September 8, 2013

Dealing with Broken Arrows

When I buy a stock, I know the edge is for the stock to go up. I know that does not mean every stock is a guaranteed winner. Or that my losses will be small. But every now and then you get what I call a broken arrow. I was in Expedia (EXPE) as it triggered in my SP500 rotation system. It was in a nice up trend, everything looking great.
brokenarrow1
And then …earnings happened.

EXPE closed down that day 27%…ouch.  That is what I call a Broken  Arrow.

brokenarrow2

 

Make sure you have a plan before you enter any trade, and stick to it

If a Broken Arrow happens to you, do you know your exit strategy?  I can tell you that if you hold trades overnight, a Broken Arrow will eventually happen.  When a trade goes Broken Arrow, as it did with EXPE, the first thing I do is go with my plan.  Not a plan that I invent after it happens, the plan I created before I entered the trade.  In this particular case, my plan was to simply hold until the next 5 day rotation period (I was trading this in a rotation system).

Then I starting thinking is this the best thing to do in this case? From previous research, I know that large drops are usually the worst time to sell.  Maybe holding for five days is not best. Maybe it is best to sell at the open on the next day. Or wait for the RSI2 to be greater than 70.

A rule I have is that I do not change a plan in mid-trade even if the research backs it up. One is still trading with too much emotion at that point to read the trade statistics dispassionately. But I wanted to know if the next time this happens (and there will be a next time), what my plan would be. Would it be the same or would I have a better exit?

Testing from 2009

Test results for my strategy start from 2009. I do not like testing too far back because markets change. We have been a bull market since 2009 and still are in it, so I believe the results are good.

 

S&P 500 Test Rules

I wanted to test that the stock was in a healthy uptrend which the moving averages do.

The Setup

  • Stock is in the S&P 500
  • Close > 40 day Moving Average of Closes
  • Today’s 40 day Moving Average of Closes is greater than yesterday’s 40 day Moving Average of Closes

The Buy

  • Yesterday was a Setup
  • Today closes down 25%

S&P 500 stocks don’t have large one day losses that often

First I looked to see how many S&P 500 stocks closed down 1 day 25% or more. But only 8 did. I like to see at least 100 trades but prefer at least 200. Next I tried down 20% or more and this only gave 34 trades. Trying losses of 15% or more gave 121 trades. Enough in a pinch but I really wanted more. So I devise a way to take a look at a bigger universe of stocks.

The New Universe

To expand the universe, I went to the top 1000 stocks by dollar volume. A lot of these will be S&P 500 stocks and big cap stocks. I looked for stocks closing down 15% and this gave 318 trades which is enough for me.

The Tested Exits

I wanted to test three exits.

  1. Exiting the next day on the open
  2. Exiting on the next open when the RSI(2) closes above 70
  3. Exiting five days later on the open

Results

Included are the results for the S&P 500 for two reasons. One, to see if they produced similar results to the top 1000 stocks. Two, more data is always more interesting.

Exit # Exit Type Universe # of trades Avg. Profit/Loss % % Winners
1 Next Day Open Top 1000 412 0.77% 56%
2 RSI(2) > 70 Top 1000 412 1.29% 53%
3 Five Days Top 1000 412 1.36% 49%
1 Next Day Open S&P 500 121 0.66% 60%
2 RSI(2) > 70 S&P 500 121 4.27% 57%
3 Five Days S&P 500 121 2.66% 52%

 

As you can see, waiting to get out at the next open on average produces a positive exit versus exiting on the close of the big down day. As painful as it is to see these big down days, it is best, on average, to at least wait until the next day to get out.

From previous research experience, I expected exiting with RSI(2) > 70 to produce the best results. Looking at the Top 1000 stocks, the results between exiting when RSI(2) > 70 or exiting five days later give about the same results. But more importantly, almost 2x better results than exiting the next day’s open. Looking at the S&P 500 results paints an even better picture. By the way, it is not trivial that the average hold for the 5 day exit is  5 days versus 10 trading days for the RSI(2) > 70 exit. Waiting for the bounce gives a much better average % profit/loss and % winners.

It looks like I will be changing my trading plan on these broken arrows. Fortunately they do not happen often.

Comments, suggestions or ideas on further tests?  Put them in the comments window below!

Backtesting platform used: AmiBroker. Data provider:Norgate Data (referral link)

 

 

Click Here to Leave a Comment Below

Andy - September 16, 2013 Reply

By the way, great post. I’m a big fan of the Connors RSI. Did you test to see if the result improve when CRSI() > 80 ? Since the introduction of the CRSI, that is the indicator I been using for my swing trading system. I’ve been successful using it.

I was thinking about a strategy exactly what you are describing here but totally forgot to test it when I got home. Buy a SP500 or large cap stock when it drop a large percentage (10 or more percentage), and sell it when CRSI > 80.

    Cesar Alvarez - September 17, 2013 Reply

    Andy,

    I did not test using ConnorsRSI but would imagine that the results would be similar. I do plan to do a future post on ConnorsRSI. Anything in particular that you would like to see?

    Thanks,
    Cesar

      Andy - September 17, 2013 Reply

      I find that ConnersRSI works great for swingtrading using EOD data, but I haven’t found it to work using intraday data (5 min bars) for daytrading. I believe it’s due to the percantage calculation that requires 100 bars of data. Therefore, opening gap plays a big challenge. I heard about gapless version of an indicator. I wonder if that would solve this problem. Any thoughts ?

      Andy

MP - September 21, 2013 Reply

Would be great to show Avg. P/L divided by avg. days held, or “Avg. P/L per day”. Without the days held data, it’s tough to glance at a table and identify the strategy variations that are the best use of capital. Thanks.

    Cesar Alvarez - September 21, 2013 Reply

    I would agree with the interest in the “Avg p/l per day” stat. One issue I have with the stat is that most of the times it favors a shorter hold. In this case since I am trading a rotation system, to me it is less important. Here is the stat for the SP500 stocks
    Next Day Open – .66
    RSI2>70 – .49
    Five Days – .53

Andy - October 30, 2013 Reply

Very interesting. I can imagine it would be extremely reassuring to know that you already have the most effective plan in place to handle a broken arrow trade should you be confronted with one.

PV - September 9, 2014 Reply

Is it possible to get earnings data so you can filter out stocks that release in the next week?

    Cesar Alvarez - September 9, 2014 Reply

    Yes it is possible but it tends to be really expensive (too much for me too afford). See http://www.crsp.com for one example

Leave a Reply: