September 22, 2014

DTAYS Weekly Breakout Strategy With Time Stops

I recently read on Don’t Talk About your Stocks about an idea that stocks that were losers after (4, 6, 8) weeks should be sold to make way for other stocks that may do better. Will this idea improve the results from the original DTAYS Weekly Breakout Strategy? This reminded me of research I did while working for Larry Connors. On a mean reversion strategy we were researching, we noticed that after 10 days, 95% of the positions end up being losers. Then came the ‘obvious’ rule to add. Exit a position if it had not bounced after 10 days. We both thought this would greatly improve the results. It did the opposite and hurt them. Why? Because it was better to wait for the bounce even if the trade was a loser.

The DTAYS strategy is not a mean reversion strategy but a momentum strategy. How will exiting losing trades after X weeks work? Maybe the strategy is freeing up cash in non-performing trades to get into better trades. Time to see what the numbers tell us.

For this test I am making a slight change from the original test. I am testing on the Russell 3000 universe. Other than that the rules are the same as the original post.

Rules

When I tested several parameters, I bolded, italicized and underlined the parameter that DTAYS uses. Testing timeframe 1/1/2004 to 6/30/2014. Maximum number of open positions (10, 20). When have multiple signals, they are ranked from highest NWEEK return to lowest.

Buy

  • End of trading week
  • NWEEK values of (10,20,30)
  • Stock closes at a new NWEEK high of closes
  • The return for the last NWEEKs is greater than (20, 30, 40)
  • The 21 day moving average of Close*Volume is greater than $15 million.
  • Stock is member of the Russell 3000. This is a change from the original post rules.
  • SPY closes above its 100 day MA

Stops

Stops are evaluated at the close.

  • Maximum loss stop of (8, 12)%
  • Trailing ATR(100)*(3, 5) stop based on highest close since in position
  • If after (NA, 2, 4, 6, 8) weeks, exit the stock if it is under the entry price. This rule is only evaluated at the end of the week. This is our new rule.

 

Results without the N Week stop rule

140922a

The highlighted row is the best variation from the original post.

 

Results – With Stop – Using Original Best Parameters

140922b

The highlighted row is the base rules without the N week stop exit. The CAR and MDD are basically unchanged when we add the stop. The ‘% winners’ and the ‘Avg. % P/L’ drop dramatically. But the number of trades increases and this offsets the drop to keep the CAR about the same.

Results – Best variation of all runs

140922c

Here we see the CAR drop with some of the stops. MDD stays about the same. We see the same pattern with ‘# Trades,’ ‘Avg % P/L’ and ‘% of Winners.’

Spreadsheet

If you’re interested in a spreadsheet of my testing results, enter your information below, and I will send you a link to the spreadsheet. The spreadsheet contains more variations tested along with yearly returns.

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

Final Thoughts

Even though the N week stop rule does not improve the CAR and MDD, one still may want to add this rule. Psychologically it is always good to get out of losers. Some people would prefer to have the more trades because it improves the chances of getting big winners.

 

 

Click Here to Leave a Comment Below

Nick Radge - September 22, 2014 Reply

This so-called ‘stale exit’ techniques are something we’ve done research on before. What we found is that market leaders tend to consolidate rather than retrace enough to hit the stops. When the market moves higher again, its these market leaders that jump first out of their respective consolidation. So, a stale exit method tends to remove these first movers. If the entry mechanism is filtered, i.e. using some kind of ROC, then there is a chance that these market leaders won’t be picked up again.

In summary we found the stale exit technique diluted returns.

Steven Gabriel - September 23, 2014 Reply

What I like about looking at this data is that it shows something that is consistent with short term data. The reason why getting out of your losers and getting in to ‘new trades” seems to have similar results is, what I believe, to be a result of just trading out of 1 sold off trade into another sold off trade. When they bounce, they bounce about the same amount.
You are really just trading apples for…apples to make you feel better by getting out of your losers. That’s it.

Marco - October 7, 2014 Reply

This is one of the best example of backtesting results changing initial trader’s thoughts…
https://nightlypatterns.wordpress.com

mike - October 30, 2014 Reply

Back in March you tested this strategy and saw CAR around 4%-5%. This test, as far as I can tell, the only change is you restricted the selection to Russell 3000 stocks – is that the only thing you changed, because you have a huge jump in your results – shocking really.

Also, you indicate that a 5 multiple on an ATR stop performs quite well. A 5 multiple for an ATR stop seems really high to me. You set your ATR stop to the_highest_high – 5 * ATR(100) – correct?

    Cesar Alvarez - November 1, 2014 Reply

    @Mike. I am not sure what you mean by “saw CAR around 4%-5%” because in the post the CAR is much higher than that. Yes, the only change I made was to restrict the universe to the Russell 3000 stocks.

    The ATR stop is based on the highest close while in the position.

      Mike - November 2, 2014 Reply

      I see now, I was looking at your Base Results table – very impressive results overall – thanks for your great work.

Leave a Reply: