Effect of timing the ‘dips’ versus ‘always in’

There is a certain satisfaction in buying a market low nearing the turning point and riding it northward.  For the momentum strategy described over the previous posts, using 12 month ROC to select a fund, what is the effect on profit and ‘time in market’ of timing the dips.

Another way of asking this question is: given the equity curve for ‘always in’, is there a way to reliably sell the high points and buy back lower?  Recessionary periods are filtered out as described here.


Annual return 34%, Max. DD 15%, Sharpe 2, Exposure 80% (time in market)

blog figs

Inspecting the above equity curve shows the difficulty in picking spots to sell in order to buy in lower.

TIMING THE LOWS (buy when > 1% of stocks are down 25% in a month)

Annual return 25%, Max. DD 12%, Sharpe 2.5, Exposure 50% (time in market)

blog figs

The result is to give up a quarter of the annual returns in exchange for 40% less exposure.  This is due to the exits, not the entries.  The horizontal green lines of the equity curve denote sidelined periods in cash, several during extended rallies.

This shows that:

1. Timing the rally tops is more difficult than the lows due to a difference in market behaviour.

2. The simple exits modeled are inadequate to detect when an uptrend is over, thereby missing several extended rallies.  (Pradeep at Stockbee uses a variety of measures to determine this qualitatively including breadth on several timescales and quantity of junk stocks making large moves, which tends to signal the approach of the end).

3. There seems to be a strong disadvantage to timing in this test, for this strategy, in terms of max. drawdown and profit.

Effect of momentum lookback period on performance

The literature shows that the 6 to 12 month momentum persists and longer term momentum mean reverts.  Vehicle selection by “Rate of change” (ROC) with a lookback period of 26 to 52 weeks, and a multiplier for the profit target [ATR(15)] of 2 to 8 were tested from 2003 – 2013 with the fund universe from the previous post.

Except for the recession filter, these were the only system parameters.

Performance was stable across parameter settings and Sharpe > 2.

blog figs


Buy =   r_filt;  Sell =  Cross(4%u,25%u); // breadth based sell condition

PositionScore = ROC(C,x);


The power of momentum coupled with a recession filter


1) Filter out recession dates using RecessionAlert real-time model.


3) Fund selection: highest 52 week ROC, single position.

4) Exit using simple breadth condition (see code below) or 6x ATR profit stops.


Annual return (2003-2013): 36% (including 20% of time in cash)

Max. drawdown: 13%

# trades 23, Sharpe 1.5

blog figs

Trade history:

Ticker Trade Date Price Ex. date Ex. Price % chg
IBB Long 3/21/2003 51.41 4/4/2003 52.29 1.71%
IBB Long 4/11/2003 49.54 5/30/2003 67.09 35.43%
FLSTX Long 6/6/2003 13.89 6/20/2003 14.22 2.38%
FLSTX Long (profit) 7/3/2003 14.2 11/7/2003 17.13 20.63%
FLSTX Long (profit) 11/14/2003 16.78 1/23/2004 19.93 18.77%
FLSTX Long (profit) 1/30/2004 19.21 2/25/2005 24.67 28.42%
FLSTX Long (profit) 3/4/2005 24.8 5/5/2006 30.72 23.87%
FLSTX Long (profit) 5/12/2006 29.45 4/5/2007 33.65 14.26%
FLSTX Long (profit) 4/13/2007 34.25 7/13/2007 38.82 13.34%
IYW Long 7/27/2007 57.62 9/28/2007 60.96 5.80%
IYE Long 10/5/2007 40.62 10/19/2007 40.32 -0.74%
IYE Long 10/26/2007 41.29 1/4/2008 42.05 1.84%
FLSTX Long 3/13/2009 14.24 5/8/2009 20.96 47.19%
IYR Long 5/15/2009 26.65 6/26/2009 28.1 5.44%
FLSTX Long 7/2/2009 20.16 8/28/2009 24.43 21.18%
IYR Long (profit) 9/11/2009 36.2 2/11/2011 54.98 51.88%
FDXAX Long 2/11/2011 12.84 3/4/2011 12.84 0.00%
IYE Long 3/4/2011 43.4 7/29/2011 42.55 -1.96%
IYE Long 8/19/2011 35.31 11/11/2011 40.27 14.05%
FBTTX Long 12/2/2011 7.92 2/17/2012 9.43 19.07%
FBTTX Long 2/24/2012 9.43 7/6/2012 10.83 14.85%
FBTTX Long 8/10/2012 10.59 8/31/2012 10.89 2.83%
FBTTX Open Long 9/7/2012 11.23 3/1/2013 12.35 9.97%

Amibroker code:

PositionScore = ROC(C,52);

Buy = r_filt ; Sell = Cross(u4%,u25%); // breadth condition (red dots in above plot)

ApplyStop(stopTypeProfit,stopModePoint,6*ATR(15) );

More testing

Further testing with the reduced universe and recession filter shows similar results to the previous post: Annual returns 25-30% including cash periods, Sharpe ~ 2 and low drawdown. The equity curve scale has been changed to log from linear.

Momentum selection metric is 12 month ROC.

One possibility is combining 2 entry methods to smooth returns e.g. 50% of capital when percent stocks down 25% in a month >1% and the remaining 50% when a breadth thrust occurs, say 4% of stocks up 4% in a day.blog figs

Testing with reduced universe

A smaller fund universe may simplify analysis and allow insight into the driver of returns. Here is the equity curve from the top Fidelity fund by momentum from this list: FATEX, FBTTX, FDXAX, FEIRX, FLSTX, OPPAX.  The holdings of these funds are summarized below and equivalent ETFs will be identified in a later post if available (expense ratios are >1%, although average hold time is only 2-3 months).

Fund Beta Top holdings
FATEX 1.09 Tech, aapl, goog, ebay etc.
FBTTX 0.87 Biotech, gild, amgn, celg etc.
FDXAX 1.15 Mid-cap, hfc, ksu, ame
FEIRX 1.06 Equity income, jpm, cvx, cmcsa
FLSTX 1.42 Leveraged co. stocks,lyb, sci, aes
OPPAX 0.94 erixf, ebay, sap, sie, ubs

blog figs

Annual return 30% (including periods in cash – green sections of equity curve)

Max drawdown 15%, sharpe 1.5 (exposure 50%).

The green sections in the equity curve are cash (no trade) therefore several large rallies were missed in these results: 2006, 2010 and 2011.

Adding recession filter

Simply filtering out recessions (see post) and buying the top ranked fund during sell-offs (>1% stocks down >25% in a month) gives the equity curve below since 2003.  Stats are 33% annual return, exposure 50%, profit factor 7, sharpe >2.  Trades are weekly and the lows are not timed precisely (e.g. by using a “breadth flip”).

Actual trades are listed with a 20% profit target.

blog figs

It is apparent that the following are important for good returns:

1) Avoiding long strategies in recessions

2) Correct vehicle selection (e.g. momentum ranking)

These factors may be more important than precise timing – this warrants more investigation.

Note that sharp market falls are possible outside recessions (e.g. 1987, 2011) but are mostly limited in duration and magnitude.