, and simply exiting when an opposite entry signal occurs. This assumes a "stop and reverse" type system.

Assuming the goal is to maximize profit, I don’t see any point in attempting to scale out gradually by combining a trailing SL with profit targets. If "letting profits run" offers significantly improved profit, on balance, then it makes sense to always let the whole position run. (However, if the goal is to smooth income, then scaling out may be preferable).

One thing that is certain about "profit exits": every approach involves a compromise between RR and win rate, which operate in inverse proportion to each other. Exit contributes to expectancy only by whatever extent it’s able to overcome this inherent balance.

I don’t want to get into a debate on semantics, especially as this topic’s been done to death elsewhere, but I believe that entry is at least as important as exit, and that those who think otherwise are taking entry for granted. Let’s just say that if a significant edge could be obtained by "enter any emerging move, cut losses quickly, and let profits run", then any MA crossover method with a tight protective SL would suffice, and writing a profitable EA would be incredibly simple. To whatever extent that’s not the case, entry must be important. The key, as I see it, is that entry and exit complement each other, depending on what aspect of price phenomena (e.g. trend-following; breakout from congestion; fading from OB/OS extremes) the system is trying to capture.

A "move SL to breakeven" strategy is probably worthy of testing. Having said that, I suspect that many traders move their stop quickly, for reasons of psychological comfort, as opposed to mathematical expectancy (i.e. revisiting the compromise between RR and win rate). However, when all wins and losses are summed, each pip of profit that’s locked in is no more significant than each pip of loss that’s locked out. Assuming that the market doesn’t move in an extraordinary fashion just because we happen to have entries and exits in a certain area, then I can’t see any justification for a "sudden leaping" of our SL at any one specific point, more so than at any other.

David
  • #20
  • Edited Jun 15, 2008 9:20am Jun 14, 2008 3:16pm | Edited Jun 15, 2008 9:20am
Quoting hanover
Hi Dan,

Obviously the result is outstanding (win rate > 50 and RR = 1.8), but IMHO the sample size (41 trades) is too small to have any real statistical meaning.

Phil (philmcgrew) has done more testing than I have. I have reached similar conclusions to his.

To test entries objectively, IMHO you must use timed exits. I plot a frequency distribution of where price was at the close of M1 candles, up to X candles away from entry, using different values for X.

To answer your question about exits, I see at least four basic approaches:

1. Trailing stoploss. The more the maxim "let profits run" improves profitability, the wider the trailing stop can afford to be, to allow maximum room. However, obviously it can’t be infinite, hence there has to be some kind of "sweet spot" area. Using ATR multiples is probably as good a method as any, since it adjusts itself according to volatility, although you might also like to try something similar to Parabolic SAR, which tends to "close in" on price the further it moves, as a possible foil.

2. Profit targets at places of probable reversal (e.g. areas of prior S/R, Fibo retracement levels, floor trader pivots, round numbers). However, some significant problems exist: (1) these can be difficult to establish objectively "codeable" rules for (e.g. from which H/L should one extend the Fibos?), and (2) the coding of "visual" concepts like S/R can become very complex.

3. Profit targets based on a frequency distribution of "significant" moves.

4. Using a fixed SL, and simply exiting when an opposite entry signal occurs. This assumes a "stop and reverse" type system.

Assuming the goal is to maximize profit, I don’t see any point in attempting to scale out gradually by combining a trailing SL with profit targets. If "letting profits run" offers significantly improved profit, on balance, then it makes sense to always let the whole position run. (However, if the goal is to smooth income, then scaling out may be preferable).

One thing that is certain about "profit exits": every approach involves a compromise between RR and win rate, which operate in inverse proportion to each other. Exit contributes to expectancy only by whatever extent it’s able to overcome this inherent balance.

I don’t want to get into a debate on semantics, especially as this topic’s been done to death elsewhere, but I believe that entry is at least as important as exit, and that those who think otherwise are taking entry for granted. Let’s just say that if a significant edge could be obtained by "enter any emerging move, cut losses quickly, and let profits run", then any MA crossover method with a tight protective SL would suffice, and writing a profitable EA would be incredibly simple. To whatever extent that’s not the case, entry must be important. The key, as I see it, is that entry and exit complement each other, depending on what aspect of price phenomena (e.g. trend-following; breakout from congestion; fading from OB/OS extremes) the system is trying to capture.

A "move SL to breakeven" strategy is probably worthy of testing. Having said that, I suspect that many traders move their stop quickly, for reasons of psychological comfort, as opposed to mathematical expectancy (i.e. revisiting the compromise between RR and win rate). However, when all wins and losses are summed, each pip of profit that’s locked in is no more significant than each pip of loss that’s locked out. Assuming that the market doesn’t move in an extraordinary fashion just because we happen to have entries and exits in a certain area, then I can’t see any justification for a "sudden leaping" of our SL at any one specific point, more so than at any other.

David
David,

Ya I know the sample is to small. Also, it is out of wack with my backtesting which had smaller profits, but bigger win/loss ratio. I just wanted to post the results of what I'm doing, for discussion.

As for trailing stops, the raw ones offered by most brokers, test with poor results. Right now I use a multiplier of ATR to set initial Stop Loss and Take Profit levels, then another fraction of ATR to close in on the price, if it's going straight up the TP target doesn't move. I'm interested in trying Parabolic SAR as an exit strategy, and just haven't got around to it yet. I know my system is suboptimal, but is better than most basic approaches. Right now, I'd just like to be in the game and in the game in a positive way.

I've tried your option 4 with not very good results. Option 3 sounds interesting since we are talking about a probabilistic system.

A guiding principle of my trading in Equities and Forex, is to reduce exposure.
I want to get in and out of a trade quick with a small profit. That is my comfort zone. With an automated system I want that to be in the matter of hours. If I never find such a profitable scalping system then I can personnaly manage the trades over a longer period and feel confident of winning.

Anyway, I hope I didn't say that entries are not important, if I did, then I got caught up in typing my thoughts. What I was trying to say is that if you look at most of the entries I've tested, there usually is an equal chance of a short term movement being against the trade as there is with it.

As for Phil's comments, I do think that timing an exit can impact profit, I just don't use it as my measure of the probability of an entry strategies accuracy. Also, I want to note that my definition of a robust entry strategy is one that works for many pairs and not just one or two. I believe that such a strategy is more likely to hold up over time.

That being said, so far this weekend I have been testing my system with different timed exits. The longer I go the better the win/loss, but the lower the profit factor and the higher the Max drawdown. So the sweet spot for my system seems to be around 4 hours. Time held is an important facet of a trading system.

Also, if someone has an entry that has a 60% or greater probability(below 40% would also be of interest) of going with the trade as measured by pips or time, I would love to test it and share my results. To date my best efforts are a pattern based analysis and trend line entry which has got me to 57%. Actually that seems to be the number for me to beat.

Dan

P.S. I just exchanged emails from a freind who follows your thoughts. I think we are probably in violent agreement as to what's important.
.
0 traders viewing now