New Backtests for ETFReplay Portfolio

I am frequently asked about various strategy and portfolio performance metrics and backtests. A reader recently asked if there are any current backtests for the ETFReplay 6/3/3 Portfolio so I decided now is the appropriate time to provide updated results.

The strategy background is available here and is updated monthly on Scott’s Investments, including a real-time  simulated portfolio spreadsheet here.

The ETFReplay portfolio is also provided free to subscribers of Portfolio123, simply search “ETFReplay” in the Ready-to-Go Portfolio section.

The backtested results below are hypothetical and do not account for commissions or taxes, so real results would be lower.

The first test was conducted using , the initial inspiration for the portfolio. The backtest invests in the Screener’s picks on the close of the first day of the next period.  Picks are updated monthly and the top 4 ETFs are purchased/held.  ETF must also be ranked about a short-term treasury ETF (SHY) in order to qualify for purchase.

This test is nearly identical to the 6/3/3 system I update each month with the exception that the test does not require ETFs drop out of the top 5 before being sold. This was a qualification I added to the strategy to limit turnover.

The benchmark for the test is VBINX (Vanguard 60-40 Mutual Fund):

etfreplay1 etfreplay2

The strategy has performed well in total, but has lagged a 60/40 benchmark in the last 3+ years.

The Pure Momentum ETFReplay Portfolio was added to my monthly updates in 2014.  It purchases 4 ETFs each month based on 6 month returns, and has no cash filter.  Backtested returns are below:

pure momentum pure momentum2

We see similarly strong early results, with returns lagging versus the 60/40 benchmark the past 2+ years.


More on this topic (What's this?)
Core ETF Report
Read more on Exchange Traded Fund (ETF) at Wikinvest

12 thoughts on “New Backtests for ETFReplay Portfolio”

  1. Scott, Your ETFreplay articles are terrific. Kudos. I have been fooling around with ETFreplay (a subscriber) for a couple of years now. Did OK in 2013 and 2014 but 2015 is a real b-tch. I have a few questions for you. First, your system where you buy the top 4 of 14 each month, and then you sell the ETF if it falls out of the top 5. I can’t do or even imagine how you would program that “sell if it falls out of the top 5″ into ETFreplay if you’re only buying the top 4. I’m very confused. Do you do that inside ETFreplay?
    Second, I saw the line in your excel spreadsheet (something like) =SUM(googlefinance(gld,”price”)*1) . I put that into my excel and get a #NAME? error. Are you using different software or some special excel setting to draw prices off of the internet.
    And lastly, whenever I backtest a system, I throw in a 6 month SMA filter rather than a SHY filter – I think it just about always gives better results. Your thoughts?
    Scott, keep up the good work, and if you’re ever in Venice Florida, let me take you out to lunch.

    Jerry Rehert

    1. I don’t / can’t program the top 5 rule in ETF Replay. However, I have successfully created the rule using Portfolio123. Google spreadsheet code and excel code aren’t the same. If using Google Docs, the code should work. Regarding SMA vs absolute momentum, I think they both accomplish the same objective of crash management. I may take you up on the lunch offer someday! Scott

      1. I see all these studies that talked how relative strength works over 3, 6, 9, and 12 months but is mean reverting over 1. Only when I use ETF replay, I can find a single relative strength backtest (except 3 months using equal weight sector etfs) that out performs the relevant benchmark. And even in the case where 3 months using ew sector etfs works, 6, 9, and 12 months don’t. In fact, the back test is horrible. Trust me, I want to use the 3 month RS as the numbers are terrific, but it looks like statistical noise and coincidence. I can’t have confidence that it will work when no other time period will. That’s just hypothetical fitting. Can you explain why in every white paper out there, relative strength works but not in real life use? And going to cash should not be necessary. Yes, the drawdown would be more than if not in cash, but I am looking vs the benchmark and it can’t even beat that?

        1. I guess I would need to see the tests you run, I have seen improvements in tests ranging from 3- 12 month momentum vs. buy-and-hold. Obviously not every test shows the same performance.

  2. Do you find it alarming that many mechanical strategies fail in the years following their discovery?

  3. Scott:
    Many thanks for the backtest on the ETFReplay 6/3/3 above! Was slippage accounted for in the test results?? If not, can you guess at the cost of slippage would have on results??

    Thank you.

  4. Hi Scott:

    What do you think accounts for the poor results vs the 60/40 benchmark in 2013-present??
    Might reducing # of ETFs to 2 help performance?


    1. This article actually explains some of the concepts behind trend/momentum and why they under-perform over long periods of time:

      2-3 years is a relatively small sample in the grand scheme, and momentum can underperform for long stretches. Also, equities have done relatively well while bonds have held their own given stable interest rates over the past 2-3 years, boosting 60/40 returns.

Comments are closed.