Skip to content

Equity Curve Lights and Best Performing Indicators

May 25, 2010

This has been long overdue—-but a necessary addition to the suite of tools required to manage strategies and systems. Please comment to indicate your interest. If the interest level is high enough, I will make this a regular addition along with the Livermore on a weekly basis for the time being.

Every Friday we will provide equity curve lights on the blog for several major indicators. They will be quite simple: red means stop, orange means you should take a small position, yellow means reduce exposure, green means you can take a full allocation. Furthermore we will also provide a list of the top performing indicators by category, as well as a category relative strength measure looking at both trend and mean reversion indicators.

Why Is This Important?

The rationale is simple: the best we can do is determine what has worked best in the recent past, and also in the intermediate term, and then apply a safety mechanism or “stoplight” to manage the strategy exposure.  There is no magical metric that can tell you whether a system will continue to work in the future no matter how careful your measurements are. Performance evaluation is most certainly not an exact science–if it was then it would be a lot easier to predict which mutual funds or hedge funds perform best in the future using past performance (hint: most methods fail to achieve their goal out of sample) . What you need is something that that also indicates whether your initial hypothesis was correct in assessing performance and a means to adjust your position otherwise.  Drawing concrete conclusions and persisting with these beliefs is a perilous path that many professors/practioners have taken. These same individuals  preach about effects that were validated using extensive testing, and then they go out and start a mutual/hedge fund and everything miraculously stops working. They stubbornly protest and defend their studies when they should be focused on trying to determine whether they should stop using the strategy and proceed to trading another.  You should not care what variables or what  indicators happen to be the best, I would trade anything that is logical and and is working well, and I would stop using my favorite strategy if I felt that it was no longer relevant —the real goal is to maximize risk-adjusted return and achieving that goal is a matter of  structured portfolio management and dynamic asset allocation .

Bringing System Testing Out of the Dark and Into the Future

Its about time we open our minds a little more and learn to change with the tide. Human beings are victims of the inherent need to have strong and persistent belief systems.  In an ideal world most of us would slip into reliable routines and rituals that ensure progress. No one wants to consider the prospect of having to move jobs and/or houses every year any more than we want to change our methods for trading. Our secret burning desire for regularity and unquenchable thirst for too much information in the face of uncertainty makes the trading game difficult to succeed as a pro. Step 1 is that we all have to limit our enthusiasm and comfort with the 60-year+ backtests–what worked decades ago has virtually no relevance for short-term strategies, especially since commissions were in excess of 1% per trade each way back then. Nothing you could find would even be tradeable net of commissions–there is no free lunch– the structure of market created an illusion of profitable anomalies or effects.  These long data histories are valuable primarily to display the ability to learn from data on a walk-forward test and thats pretty much it. Step 2 pertains to avoiding futile attempts at unconstrained mass parameter optimization without considering degrees of freedom or the false discovery rate. Even worse is the tendency to assume that such ridiculously over-fitted strategies will work for years to come after 5 minutes of optimization (I guess even worse than that is trading these “systems”).  Step 3 pertains to becoming  too exuberant about trying to make a strategy “robust” –we have to limit painful and meaningless exercises in trying to establish robustness by verifying that  a short-term S&P500 strategy that must also work on the Euro or the Solar Industry ETF to be considered valid. Guess what, markets are different–even the same index traded using different instruments or time zones can behave very differently (great post here)   The only way to survive is to adapt to a particular market, and adapt to the times, and trade the best effects/anomalies you can and be willing to hop off the train when things stop working —anything less is asking for trouble.

24 Comments leave one →
  1. Brad permalink
    May 25, 2010 4:49 am

    High interest, David!

    • david varadi permalink*
      May 26, 2010 1:57 am

      thanks to all who commented—-i will try to follow through with this as soon as possible.

  2. J Marc permalink
    May 25, 2010 5:35 am

    Would really appreciate this David. Thanks again.

  3. May 25, 2010 7:15 am

    The EC Lights would be a mild curiosity if they are simply posted. However, if the workings of the indicators are disclosed then they can be tested and we can become confident enough in them to make use of them. Traders tend not to use that which they can not independently verify and/or experiment with.

    As to the discussion of the usefulness of historical data. I agree with all that you say. I do, however, think you may overly diminish the value of testing over very long stretches. This is especially true re the performance of “adaptive indicators”. Unless you believe that we are permanently in a mean reversion regime (which I know you don’t) it seems essential that any truly adaptive indicator demonstrate its ability to handle the monumental changeover from trend following to mean reversion in the 1998 to 2000 time period. Ten years or shorter lookbacks don’t do that. Thanks for all you do,

    • david varadi permalink*
      May 26, 2010 1:55 am

      hi jerry, i agree with point one and we will be showing commonly used indicators or measures in addition to publicly disclosed dv indicators. as for point two, i mention in the paragraph that indeed this is the primary use for historical data–to use for out of sample testing. as for point three– i have no idea what regime we will stay in, but I can say that the “trend” or follow through regime was artificially induced by the inability for market players to quickly cash out of positions that were winning at a profit.this created long up and long down runs. in the current environment with very low transaction costs, and lots of high frequency trading, most players cash out after up days and vice versa. this creates a mean-reversion effect. note that the adaptive rsi handles this transition easily.


