I have been trading an equity trend following system for ~3 years now with a fair degree of success. My system enters when a strong trend is in place which has a short pull back before continuing in the direction of the trend (not rocket science and not much optimisation involved). My exits are based on a wide volatility trailing stop which keeps me in winning trades for up to 18 months (average ~10 months)
I am struggling with two issues and would appreciate any guidance you may have:
1. What is the best approach to take when i have allocated all of my account to trades but still keep getting new signals? Literature on trend following (primarily futures focused) indicates that you should take EVERY signal to avoid missing that one big trend that might make your year's profit...but this does not seem to be possible with equity trend following as you can generate so many signals. I have considered applying various filters to reduce the number of trades to a reasonable level, but this just feels like curve fitting.
2. How can i improve the capital efficiency of my system - since i am using a wide trailing stop, I can often sit in trades for weeks / months that go nowhere but don't get an exit signal because my trailing stop has not been hit. I have experimented with combined trailing stop / indicator exits but can't seem to improve the basic system statistics compared to a simple wide volatility based trail exit. The system works very well with the wide trailing stop...but i would just like to improve it be capturing more of the upside in some cases rather than waiting for the price to reverse and move all the way back to my trailing stop before exiting.
Any help or suggestions would be greatly appreciated.
Equities trend following - Take all signals?
Re: Equities trend following - Take all signals?
I don't know what the BEST approach is, but here are some thoughts.
You probably have a history of which trades paid off the most, do some data mining and see what they have in common, maybe you can distill that to see which of your signals have the most "giddyupedness" and stick with those. Alternately, perhaps some measure of momentum should be the filter between your signals taken and signals ignored.
To some extent, your answer to the first question may get you out of those sidelined-but-not-stopped-out stocks.
Perhaps narrow the range of stocks that are subject to systems trades. Maybe some fundamental metrics on the underlying companies? Maybe trade only some small group of stocks (Dow 30? Nasdaq 100) for systems?
You probably have a history of which trades paid off the most, do some data mining and see what they have in common, maybe you can distill that to see which of your signals have the most "giddyupedness" and stick with those. Alternately, perhaps some measure of momentum should be the filter between your signals taken and signals ignored.
To some extent, your answer to the first question may get you out of those sidelined-but-not-stopped-out stocks.
Perhaps narrow the range of stocks that are subject to systems trades. Maybe some fundamental metrics on the underlying companies? Maybe trade only some small group of stocks (Dow 30? Nasdaq 100) for systems?
Futures guy with scant stock trading experience, spews random halfbaked ideas willy nilly:
1. When fully loaded and you get a new signal, choose a "Victim" from among your existing positions and exit the Victim trade. Victim might be (a) most profitable trade, (b) least profitable trade, (c) oldest trade, (d) trade with least-promising indicator_value (relative strength? Rate of change? Linear Regression Slope?) (e) whichever existing trade is in a stock (or industry group) that is most closely correlated to the stock of the new signal, (f) divide the existing trades into the top-half and bottom-half according to some indicator, then pick a Victim by random choice among the bottom-half. Coin flipping. Caprice. Nondeterminism. Use the cash freed up by selling the Victim, to enter the new trade.
2. Exit a small piece of each existing position to free up enough cash to enter the new trade. For example, if your total portfolio value is $56,000 and you have 7 stocks, sell $1000 worth of each of those stocks. Now you've got $49,000 worth of stock and $7,000 cash. Use the $7,000 cash to buy the new stock. Or instead of selling equal dollar amounts of each position, you could sell off equal percentages (sell (1/(7+1)) = 12.5% of each position, and use the new cash to enter the new stock).
3. Use the occasion to "rebalance" the portfolio. Sell or buy just enough of each one of the existing N stocks, that each one will represent (1/(N+1)) of the total account value. Use the new cash to take the new signal in the new stock.
4. The "Manana Principle": Temporarily go on margin to enter the new position, with the opinion that you'll very likely get an exit in another stock pretty soon. If you wait a while and still haven't gotten an exit and are still on margin, sell off a piece of each position (or pick a Victim, or ...)
5. Run backtests to find out what would have been the maximum number of simultaneous positions in the past, if you took all signals. For sake of discussion, say this turns out to be 60 simultaneous positions. Use this number to help you choose a positionsize (let's say you choose: 1/50th = 2.0% of the account) and then live with it. If you ever get more than 50 simultaneous positions such that you have no cash to enter new signals, live with it. Skip those trades and don't let them worry you. Those are somebody else's trades, they aren't your trades.
Trying out some or all of these ideas in backtesting, may take some time.
1. When fully loaded and you get a new signal, choose a "Victim" from among your existing positions and exit the Victim trade. Victim might be (a) most profitable trade, (b) least profitable trade, (c) oldest trade, (d) trade with least-promising indicator_value (relative strength? Rate of change? Linear Regression Slope?) (e) whichever existing trade is in a stock (or industry group) that is most closely correlated to the stock of the new signal, (f) divide the existing trades into the top-half and bottom-half according to some indicator, then pick a Victim by random choice among the bottom-half. Coin flipping. Caprice. Nondeterminism. Use the cash freed up by selling the Victim, to enter the new trade.
2. Exit a small piece of each existing position to free up enough cash to enter the new trade. For example, if your total portfolio value is $56,000 and you have 7 stocks, sell $1000 worth of each of those stocks. Now you've got $49,000 worth of stock and $7,000 cash. Use the $7,000 cash to buy the new stock. Or instead of selling equal dollar amounts of each position, you could sell off equal percentages (sell (1/(7+1)) = 12.5% of each position, and use the new cash to enter the new stock).
3. Use the occasion to "rebalance" the portfolio. Sell or buy just enough of each one of the existing N stocks, that each one will represent (1/(N+1)) of the total account value. Use the new cash to take the new signal in the new stock.
4. The "Manana Principle": Temporarily go on margin to enter the new position, with the opinion that you'll very likely get an exit in another stock pretty soon. If you wait a while and still haven't gotten an exit and are still on margin, sell off a piece of each position (or pick a Victim, or ...)
5. Run backtests to find out what would have been the maximum number of simultaneous positions in the past, if you took all signals. For sake of discussion, say this turns out to be 60 simultaneous positions. Use this number to help you choose a positionsize (let's say you choose: 1/50th = 2.0% of the account) and then live with it. If you ever get more than 50 simultaneous positions such that you have no cash to enter new signals, live with it. Skip those trades and don't let them worry you. Those are somebody else's trades, they aren't your trades.
Trying out some or all of these ideas in backtesting, may take some time.
Thanks for your reply - good thoughts. I will spend some time going through my actual and simulated big winners to identify any similarities in their behaviour at the end of the trend to generate exit signals before the trailing stop is hit.
I like the idea of incorporating fundamental data into both entry and exit decisions, however, I prefer to be able to backtest each element of my systems. Are you aware of any reliable historical databases of fundamental equities information (PE Ratios, book value, ROE etc) that can be used in backtests? I primarily trade Australian equities (and haven't managed to find this data locally) but would be interested in trading us stocks with a similar combined technical / fundamental trend following approach if i could get hold of the data to test such a system.
I like the idea of incorporating fundamental data into both entry and exit decisions, however, I prefer to be able to backtest each element of my systems. Are you aware of any reliable historical databases of fundamental equities information (PE Ratios, book value, ROE etc) that can be used in backtests? I primarily trade Australian equities (and haven't managed to find this data locally) but would be interested in trading us stocks with a similar combined technical / fundamental trend following approach if i could get hold of the data to test such a system.
Sluggo,
Thanks for your thoughts - I will spend some time looking at how to back test these...I am a little conflicted by the idea of slimming down my existing positions in order to open new ones as this seems to be against the idea of letting profits run and maximising P&L impact from winning trades, but I will experiment with slimming down trades which are yet to show the "10 bagger" type trend which my approach capitalises on.
Your point about going on margin temporarily is an interesting point - I usually limit my total exposure based on the total amount of cash in my account + the amount of leverage I am comfortable with taking on (i.e. the amount of leverage which gives a reasonable tradeoff between total return and drawdown in historical testing). With CFD's the amount of leverage available on stocks is much more in line with leverage in the futures market (though the interest costs are pretty punishing for a longer term system)...Maybe i should consider taking advantage of this levarage and control my portfolio using overall portfolio heat rather than the total exposure - Does this make sense? I have been reluctant to do this to date because the worst case scenario in a stock market crash is pretty ugly if you lever up your account much over 50% and carry overnight risk (as is inherant in the longer term approaches)
Thanks for your thoughts - I will spend some time looking at how to back test these...I am a little conflicted by the idea of slimming down my existing positions in order to open new ones as this seems to be against the idea of letting profits run and maximising P&L impact from winning trades, but I will experiment with slimming down trades which are yet to show the "10 bagger" type trend which my approach capitalises on.
Your point about going on margin temporarily is an interesting point - I usually limit my total exposure based on the total amount of cash in my account + the amount of leverage I am comfortable with taking on (i.e. the amount of leverage which gives a reasonable tradeoff between total return and drawdown in historical testing). With CFD's the amount of leverage available on stocks is much more in line with leverage in the futures market (though the interest costs are pretty punishing for a longer term system)...Maybe i should consider taking advantage of this levarage and control my portfolio using overall portfolio heat rather than the total exposure - Does this make sense? I have been reluctant to do this to date because the worst case scenario in a stock market crash is pretty ugly if you lever up your account much over 50% and carry overnight risk (as is inherant in the longer term approaches)
-
AFJ Garner
- Roundtable Knight
- Posts: 2071
- Joined: Fri Apr 25, 2003 3:33 pm
- Location: London
- Contact:
I chatted to a Wizard a year or so when he was over in the UK. Although a futures Wizard, he was spending much of his time testing large portfolios of equities on very long term TF systems, quite of bit of which was concentrated on just this problem.
At that stage, he reported that his most profitable tests resulted from selling out the longest lasting and most profitable trades or at least cutting them back. His theory: these had or were likely to have reached a plateau (even McDonalds couldn't keep gowing for ever) and that this method made way for higher growth stocks.
Daunted by the sheer magnitude of the task of testing (and then trading) hundreds or thousands of stocks, I must report that to date I have not set out to confirm his findings!
At that stage, he reported that his most profitable tests resulted from selling out the longest lasting and most profitable trades or at least cutting them back. His theory: these had or were likely to have reached a plateau (even McDonalds couldn't keep gowing for ever) and that this method made way for higher growth stocks.
Daunted by the sheer magnitude of the task of testing (and then trading) hundreds or thousands of stocks, I must report that to date I have not set out to confirm his findings!
That same level of "daunted-ness" has led me to limit the universe I choose from, limit the number of positions, and focus on momentum as a selector.
I don't know if it's the best solution, but it seems to work well in the testing done so far.
[edited for spelling - whoops!]
I don't know if it's the best solution, but it seems to work well in the testing done so far.
[edited for spelling - whoops!]
Last edited by nodoodahs on Thu Feb 14, 2008 1:54 pm, edited 1 time in total.
If you're fully loaded and you get a new signal, you either take it or you don't. If you don't, that's suggestion (5): Skip those trades and don't let them worry you.
If you do take the new signal, you either leave your existing positions intact (and thus buy the new stock on margin), or you don't. If you do leave your existing positions intact, that's suggestion (4): Temporarily go on margin to enter the new position
If you do take the new signal and you don't leave your existing positions intact, you cut back existing positions either by exiting completely (suggestion 1) or partially (suggestions 2 and 3).
That's the entire decision tree. It's not very complicated.
a. New signal arrives when fully loaded: ignore or take
b. If take, buy it on margin or don't
c. If take and don't buy it on margin, exit existing position completely or partially.
If you do take the new signal, you either leave your existing positions intact (and thus buy the new stock on margin), or you don't. If you do leave your existing positions intact, that's suggestion (4): Temporarily go on margin to enter the new position
If you do take the new signal and you don't leave your existing positions intact, you cut back existing positions either by exiting completely (suggestion 1) or partially (suggestions 2 and 3).
That's the entire decision tree. It's not very complicated.
a. New signal arrives when fully loaded: ignore or take
b. If take, buy it on margin or don't
c. If take and don't buy it on margin, exit existing position completely or partially.
There is obviously no clear cut answer to the problem to how (if at all) to take new trades when your equities account is fully loaded...but a pragmatic approach based on some of the comments above might be:
1. When fully loaded, skip the first signal in the stock (with my LTTF system the first level trades tend to be less profitable than trades at levels 2-5) but watch the stock for subsequent signals
2. If a pyramid signal is given in the same stock (and you are still fully loaded) look at the decision tree outlined by Sluggo above - If there is an obvious 'victim trade' in the portfolio then reduce the size of that position to fund the new one
3. If there is no obvious 'victim trade' open the new trade on margin and look for weaker trades (probably ones which have had a big runup and subsequently slowed down) and reduce size on these trades to fund the new position
It seems to me that this approach gives a good trade-off, because in LTTF, most of the signals end up loosers (~60-70% losers in my case), so it won't matter much if I miss the first level trade, but if a strong trend develops I should get a pyramid signal and won't miss out on the whole trend - Kind of like a fail safe entry i guess.
This is probably difficult to back test accurately...Is this logic sound - it seems like it takes into account a lot of the points made in this discussion.
Adrian
1. When fully loaded, skip the first signal in the stock (with my LTTF system the first level trades tend to be less profitable than trades at levels 2-5) but watch the stock for subsequent signals
2. If a pyramid signal is given in the same stock (and you are still fully loaded) look at the decision tree outlined by Sluggo above - If there is an obvious 'victim trade' in the portfolio then reduce the size of that position to fund the new one
3. If there is no obvious 'victim trade' open the new trade on margin and look for weaker trades (probably ones which have had a big runup and subsequently slowed down) and reduce size on these trades to fund the new position
It seems to me that this approach gives a good trade-off, because in LTTF, most of the signals end up loosers (~60-70% losers in my case), so it won't matter much if I miss the first level trade, but if a strong trend develops I should get a pyramid signal and won't miss out on the whole trend - Kind of like a fail safe entry i guess.
This is probably difficult to back test accurately...Is this logic sound - it seems like it takes into account a lot of the points made in this discussion.
Adrian