My Photo
Note: Jeff does not accept guest blog posts on A Dash of Insight.

For inquiries regarding advertising and republication, contact

Follow Jeff on Twitter!

Enter your email address:

Delivered by FeedBurner


  • Seeking Alpha
    Seeking Alpha Certified
  • AllTopSites
    Alltop, all the top stories
  • iStockAnalyst
Talk Markets
Forexpros Contributor
Copyright 2005-2014
All Rights Reserved

« Grading the Fed: Was TAF a Success? | Main | Is the Fed Ahead of the Curve? »

December 20, 2007


Lloyd Nirenberg

You are correct in your general analysis. I would like to draw your attention to the fact that most ETFs and mutual funds are already very highly correlated (highly coupled), to such a degree it is really hard to construct a truly diversified portfolio (i.e., a portfolio with uncorrelated or even weakly correlated securities).

Evidence: I regularly publish a 64-security matrix of ETFs and indexes and compute their cross- correlations for all to see:

You will immediately see how beastly a task it is to get uncorrelated assets!


very interesting.
i'm adding in RSS Reader

Mike C

FWIW, I think Bill's point about mixing different strategies is an excellent one, and I would be interested (perhaps maybe the subject of a future post?) about how you approach mixing different strategies within an overall portfolio.

Perhaps some discussion of the specific strategies and how to divide amongst them (like say the TCA-ETF system versus your fundamental stock-picking)


Chris -
Thanks for your comment and also thanks to Bill. This discussion will help me sharpen up the writing in future efforts.

It is a real joy to have smart readers!



Jeff and Bill,

The combination of your two explanations made it extremely clear. I think Bill's summary is key because I was confusing activity with diversification in the same way that Blake was confusing them. Stating that "diversification is about mixing different strategies" makes it all make sense.

Bill aka NO DooDahs!

Summary: getting into the long run is about activity, and increased activity in a given strategy and given unit of time will get you into the long run faster (statistically measurable results), while diversification is about mixing different strategies.

Think of each strategy which is traded as being a series of trades, and each strategy has its own payout distribution, timeframe, asset class traded, etc. In this context, "buy an S&P 500 index fund and hold it forever" is a strategy, as is "buy the GSCI index ETF and hold it forever" or "buy the REIT ETF and hold it forever" or "buy GE and hold it forever." Another strategy may be Gil Blake's 50+ trades per year, holding each for a few days. Yet another might be the hypothetical casino taking a very small edge off of hundreds of Roulette spins every day.

Getting into the long run is all about having enough independent results to apply some statistical thought to whether the outcome is predictable, and whether the outcome is desirable. The statistical tests are dependent upon having a relatively large number of trades to be confident in the result. The *theory* is that as the number of trades increases, the likelihood that we are observing something resembling the actual system results increases.

For a Roulette wheel in a casino, after a week's worth of spins (probably thousands of spins), we're pretty confident that our take (dollars kept out of dollars wagered) is predictable, and one week will be darn close to the next week, etc.

For a system with 50 trades in a year, maybe we're confident after a year, and maybe not. Let's see the results and test them. Maybe after two years? When checking results on the typical buy/hold low turnover mutual fund manager, we might want several years to assess if the results are predictable and/or provably "better" than index returns.

Getting into the long run with a "buy and hold the SPY" strategy is tough. We have a LOT of variation in month to month returns, and it's only over many years that we get a convergence to something we might call predictable. I can tell you only a wide range of SPY results over a year that I will be confident with. However, I can tell you with pretty good confidence what a typical 50-year buy/hold result will be for the index.

Similarly, I can't tell you what one trade in an active system will pay off, but if I've analyzed the system thoroughly, I can tell you what 200 trades over 4 years might look like.

Diversification is something totally different. The only similarity is that both increased activity and increased diversification will tend to smooth out the profit/loss results over a unit of time, which is why I made the statement that I thought Blake had the two of them confused.

Diversification occurs when one mixes strategies. Holding the SPY, the GSCI ETF, and a REIT ETF would be a diversification. Since each strategy has long-term positive returns, and since each strategy has a low de-trended correlation to each other strategy, this has good diversification, and will lower the volatility of the portfolio, improving risk-adjusted returns. Another example would be if Gil traded his 50/year system with 1/3 of his money, a value investing campaign with 1/3 of his money, and gave the other 1/3 to Bill Gross to manage.