      • May 26, 2010 9:43 pm

        That’s the best explanation that I’ve ever heard for the crossover from trending to mean reversion in that period! Although the 2009 crossback from MR to Trending seems to have other causes, my point was simply that when we are testing indicators that are designed to be adaptable that we test them over the trend to mean reversion crossover period in 1998-2000 as well as the more recent period because it was so significant, and was in the other direction. That shift was just the type we want to be sure an adaptable indicator can handle. Thanks, Jerry

  4. May 25, 2010 7:54 am

    Sure…. Let’s give it a go !

  5. Aristotle permalink
    May 25, 2010 8:43 am

    Sounds like a very interesting idea, I’m all for it!

  6. Emil permalink
    May 25, 2010 12:27 pm

    Very interested, how could anything from this blog ever be boring?

  7. Peter permalink
    May 25, 2010 1:11 pm

    Very interesting!

  8. Roger permalink
    May 25, 2010 6:05 pm

    Interested but strongly agree with Jerry above – more information about the workings of the indicators would make them very valuable.

    Following black box indicators is not something I’m that comfortable with. I’ve appreciated the insight into your indicators you have given on the blog without always revealing the whole “secret sauce.” For instance, I have conceptual understanding of how DVO works via your blog postings but don’t know the secret sauce (use the Excel/Amibroker versions for this).

    Keep up the great work

    • david varadi permalink*
      May 26, 2010 1:56 am

      roger, point taken, i will attempt to move more towards open box solutions in the future.

  9. Wilson permalink
    May 25, 2010 6:50 pm

    Hi David, I’m a new reader of your blog, and this is my first comment. I’ve just been reading through a number of your posts, including this one. Very thought provoking indeed! If the short term strategies are so time/market sensitive, the development of such strategies would require one to work with data of the recent past. Would this leave little data for out-of-sample verification, and therefore less confidence in the strategies once they go live? -wilson

    • david varadi permalink*
      May 26, 2010 2:00 am

      hi wilson, i think the idea with short-term indicators is that only a 2-5 year track record is required. out of sample validation is less effective than simply hopping on to an indicator that has worked well recently and stopping trading if it has a negative performance over a 1 year period. this is based on extensive testing.

  10. Victor Martindale permalink
    May 25, 2010 7:20 pm

    Absolutely! Sounds like an excellent idea.

  11. BMB permalink
    May 25, 2010 7:46 pm

    Very interested.

  12. Carl permalink
    May 25, 2010 9:22 pm

    Very interested but agree with Jerry.

  13. George permalink
    May 26, 2010 7:46 am

    Great. Very Interested.

  14. Jim permalink
    May 26, 2010 5:16 pm

    Look forward to to it.

  15. May 26, 2010 6:03 pm


    1. How does one know whether in this HFT world the equity lights are blinking fast enough? Is there any way to backtest the lights?

    2. If we’re trading more than one security, then other factors start coming into play. 5 stocks x 2 trades/day x 20 days/month x 0.01% slippage & commissions (ie a penny on $50 stock) = 2% per month. Of course that’s on the high end for frequency, but that’s probably more than offset by the low slippage.

    Don’t get me wrong – I like where you’re going with all this, but depending on whether you’re advocating switching strategies or stocks, I’d want to hear what you have to say about 1 or 2 above.

    • david varadi permalink*
      May 27, 2010 2:08 am

      hi alex, this would be a medium to long term overlay to add value to risk management. the equity curve management would not cause extreme turnover since it would not intervene very often and many adjustments would simply involve position size. it would be used to assist with monitoring systems versus switching stocks.


      • May 27, 2010 4:59 pm

        Aren’t the regimes stock- and not market-specific? I’m sure there’s correlation, but i still don’t see why we’d use these if we can’t backtest how the signal is being generated.

        (You’re right DV Panel doesn’t have that much turnover. I guess it would then depend on LTR turnover, but with weekly updates, we’re probably looking at relatively minimal costs)

  16. kostas permalink
    May 27, 2010 10:39 am

    Hey David, great interest but the lights seem off and I would rearrange them: red=stop, orange=reduce, yellow=small position, green=full throttle. Even better I would substitute BLUE for the yellow (red-orange-blue-green), for a more intuitive/natural progression from sidelines to full exposure

  17. derek permalink
    May 29, 2010 7:16 am

    It’s absolutely a good idea. The balance between patience and abandonment with a system is up to the individual’s time frame, but in this case I think more info is better

Leave a Reply

Fill in your details below or click an icon to log in: Logo

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

Google+ photo

You are commenting using your Google+ 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 )


Connecting to %s

%d bloggers like this: