Tweaking the Sector Rotation Strategy (Part 1)


In the next two posts I’ll test some tweaks to the Fidelity Select sector rotation strategy that we introduced last week.

Side note: turns out Woodshedder did a series on the same starting strategy back in January. I’m NOT going to dig into Wood’s work just yet. It should be interesting to see if we both came out in the same place.

[logarithmically-scaled, growth of $10,000]

The graph above shows the results of the original strategy (light red) and a modified strategy (dark red), versus the Vanguard S&P 500 fund VFINX (grey), since 1987.

Geek notes: (a) in my last post I used the S&P 500 price index as my benchmark, but VFINX (an actual investable asset) really is a better choice, and (b) to reward the strategy for time spent out of the market, I’ve assumed a return on cash of half the interest rate composite.

The first two tweaks to the strategy are related to the two issues I raised in my original post: (a) the original strategy tended to pick volatile sectors rather than ones that were showing momentum relative to volatility, and (b) there wasn’t a mechanism to sidestep bear markets.

To combat the first issue, rather than buying the sector fund with the highest % gain over the previous 25 days, here we’re buying the one with the highest average daily % return divided by standard deviation of daily returns. Now, less volatile sectors have a chance to get in the game.

And second, we’re not initiating a position until the 50-day (simple) moving average of VFINX is above the 200-day (read more about the Golden Cross). If the 50-day falls under the 200-day while a position is already on, we still hold the position until at least 30 calendar days have passed (to get past the minimum holding period).

This is trend-following at its simplest. In real-life I would recommend a more nuanced approach (like we take with the State of the Market report) where multiple long-term averages are examined and the signals averaged (so that you could be long say 25% or 50%, not necessarily just 0% or 100%).

Numbers for the number-lovers…

This is fast becoming a worthwhile trading approach. Changing the way in which sectors were selected and adding a mechanism for sidestepping bear markets significantly reduced portfolio volatility and drawdowns and increased risk-adjusted returns.

That’s all for this first round of tweaks. In our next post I’ll try to further refine and improve our little sector rotation strategy.

[Edit: click for a summary of posts related to this simple sector rotation strategy] 

Happy Trading,

. . . . .

To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.

26 Responses to “Tweaking the Sector Rotation Strategy (Part 1)”

  1. 1 John French

    Michael, I must say I like the idea of normalizing the av daily %age using the SD of av. daily returns. Did you use a 25 day lookback here too?

    • 2 MarketSci

      RE to John French: yes sir…just to clarify:

      For each sector on each day I…

      1. Calculated the (geometric) avg. daily % returns over the previous 25 days.
      2. Calculated the std. dev. of daily % returns over the previous 25 days.
      3. Divided #1 by #2.
      4. Used the highest result as my top sector on that day (which might or might not have been selected depending on whether we were past the 30 day minimum AND in an “uptrend”).


  2. 3 ashpak001

    Impressive. Especially given that the original strategy (called Pankin strategy) was considered by Formula Research at one point as one of the best they ever analysed.

  3. 4 kostas

    Michael, what would the results be had you used the simple momentum criterion (best return over last 25 days without adjusting for volatility) but had added to that the golden cross? My guess is that the results wouldn’t be much worse and quite possibly even better. It could be that the adjustment for volatility is not adding much here…

  4. Michael – How much better is it ? (“the original strategy tended to pick volatile sectors rather than ones that were showing momentum relative to volatility”)
    Is this new way THE way to pick sectors? yers, paul

  5. 7 kostas

    Since I asked the question on the contribution of the volatility adjustment of the momentum criterion above and beyond that of the golden cross, I ran some numbers of my own (using a different momentum criterion) and the results are very strong; adding any volatility adjustment dilutes the momentum criterion to the pont that the results ove time are worse in raw returns and even worse in terms of drawdowns. Of course I did not use the golden cross to modulate the downturns but something of the same functionality I am anxious to see if Michael can verify my point with his setup.

    • 8 MarketSci

      RE to kostas and paul jakob: adding the “relative momentum” tweak actually worsens performance a bit in terms of r/a return and drawdowns.

      But here’s the rub: we’re dealing with a relatively low number of trades with very long hold times, so this strategy is (more than most) subject to curve-fitting. I’m okay with the MA crossover idea b/c it’s been shown to work (on broad indices) going back a very, very long time.

      The relative momentum tweak is based on common sense more than what a curve-fitted test says. Consider this hypothetical: we’re looking at two assets in a bull market: an S&P 500 ETF and an S&P 500 2X ETF. Which is showing more momentum? By a simple % return method, the 2X fund is. But of course, it’s simply leveraged (similar to a high beta sector).

      The relative momentum metric balances that out, and (trading frictions, tracking error, etc aside) would show them as equal. I’m taking that same logic and applying it here.

      Just my $0.02 and by no means the only way to do it.


  6. 9 steve

    holding for how long? problem with FID funds is that they charge a .75% penalty for funds held less than 31 days. Monthly tets do not factor this in.

    • 10 MarketSci

      RE to Steve: see original post. The strategy is holding for 30 calendar days. michael

  7. 11 steve

    what if you had bought top 3, 5, 7 and 10 instead of top 1?

    • 12 MarketSci

      RE to steve: don’t steal my thunder =) likely a future post topic. michael

  8. Hi Michael,

    Great research! Forgive me if you already have but can you share a list of the funds that you used during your test please.


  9. 15 balazs

    Dear Michael, I have some issues concerning your analysis:

    1) Your modified strategy – which exits whenever the Golden Cross is down – should be compared not with a B&H but with a classic Golden Cross’ed B&H (which exits when MA50 is below MA200). But this is less important.

    2) The serious one is that your results depend heavily on the specific starting day of the strategy. I computed the distribution of annual returns for all 30 possible “rotations” and the one you published is among the best ones. I really wouldn’t use it as is, but the diversification of sectors – as you also mentioned – does wonders :)

    Anyway, brilliant blog, I’m stealing all your ideas :) keep up the good work!

    • 16 MarketSci

      RE to balazs: thanks for the kind words sir.

      RE #1: I agree and was thinking about doing a post about this. In short, a good bit of the “good performance” of this first tweak is a result of the MA cross that you could have taken advantage of using something like a simple index ETF/MF. But there is a consistent excess return (even accounting for volatility) that it coming from the sector selection itself.

      RE #2: I addressed this at the bottom of the first post. Yes, when you start the strategy would impact performance, but because it can hold for longer than 30 days (and now b/c of the MA crossover) all the possible variations “synchronize” pretty quickly. I ran multiple start dates and all sync’ed within the first year. Not a significant issue for this test.

      Thanks for the very well thought out comments…michael

      • 17 JC

        Hi Michael,

        I have also did some backtest and result the same as Balazs suggested – once the backtest starting date selected, the serie of desicion/fund turn over days are also fixed with their corresponding return. For your monthly (22 trading day assumed) revaluation strategy, you got only 22 independent series. I have to agree, all the series beat the S&P500, in which Golden Cross plays important contribution though.

        Also refer to

        I am not quick sure I understand you reply about “it can hold for longer than 30 days (and now b/c of the MA crossover) all the possible variations “synchronize” pretty quickly” – Isn’t the desicion made every month, or you have various valuation interval?



      • 18 MarketSci

        RE to JC: Two response. First, you need to look at 30 calendar days, not X number of trading days. Second, because the fund can possibly continue to be held for more than 30 calendar days, eventually (regardless of when you start the test) the strategy will synchronize and switch to a given new fund at the same time, because regardless of when you start the test at some point a fund is held for longer than 30 days and all starting dates switch to a new fund at the same time. Hope that helps. michael

      • 19 JC

        Hi Michael,

        I re-thinked your comments, basically you only valuated one series, which is month end. What about same rule but trading 1st day of the month, 2nd day of the month and so on, which gives you 31 series. I did some test, it almost same result pattern as trading base on trading days.


    • 20 MarketSci

      RE to JC: has NOTHING to do with month end. The strategy doesn’t HAVE TO switch after 30 days. It has the OPTION to switch after 30 days. That means regardless of when you start the test all potential start dates will sync up because some funds will be held for more than 30 days and the SAME NEXT FUND chosen regardless of which of which start date you selected. I’m not sure how to better explain it. If that still doesn’t make sense, I would suggest moving on to another less complex strategy. michael

  10. Top quality article! Have you seen Meb Faber’s related work?

    I have a similar system that I put up on C2. It uses a more nuanced SMA test, and 3 other tests to decide which global asset to buy (not just sectors). The back-tested cagr is higher than the system under discussion in part because I designed it to be able to go short using inverse ETFs.

    In my framework I was able to get pretty good improvement by adding a test for funds near their 52 week high. That may be interesting to mix in here…

    I am also interested in the answer to Kostas’ question about what volatility normalization does.

  11. 22 kevin

    Michael – I spent the morning building a spreadsheet to rank the 40 Fidelity Sector Funds per your rules above. For the 25 days ending 4/30/10, I show Fidelity Select Gold FSAGX as the number ranked fund with a volatility adjusted score of 41% (25 day geometric avg return of 0.57% / std dev of 1.41% = 41%)

    Can you confirm if this is consistent with your results?

    Rank 2 is Leisure at 30%
    Rank 3 is Construction & Housing at 29%

    Thanks for your analysis and ideas.


  12. Michael, lots of good stuff here! I’ve got lots of catching up to do. I’m behind in my testing and reading about other folk’s testing but I think this post will be a kick in the rear that I need.

  13. 24 Robert Jennings

    I’m becoming more interested in systems trading and I am really enjoying browsing your site. I’m pretty much a beginner but liking the idea of rotation, seasonality, MAs and stuff. I wonder if buying the best performing company on the FTSE 100 in the last month and then holding for another month + a 20 day MA filter would work. Hmmmmm will check it out.


  14. 25 Dan

    OK perhaps a dumb question, but I like this approach but would like to know where to find the daily and 25 day returns. I am not a member of fidelity. Thanks for the direction

  1. 1 Tuesday links: sticky capital Abnormal Returns

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s