Gaining an edge from your exit

Hello,

A while ago I posted a thread on this site about gaining and measuring the edge from a trading system's entry and got a lot of good response and helpful feedback.

So I thought I'd see if anyone wanted to discuss ways to gain an edge from a mechanical exit?

I have put one of these exits on my site/blog http://www.myforexdot.org.uk/SystemV.html the exit's edge is exiting at a favourable price, in this case exiting when the price closes at a new 3 day closing high when you are long and exiting at new 3 day closing low when you are short Vs. a time-based exit.

Other exits I have compared are -

1. Waiting for a trend reversal signal - works well in long trends but involves giving back a lot of profit as you wait for a signal that the market has finally turned and gone against you. Also produces more losing trades than winning ones.

2. A time-based exit. Usually produces slightly more winners than losers but doesn't make the most of long trends.

3. A lack of follow through. A compromise between exits 1 & 2. We might exit if we were long when the market failed to make a new high for 2 weeks etc... Doesn't need a full blow reversal to get out, just a sign that it's not quite going our way any more.

4. Taking a favourable price when you have it. I.e. the market makes a new high when you are long.


What are your exit strategies and which one produces the best results for you? And how have you measured these results?

This thread was created more than 7 days ago, and yet, despite everything that I read about exits being more important than entries, apparently nobody here is brave enough to offer comment. So I will be the first to give it a try.

Despite the obvious cognitive distinction between 'entries' and 'exits', I don't see any mathematical difference between them. Both are merely vehicles to position a trader by increasing or decreasing his directional exposure to subsequent price movement. Market behavior is not going to change radically because of the net exposure of one insignificant retail trader. Hence, to whatever extent profitable 'entries' involve exploiting inefficiencies – or at least semi-predictable idiosyncrasies in price behavior – then profitable 'exits' must necessarily do likewise.

Profitable trading equates to being net long while price is rising, and net short when it's falling, frequently enough to overcome costs. The bottom line is that the more accurately we adjust our exposure ('enter' and 'exit') relative to price reversals, on average, the more pips we will ultimately bank. The P/L of our current 'trade' – or indeed, our entire approach toward managing risk – has no bearing on probable market behavior, and as such can offer nothing in the way of an edge.

Reverting to a less theoretical line of thought, I'm familiar with the 'cut losses quickly; let profits run; add to winning positions' strategy employed by the likes of Richard Dennis' turtles. This strategy is profitable to whatever extent markets 'trend': how far, and how often (Thursday was a great case in point, LOL). Black swans aside, however, a major key to improving the probabilities is to trade the strongest trending markets (pairs, timeframes), on the assumption that their moves are the most likely to continue. (In this case, the market's 'trendiness' is the exploitable inefficiency). Personally, I use a simple ATR-based trailing SL, which is in keeping with the OP's point #1: the expectation is that pips forfeited as price reverses to the SL will be more than compensated for, on balance, by the pips harvested by simply leaving the position open.

Point #2 (time-based exits) these can provide an edge only to the extent that they have demonstrable statistical backing, e.g, in a certain scenario, the 'average' length of a move is X candles. And of course time of day can have big bearing on volatility, a potentially significant consideration. But in all other respects, time-based exits are little more than an arbitrary means of exit.

Point #3 (technical exits) and point #4 ("taking a favorable price when you have it"). In trading the M15/M30 charts, I find that long candle (preferably backed by high tick volume) frequently results in the exhaustion point of a 'surge', and given that this complies with both #3 and #4, I'll look to scale out as a 'hedge' against my trailing SL. I'm not sure to what extent this provides an edge, but it helps to smooth out my P/L. How much I scale out depends on my assessment of the strength of the prevailing move.

Of course everybody trades differently, and the above is particular to my own strategy. Speaking more generally, it's my view that a good exit methodology is one that (approximately) fits the sizes of the moves one is attempting to capture, complements the setups being traded, and is compatible with the strategy as a whole (in my case, attempting to seek out the biggest moves).

@Davidee: I must admit that (and perhaps it's partly because I'm a discretionary trader) I haven't specifically measured the results of my exits. I read your previous post on measuring entry efficiency by comparing favorable and adverse price excursions from the entry point. On that basis, the efficiency of exits must necessarily be the ratio between overall profit factor, i.e. total system efficiency, and the efficiency of the entries. So, for example, if the system PF is 1.5, and the average MFE/MAE of the entries is 2.0, then the exit efficiency would be 1.5 / 2.0 = 0.75. (In this example, the exit method is actually degrading overall performance).

IMO this is superior to simply making the assumption (as some apparently do) that win rate is a direct measure of entry performance, and average win size is a direct measurement of exit. As an example of how this is flawed, obviously exits have a bearing on win rate: everything else being equal, a tighter initial SL will decrease win rate.

I saw this thread, and was expecting replies to study (thanks as usual Hanover), ill take a go:

Im a firm believer in your 'entry' being your 'exit'. Actually keeping in mind what the previous poster has noted, your exit cannot provide you an edge in itself (am i correct?), so if this is one floating P/L graph, where you are trying to be long when the market is long and vice versa, your exit and your entry are interchangeable. (*I am making an assumption, and let me clarify, i also, believe no money management system can enhance the existing edge or create an edge where none exists). Why not use the same criteria to exit (and reverse occassionally), assuming your system setup is +ev, the edge should reply in reverse.

Also a firm believer in not having price targets (rather let the market take you out), if i am following the trend purely. Whatever criteria caused you to enter when present against you (or some obvious S/R), just tighten your stops (possibly below the last swing low etc).

Of course having obvious technical targets works too, but the point being, there has to be a logical reason. For ex if you are long into a nice leveraged position risk free on a positive carry (AUD/USD), im sure you are in no hurry to let go of the juice you make every day, just yet. This is where an ATR stop does the job too (1ATR,2ATR, 3ATR). This belief structure should also coincide with how trend traders tend to think (far rightfield opinion fwiw). The reason I am concentrating on the trend here is because OP's orignal system in the past thread on 'edges' was a trending one.

Well these firm beliefs will be put to test very soon .

Chuck Lebeau gives a tip, which makes sense to me (if we were looking to quantify the efficiency of exits). He says, if you compare your profit to the max profit you would have made if you had held the trade twice as long, and you get a figure less then 50%, your exit strategy needs tinkering.

I personally think, holding period returns and hence 'time' is also a factor worth considering (making efficient use of capital on hand).

I do not have any math data on this, but will contribute the following...

The biggest impact on my trading came from truly understanding how important it is to know the "mark-up" percentage of your business. If you ask a traditional business owner what their profit margin per product or service offering is, they'll usually be able to tell you. So the hot dog vendor that buys his hot dogs for $ 1.00, and then sells them @ $1.50 is operating on a 50% profit margin (or mark-up). However, when you ask a trader what their profit margin is... well you know how that goes. "I let the profits run", or the usual, "I let the market dictate what it gives me".

When I buy my hot dog for $1.00, I know what I'm going to be selling it for. There's none of that traditional trader ideology about exits. For me that way of thinking was a losing proposition. Fixed TP levels. Period. No trailing stops. No set to BE. I either take a profit, or a loss. Coincidently, if I can't find someone to pay my asking price that day, and the hot dogs spoil... then so be it. I'll buy fresh ones tomorrow.

It worked wonders for me. If nothing else it took the dark cloud off this whole "exits" mystery.

thanks for sharing this

So it sounds like you're going for a variation option 4 then? Taking a favourable price if/when you have it. You accept a price you deem as favourable (your asking price) the moment the market offers it.





Yes, thanks That's something I'll be looking into carefully...

It probably does provide an edge if it's done right as it's just another method of taking advantage of a favourable price when it's there.

I guess it could be tested easily enough though by defining exactly how long the candle had to be and then comparing the favourable and unfavourable price movements after x amount of time of a candle that met the criteria appearing.

Exits -

While many traders enter too fast and impulsive, they are confused by immediately following price action in the decision making process for exits.

When you exit a trade, there is no way of knowing if it was the best time, too early or too late (the same goes for entries in fact, but we make an assumption that the entry is based on solid research into the to be exploited phenomenon).

My view is to reverse this: you [the trader] dictates the entry - which should be well thought through and have a solid reasoning, while the market [price level] will ultimately dictate the exit.

You can control where a stop or a profit target should be - but make no mistake, it is not in your control whether the price will reach one or the other first, making it a profitable or losing trade.

Its a bet.

You bet the market will rise. You know where you take a loss if the price will instead fall. If the market will rise, it can do so fast, slow, fast a little then reverse slowly but staying above your stop, rise slowly then reverse fast but still not hitting your stop, rise fast and hit your target within a very short time, rise slowly to hit your target after quite some time.....and so on and so forth - the possibilities are infinite and every time a little or a lot different.

