How to use Trailing Stops?

Research
1

How do traiing stops work when stocks are bought at different times (either to average in or if different parallel running systems recommend same stock at different times). I use IB for instance. I don’t see a way to separate out and list trailing stops separately for each separate buy. It seems to lump them together. If I set a 30% stop and the stocks were bought at 20% different prices originally, does the trailing stop get set to the stock bought at the lower or the higher price? Do different brokerages handle this differently?

Thank you.

2

Hi,
With Scottrade the trailing stop is set whatever % (or points) below the high of the stock since the time the trailing stop order was placed. It doesn’t matter when or at what price the shares were purchased. I haven’t tried it but it should be possible to place separate trailing stop orders with separate stop levels for different shares of the same stock.

FYI, occassionally Scottrade deletes Trailing Stop orders without notice, I think at the end of the month following the month the order was placed - most inconvenient.

Don

3

Hi Don,

Thank you for your reply.

I have learned that IB does indeed separately track a separate trigger price for each trailing stop according to when it is placed. So they don’t reset the trailing stop to a common price if the same security is bought at different prices; they track them separately, even though they aggregate the total number of shares held when they display the positions. I have not verified this difference in trailing stop trigger points with actual trades at this point. So, this makes me more comfortable. It allows me to mix the same stock bought by 2 diffferent systems and keep their trailing stops distinct if I so chose.

I don’t know how to determine from their Trader Workstation which trade a TS relates to…it seems I must keep a separate record of that to keep track of it.

To be honest, I am not always firm on mixing the same stock from 2 systems. Sometimes if I have already bought a stock with one system, I will bypass it and pick the next lower ranked stock. My main reason to buy the same stock twice from 2 different system is to minimize deviation from backtested methodology. The countervailing reason to not double up on a stock is for postion size/risk control.

4

Related to Trailing Stops, has anyone created a solution to this problem: When a Trailing Stop is in the sell rules, that rule is only invoked once per week (two weeks, month, etc.) as we run our Sims and generate estimated returns, drawdowns, etc.

However, when we put real-life trailing stops on the stocks, they get sold on the weekdays, they don’t wait until we’re ready to rebalance on the weekend. This obviously throws the entire Sim out of whack. Any sim with a TSL built into it is therefore an approximation at best.

If you have done testing for robustness (as you should on every sim), you know the extreme variation in results that can be caused by a different start date or a different rebalance time period applied to a sim.

It seems to me that when we create a sim with a sell rule such as “PctFromHi<-14”, but have actual Trailing Stops on our real life portfolios that sell the stocks daily, during the week, the forecast of the Sim is thrown out the window.

Turnover and trading costs are drastically increased, for one. Additionally, The stocks that are produced in a sim may look nothing like the real-time portfolio’s stocks after a volatile month or two with a high number of TSLs, causing a potentially extreme "Butterfly Effect " (“for example, a ball placed at the crest of a hill might roll into any of several valleys depending on slight differences in initial position”).

Is there a work-around to this problem that anyone has found? Are there any ways that it can be implemented using the cuurent tools available? What about creating “Sub”- Buy/Sell Rules that can have a seperate rule or two with a separate rebalance period? All the fundamental Buy/Sell rules can be rebalance weekly/monthly, but the “Sub-Rules” can be rebalanced daily. This would also have an application for TA functions (“Sell a stock if its price drops below the 50-day moving average”).

You could have all your other Buy/Sell Rules Rebalance weekly, but the Trailing Stop sell rule would be checked daily and a rebalance ran to replace the stock that got sold.

Is anyone interested in this enough to warrant a Feature Request? Personally, I would love to have it, instead of my ports being only approximations of what could happen from my human intervention when a TSL is triggered or I think a stock should still be held. I personally have a % TSL on every position purchased, so this would affect every sim/port I have.

In the meantime, let me know the work-around you guys have come up with (if any) for a weekly rebalance combined with a real life daily TSL. Thanks!

5

Hi,
There is an open request for trailing stop loss, you can vote for it here:
http://www.portfolio123.com/feature_request.jsp?view=open&cat=-1&featureReqID=23

There is another request for a fixed stop here:
http://www.portfolio123.com/feature_request.jsp?view=open&cat=-1&featureReqID=360

Don

6

How does everyone handle this?

Denny, I know you are a big TSL user. If you set your TSLs a bit higher than the weekly rebalance TSL, as I do, you will occasionally get a lot of sales and turnover when the market slumps and you are still holding positions.

As you know, I now have a policy of going to cash a few days in advance of any significant correction. But I still get quite a few of my volatile small caps that are occasionally sold during the week.

Additionally, I may have a higher amount of TSLs than others, because I use them occasionally as a “take profits” part of the system. I start with a basic TSL rate of -15%, then as the profit of a stock moves up, I have P123 tighten down the TSLs. When a stock has gained 36% or more, the TSL is only 3%. Here’s the formula I use, which Dan Parquette and I put together::

Eval(GainPct<=15,PctFromHi<-10,Eval(GainPct<=35,PctFromHi<-5,PctFromHi<-3))

Is there any way to simulate a “during the week TSL” while the Sim only rebalances on weekends?

This is obviously affecting the sims dramatically, but we don’t know if its an up or down effect.

