When to Hold ’em, When to Fold ’em

One of the big questions facing a trader or fund manager occurs when previously successful methods no longer work.  There are three choices:

  1. Wait for things to return to "normal";
  2. Tweak the method or model to reflect the new conditions;
  3. Quit trading.

The choice is a matter of psychology as much as professional evaluation.  The first step is in recognizing the nature of the problem.

What to Tweak

In Brett Steenbarger’s book, Enhancing Trader Performance, he presents the concept of "finding a niche."  There are many ways to succeed at trading.  Sometimes a method works for a while, and then seems to quit working.  Traders and system developers alike must ask whether it is a temporary change in the markets or whether it is time to find a new niche or method.  Steenbarger’s book provides a lot of practical advice on this question.

At "A Dash" we believe that there are different eras of trading.  We believe that long-term sentiment is helpful in defining these eras.  Summarized briefly, these include the following:

  • 1980 – 1998, where money flowed into mutual funds, there was sector trend persistence measured in months, and market valuations followed the stock/bond comparison reflected in the Fed Model, as we described in our earlier analysis.
  • 1999-2001, where some mutual funds pursued high momentum strategies with large commitments to a small number of stocks, where newbie investors jumped on IPO’s and Internet stocks, where some "real companies" who served the Internet IPO’s had temporarily inflated prospects, and where normal valuation measures were suspended.
  • 2002 to date, where mutual funds investing in U.S. equities experienced continual outflows, where new money rushed to real estate, emerging markets, and hedge funds, and where normal valuations measures were (once again) suspended.

What we expect to demonstrate (in forthcoming posts – one page at a time) is that some things should be "tweaked" and others should not.

When to Tweak

In our own trading, we always evaluate and update our short-term models to reflect market conditions.  Trading, therefore, should be tweaked to fit current conditions.

The fundamentals of valuation — forward earnings of stocks versus the return from bonds or other assets — do not really change.  (While we are not wedded to this specific approach, we use the Fed Model as a good indicator of the general condition.)

Please consider this more carefully.

A model may be either empirical (descriptive), normative (prescriptive), or both.  The Fed Model was excellent in both descriptive and prescriptive terms for many years.  When it failed to "explain" market behavior in the Internet Bubble era, many big-firm analysts thought they had identified a new paradigm — an era where super growth and productivity rendered the old models obsolete.  They started "tweaking their models" to force them to show that the current market prices were logical.

With the benefit of hindsight, we know that this was wrong.  Market prices were a reflection of (irrational) sentiment, not fundamental value.

Should a modeler "tweak the Fed Model" to reflect this?

Our answer is a resounding "NO!"

Why take a model that provided great advice, and tweak it until it fit what really happened?

Valuation models based upon logical principles are prescriptive (always) and descriptive (usually).  Anyone following the Fed Model principles would have traded cautiously through the Bubble Era (as we did) and understood that the market could decline — not just a little, but by a lot.

Many traders are schooled to think in terms of "days of decline", time since the last bounce, etc.  These traders, relying on the concept of local efficiency, did not realize how over-valued the market really was.  There are many sad stories of traders and managers who held positions that lost 90% of their value.

What about Now?

Current Fed Model readings show that we now experience the opposite of the Bubble Era effects.  Should we be "tweaking the model" to fit the current data?

Regular readers of "A Dash" know that we are skeptical of tweaking valuation models.  It is an interesting question, and one which we shall explore more carefully.  Our preliminary conclusion is to adjust short-term models for trading purposes, but to adhere to long-term valuation methods.  Deviation from long-term models (the Fed Model and similar methods for asset allocation) is better viewed as a measure of sentiment.

We understand that many (most?) analysts disagree with this view.  We shall consider some of the proposed "tweaks" to valuation models in future posts. 

You may also like


  • Jason Ng February 11, 2007  

    Nice article. I find that every trading strategies or systems has its own cycles too. It cannot be expected for any systems to have a homogenous performance throughout the year as the stock market is not homogenous too. A reliable system should therefore be evaluated on its annual net returns instead of whether it seems to be waning or not.
    This is, of course, true only for systems that are created for every market conditions. I have seen systems created only for the good old pre-2000 bull market which completely collaspe in this volatile market condition today.

  • RB February 12, 2007  

    Look forward to the rest. In a synchronous global economy where there is a flight to safe assets at the end of the cycle, one could perhaps expect valuations to return to trend in accordance with the Fed model.

  • Bill a.k.a. NO DooDahs! February 12, 2007  

    This points out one advantage of the multi-strategy approach, provided that the detrended equity curves of the strategies are uncorrelated and that the portfolio is rebalanced. If the strategies in question are close to each other in terms of total return, then the combination may provide actual improvements in total return instead of just reduction of risk.
    Deviation from rational valuation methods is indeed a measure of sentiment. This applies to individual stocks as well (“value” investing).