All these possible combinations have an influence on most traders in terms of their decision to either stay with a trade or abandon it (I should add that abandoning it can be positive in some cases while in others it can have a negative outcome).

Here lies the problem: TIME

Traders need to take into account that there is the possibility of interferences with the time perspective once they are in a trade. Being in profit has a time compressing effect in that the average trader looks at the price (almost trying to hypnotize it) wanting it to move in fast forward to the desired profit target.

When in negative territory (but not stopped out) time makes many traders feel the agony and everything seemingly slows down. That often has the result that IF the price comes back to break-even or a showing a small profit, the 'feeling' of the elapsed time seems so overwhelming that trades are closed out based on the huge relief that is to follow first - then often comes the regret for not sticking with while the price keeps on moving in the originally anticipated direction.

So, whatever exits are chosen, it will be imperative to investigate the average length for the price to either reach target or stop. When using timed exits, it has to be investigated that in fact using them will have a statistically positive expectancy (as already explained by Hanover).

Relying on past data (even if it is a large sample and statistically significant) still bears the possibility of future results deviating from the past positive expectancy.

Ever watched a school of sardines in the ocean? Very difficult to mathematically constrain the possible speed, movement and speed of movement.

Sorry for the lengthy post - hope it will give you something to thing about.

regards
daytrading

Thanks for your thoughts daytrading. Merging targets with time is something I've tried but I haven't (yet) written the results up to put on my blog.

One exit method that worked well on long term trend following systems was waiting ten days after an entry then, after at least ten days had passed taking the first 5 day high as a exit when the trade was long and the first 5 day low after a waiting period of ten days when short.

Your method of looking at the average time is something I'd also like to test, with targets, i.e. if the target isn't hit in x number of time assume that it's not likely to be and just exit taking whatever profit or loss one has at that point. This is something I'll have to test.

Thanks for that

I'm personally using the same as hanover- ATR trailing stop. I trade on longer time frames so it is not very prone to noise. If it hits, too bad. If it doesn't hit, just wait until it does or an exit signal shows up. I generally make more than I lose using this method.

However I've been demoing shorter time frames lately and found that doing method 3 (exit when it doesn't quite go your way) produces much better results. For example, I would pick an entry based on 4 hour TF and and exit based on the 1 hour charts or enter based on 1hour charts and exit based on 15 mins charts.

Since I'm still using the same ATR trailing method on shorter TF, you tend to make less when you are right and get stopped out more often. I find the need to take profits more often because prices whipsaws more often. I move stops to breakeven after a good rally because I know that prices can reverse in less than an hour. I can be up 80 pips now and be -40 pips 2 hours from now. Profits are harder to protect on shorter time frames simply because the percentage move is much much larger than on longer time frames. This is never a problem on daily charts.

Having said that, I'm not sure if this is the best way on shorter TFs. I know it works well for daily. Maybe you can come up with some test results I look forward to it.

I find whole number s/r levels perfect for exits. On any timeframe h1 or higher you'll see price action tends to stop/reverse or break through whole number levels. Anything ending in 00 is where I look to exit/take profit.

It is a known entry trades does not matter, its the exit that determine a trader profitability.

I never tried time based exits, but I have one rule that says that if a new position doesn't do its job quickly I look to fire it (exit) without hesitation. See below.

I'm a trend trader and try to get as much out of the trend. My entry is discretionary, PA based, but the exit is completely mechanical.

My exit strategy is composed of 2 trailing strategies, consider a long position:

1. After entry I trail behind each bar low until I get to BE. The idea behind this aggressive trail is to quickly fire any position that doesn't do its job quickly & properly - that is taking off as soon as I get in. This way I rarely have full 1R losses.
2. After I reach BE I allow more room for price and use a simple volatility trailing stop based on ATR.

Doing well so far.

Can you explain this further? Are you saying that we might as well toss a coin in deciding whether to go long or short, and hence that all charts and analysis are a total waste of time?

Can I ask what settings seem to work well on ATR stops? I trade J16 Pinbars using just daily set-ups.

If someone can give me some idea, based on their experience, as to which settings suit daily set-ups I would be very grateful.

Thanks

Glynn

I think that's just a myth, or at the very least not 100% correct. Entries can have an edge and therefore do (at least to some extend) matter.

Don't get me wrong, I don't doubt that it's possible to have a profitable trading system that uses random entries and an exit that provides an edge (in this case the exit would most probably be closing the trade quickly if it goes against you and letting the winners run), but entries do matter. Entries can provide an edge, I know they can as I've tested a number of entries with a time-based exits and some of them can provide a significant edge, which random entries wouldn't do with a time based exit.