Summary: getting into the long run is about activity, and increased activity in a given strategy and given unit of time will get you into the long run faster (statistically measurable results), while diversification is about mixing different strategies.


Chris -

I really appreciate your questions. As you may know, I am trying to sharpen up many of these points for a future book. It is sometimes quite clear that I have not explained something well at all. Back in the classroom I could see some confused faces and try again. At "A Dash" I rely upon astute readers to tell me when I have not been clear!

Let me try again. Blake does not diversify. He does not believe in it. Instead, he has a system that has demonstrated edge, something he knows from his testing. This type of trading converges to the expected result (the long run) when the number of trials is sufficient. This can be simulated or tested mathematically.

We do both in our methods. If one views the "long run" in this way, it might be accomplished in a single day. There are many trading firms, making up much of the daily volume, that use such algorithms for intra-day trades. Some of them have an extremely high proportion of winning days, since they get into the "long run" every day.

Defining the long run is a matter of figuring out how many trials you get in a given time frame.

Is that explanation any better? Maybe Bill Rempel can add a thought.

Thanks again,



Ok, I think I get it now. You are using "into the long run," as in "in the long run, this system beats the S&P by 5 percentage points per year" or something like that.

I guess I'm still a little unclear on how this relates to diversification. Are you saying nothing more than that with more traditional diversification, where say you make 30 trades on day 1 and hold for a month, you see the benefits of the long run sooner? Whereas with Gil Blake's system, where maybe you make one trade per day for a month, the benefits of the long run may take longer to materialize? In the first scenario, you could presumably see the benefits of diversification on the first day, whereas you would have to wait at least two days to see diversification benefits with the Blake system.


Chris --
Humor is welcome here! We need more of it, and your comment is indeed worth a laugh.

Bill (who is both intelligent and informed about system development) is right on several fronts. I had to make a trip into the Wisconsin winter for the funeral of an old and dear friend.

Bill's answer is pretty much on target. Gil Blake famously invested his entire portfolio in a single Fidelity sector fund each day. Later, he used hourly trade data to take similar actions.

The idea here is that diversification is important to the average investor. We certainly diversify our individual accounts.

Should someone trading a system behave differently? One way to look at it -- Gil Blake's way -- is that you are applying the system in many different instances, prepared to adjust each day. If the system really works, you can get enough cases to converge on actual performance fairly quickly.

We are not "all in" on a given sector each day, but we recognize some correlation in the top choices. Since we evaluate and adjust daily, we are getting more decisions than we do in the long-term portfolios.

Your questions and comments have made it clear that I need to sharpen up this explanation, but I am pretty confident of the basic conclusion -- both from our testing and our results.

It is a joy to have intelligent and thoughtful readers who are willing to share their ideas.

Thanks for your comments.


Bill aka NO DooDahs!

I'm sure it was taken as 1/2 question and 1/2 joke (at least, I took it that way), but maybe give the folks some allowance for holidays to answer? Maybe Jeff and crew are out of town or something?

My take on the answer, one needs enough independent trades in a system to get a good idea of whether the system results are statistically sound, which is "into the long run." That's my interpretation, but I'm just a reader of this blog, not an author.

If you're a value trader, oops, "investor," and you cycle through annual turnover of less than 100% with transactions every quarter (maybe!), then most likely your trades are playing on the same themes for many months at a time and it will take many years to calculate some "by trade" statistics on the robustness of your system.

On the other hand, a "slow" day trader (no insult intended) may make 10-20 trades a month, each closed at the end of the day, and will have, within a few months, good statistics on their trades.

Just my take on it.


Sorry to sound so flippant (I still haven't learned that jokes don't go over well on the internet), but I really didn't understand what "getting into the long run" means...


I'm a little confused as to what you mean by "getting into the long run." Is that some sort of investment club? If so, how can I get it?

Bill aka NO DooDahs!

It sounds to me as if Gil Blake is confusing the benefits of diversification with the benefits of increased activity. Both have their merits, and assuming that the all other things are equal (such as the expectancy of the system), increased activity will reduce volatility of results over unit time (which is also a benefit of diversification). I think that both also have some overlap, since I find it hard to conceptualize using increased diversification in the same system without necessarily increasing the number of bets made per unit time.

It is interesting to look at trading similar systems in the same equity class, BUT ON DIFFERENT TIMEFRAMES, as a diversifier.

The comments to this entry are closed.