The sims are being run and may allow a stock to drop to half or less of its original value during the week, and you won’t get a TSL sell until the weekend. It seems that if you are locking in those losses with a live TSL during the week, the sim may be overestimating the annual return. I guess the question would have to be, “How well did my stocks do after I sold them out of the system compared to the replacement for it that I chose.”

Does anyone have time to take on that project?

7

That’s just the half of it. If the stock recovers to above the TS by the end of the rebal period, P123 won’t even recognize that a Sell took place.

Here’s a workaround that at least triggers a Sell in this situation. A TS rule like PctFrmHi < -20 will be evaluated only at rebal time. This workaround might work (I haven’t tried it yet):

Lowest(#Low,5,0) < Highest(#High, NoBars,0) * 0.8

The first term is the low since last rebalance (replace the 5 for rebalance periods other than weekly). The second is the high since the position opened.

Unfortunately, as P123 is implemented, the sell will be evaluated as of the rebal time, not at the trigger price. Still, this should be a small but positive change toward more realistic sims.

8

Since we seem to be heading into our second market correction in two months, I’m starting to set trailing stops in my Scottrade account without messing with my P123 sell rules or trading system. I’m using a setting of -10%. If I get stopped out on any particular holding, I’ll manually enter that sell transaction in P123, wait a few days to guage market direction, then try a rebalance and see what P123 suggestions come up.

I’ve used stop loss provisions in simulations and found that most of the time, they increase turnover but don’t add much to the overall return. But I think that’s because my sims and ports typcially rebalance only once a month and stop loss orders need to be much more timely to be effective. So I’m going to try this in real life and see how it works for me. I’m not putting in TSL orders on every holding, just the ones that have gone up the most, or that show very low rankings in the Valuation portion of my ranking system.

Does anyone else have real world experience with TSL orders that they’d like to share?

9

Does anyone else have real world experience with TSL orders that they’d like to share?

Given that the pullback systems (see Olikea’s challenge) almost certainly work by taking the other side of people’s stop losses, I would say stop losses are a bad idea.

If you want to get out, you want to get out on a high, a stop loss always gets you out on a low.

10

You could of course write calls on your position instead of using trailing stops …

11

Olikea is right, on one point, the pullback system works well. However that doesn’t mean that stop losses don’t work. They work very well for me, and I think that they will also work well for everyone else except in some unusual trading approaches. See my Stop Loss Challenge here .

Ginger,

Here are 2 approaches that I use:

In the first approach, I set the real trailing stop loss to 3, 4, or 5% higher than the Port’s Stop loss; 3% for my ports with the lowest volatility, and 5% for the Ports with the highest volatility. I do this for capital preservation, not for performance. This way the real stop loses are rarely triggered during the week when the Port’s stop losses are not also triggered on the weekend rebalance. The difference is that on the average of 81 real stop losses that I have experienced in the last 3 years since I have been keeping the data, I got out of the stocks at a price a little better than the Port’s price.

In the second approach, I have developed some Sims that more closely approximate real stop losses triggered during the week. I have a few Ports in which the stop loss is the only sell rule that has anything to do with price or volume. The Sims for these Ports were run with daily rebalance. This causes the Sim to sell due to a stop loss during the week since Price and Volume are kept in the data base daily, but the Sim only sells stocks over the weekend due to the other sell rules since P123 only takes a snap shot of the other functions weekly in the data base.

The short coming of this approach is that the Sim only sells at the close price if the price closes below the stop loss instead of at the stop loss like the real stop loss would do. This would be greatly improved if this Feature Request were implemented.

With the Ports that were developed this way I set the real stop loss the same as the Port’s, and I rebalance daily.

Denny :sunglasses:

12

Denny, while I don’t disagree that you should get out of a trade that is going sour, I find it logically incompatible that it is both a good idea to buy into a stock going down (as in the pullback system) and sell a stock going down (stoploss).

However, I have noticed that the rule “close(0) > close(1)” gives a boost to returns if it is included as a necessary sell rule.

Why don’t you try “pctfromhi < -20 AND close(0) > close(1)” ? Daily rebalancing is absolutely necessary, but I think you may find it does improve matters to get out on a local upswing.

13

Olikea,

I agree with your statement “I find it logically incompatible that it is both a good idea to buy into a stock going down (as in the pullback system) and sell a stock going down (stoploss).”

However, this statement only addresses the stock in question, and ignores other issues that may affect performance.

There are at least 2 situations that will change the actual performance.

  1. The stock that replaces the stock that is stopped out may increase in value faster than the stock that was sold if it had been kept.
  2. The pullback system works well for certain stocks under specific conditions. For other stocks that are going down they just keep going down, so when do you sell them and move on?

A stop loss properly tested will set a value that moves out of the losers at the best point to optimize the affects of the above 2 issues.
Again I ask, “If a stop loss improves the performance and drawdown of properly tested and robust Sim, why not use it?”

I’ll try your ideas for PctFromHi < -20 AND close(0) > close(1). That looks intersting.

Denny :sunglasses:

PS: Thank you for the Ideas of the Pullback System and Steve for his hints to his Holy Grail. My hybrid Port using these ideas is up 125% since 05/23/07, and only experienced a 7.4% drawdown during the July-Aug correction.