You don't need to have an entry with an edge to have a profitable trading system, but in my opinion it's probably best to take advantage of every edge you possibly can.

Unfortunately I have not done any testing beyond my own strategy. You will have to find out thta number yourself.

For me it is 2 atr on the daily charts. Period of 20 days back. Anywhere between 14-20 days shouldn't give too much difference. But preferably you do not use anything less than 10 days as it is too prone to noise.

What you will want to see is when the price is trending, you do not want to be stopped out. There will be occasional price movement that breaks your 2ATR. That is normal, however if it happens 2 or 3 times in a trend that means your settings are too tight. I would suggest testing it on EURUSD because they trend the strongest.

Really helpful. Many thanks.

I've been thinking quite a bit about different order types lately. Rather than post these in a new thread, I thought I'd add to the really good commentary in this thread. I think most traders are aware of the basic order types, but it's good to remember that the type of orders used to manage open positions can determine or add to whether a strategy is a loser, profitable or even highly profitable, given the same entry under multiple test conditions. And optimal will vary with the time frame and length of holding period.

The basic ones everyone knows about are:

Take Profit or profit target - a fixed price profit level.

Stop Loss - a fixed price loss level

Breakeven Stop - Once the market goes so far into profit, this stop ensures that the position will break even at worst.

Trail Stop - There are a couple of different flavors. The simplest is the point or pip trail stop. You pick a number of pips to trail the position (like 30), and then when price pulls back 30 pips your position is stopped out. This can also be implemented with a floor or a requirement that the market has to go a certain distance in your favor before the trail stop kicks in - that certain amount is the floor.

You can do the same thing with percentages instead of fixed pip amounts. For instance as a percentage of maximum positive equity for a trail stop, when price pulls back a certain percentage from maximum open position equity, you stop out. This type must be used with a floor to prevent instant stop outs.

Some order types that are talked about just a bit and can be useful, though you probably have to custom program them:

Time orders:
Time stop - exit after a certain period of time if position is in loss.
Time target - exit after a certain period of time if position is in profit.
Time exit - exit after a certain period of time, whether profit or loss.

Trailing Target: Everyone has heard about a trailing stop but I've never heard anyone talk about a trailing target. Perhaps I should patent the term lol. A trailing stop works on a long position by moving the stop up as the market makes new highs - keeping the trailing stop a certain distance below the market - when it pulls back the distance of your trailing stop, your position stops out.

With a trailing stop for a long position you're comparing the maximum price with your trail stop amount. With a trailing target you compare the largest pullback with your trailing amount. Let's say you pick a trailing target of 30 pips. If price goes straight to your target, you make 30 pips. You effectively place a profit target 30 pips above the lowest pullback. If price goes up 10 pips then down 30 pips your trailing target will move down to breakeven on a long position. If price goes up 10 pips then down 20 pips, your trailing target will be at +10 pips.

A couple of new takes on the standard order types can be achieved by using a time element combined with a take profit or stop loss. You place a timed stop loss at -10 pips with a 2 minute time limit. What happens is that price doesn't stop you out immediately when it reaches -10 pips - it starts the clock ticking. If 2 minutes later the price is -10 pips or worse, you stop out, otherwise you reset the clock to zero. Same thing in reverse for a take profit order. Rather than immediately exiting when the +10 profit target is hit, the clock starts and as soon as the time is up and you are still in profit, you exit the position. Another good way to use this to let profits run for take profits is to reset the clock whenever a new equity high is made. So your 2 minute clock could reset indefinitely as long as price keeps making a new equity high, allowing profits to run, but it exits after price stalls and hopefully before the lion's share of your profit deteriorates.

These and other position management tricks are ways to add a bit of advantage to a trading strategy. I hope this has spurred some thought to think outside the box of the usual order types and illustrates how it is possible to think of new ways to manage positions using traditional stop and profit targets that might help improve the bottom line.

You need to do your own testing to see if the settings meet your trend following expectations. The larger the multiplication factor the longer you can ride trends but you will also give back a hefty amount of pips when the market turns against you.

In my testing I found out that using a factor of 2 allows you to ride swings of trend while a factor above 3, more like in the range of 4 and 5, allows you to run big trends (multi month) when you catch one.

I personally use a factor of 2, look to catch swings of trend that last from a few days to a few weeks, but you have to do your own testing.