This entry was posted on Thursday, October 11th, 2007 at 7:29 am and is filed under Education, system trading. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.
This morning I came across an article by Alligator Investor on the Seeking Alpha website noting a trading system in the Dow Jones Industrials over the past 50 year. In the article he mentions a commonly mentioned system of using a moving average crossover system that uses the 50 and 200 day moving average. When the 50 day crosses above the 200 day moving average we buy and when it moves below we exit. Although I think there is some merit to the basic premise of the article I think it leaves the reader with the impression that the system did pretty well overall.
The problem is that over time it leaves you way way behind. The data and charts I am about to show you do not even include dividends or the effect of short term tax consequences. Looking at the last 50 years as he did in the article let us compare buy and hold versus usage of that system.
And here is a visual of what that looks like in your portfolio.
That is not good.
Hey why stop at 50 years? Lets go back to 1919 !
Number wise this is bad enough. But this next visual is very enlightening.
Ok your so far behind - your not even in the race.
Here is a PDF with all the trades listed
Surprising? I think it is when you see it in such stark contrast of data and pictures. Backtesting allows that. It allows you to challenge the conventional wisdom that permeates the trading environment. When people ask me “Dave what does work?” I tend to say things like “it’s the opposite” or “not what you may think”. I say these things because so much of what I have tested through the years has been so counter-intuitive to what I had been taught up until that time. It cuts to the core of your belief systems. It’s that knowledge that now allows me to feel comfortable stepping away from the crowd.
As an example of being totally crazy lets change the test a bit. We will not buy on a cross up - but a crossunder of the 50 day moving average and we will use the same exit as the first test- a cross under of the 50 below the 200 day moving average. So our system will buy the so called “Death Cross“. Yes the dreaded Death Cross. Those are considered very very bearish. So we buy that and wait for a cross up and use the crossunder as an exit. That has to have much worse performance right?
That is the past 50 years. So let me get this straight. Drawdown the same. Return of 5.15% in Death Cross system and 4.42% in the “good/safe” system I showed you above.
The purpose of this post is to make you look at the markets differently. Remember what Dr. Koch mentioned in his interview with me?
The Trading Digest - Can you tell me any consistent themes you find are more successful in trading systems? In what timeframes?
Dr. Koch - To tell you the truth: after years of research in the stock market I
know one single principle which works consistently: The effects of fear
in the markets. In terms of trading systems this is the pullback, bottom
fisher or mean reversal which in effect talk about the same statistical
effect. You can find this effect in hours, days, weeks, months and even
years. I found the most interesting time frame in the range of 2 to 5
days. Let me emphasize that this statement means at the same time: None
of the so called classical technical indicators works. No RSI, no MACD.
none of them.
Did you catch that? He said “The effects of fear in the markets. In terms of trading systems this is the pullback, bottom fisher or mean reversal which in effect talk about the same statistical effect. You can find this effect in hours, days, weeks, months and even years.”
Do you think something we call the “Death Cross” might be fear creeping into the market at that time? Sure is. And as I have shown you here today thats an edge over conventional wisdom.
Have a Great Day!
Dave Johnson
October 11th, 2007 at 7:47 am
Hi Dave,
very nice article. Looks like these MA-Crossovers will never die…
One remark: You should display long term charts with an log scale always. This scale makes Percentage changes the same size no matter where prices are.
October 11th, 2007 at 8:21 am
Thanks for the note on using the log scale. In my haste to get the article published I ended up snapshotting the non-log one.
October 12th, 2007 at 2:16 pm
Do you know what tools, if any, Dr. Koch uses to identify “fear,” e.g., VIX, p/c ratio, standard deviation, etc?
October 12th, 2007 at 2:33 pm
Steve thats a great question. Let’s see if he responds here. Otherwise I will email him.
October 12th, 2007 at 2:49 pm
Please note: “fear in the markets” is already an interpretation. Same for “panic” and similar terms. The only thing that counts: after a “strong” down move, there is a higher than 50% probability for up moves. This is observable on various timeframes, the term “strong” meaning something different in each timeframe, of course.
Just plot past price changes vs. future price changes and you’ll see this “pattern”.
May 22nd, 2008 at 8:21 am
Did you include the contributions from holding cash during the times the crossover system was out of the market?
Did you evaluate the crossover results from a risk-adjusted perspective, perhaps that of an investor who needs to withdraw funds to live?
May 22nd, 2008 at 9:07 am
Bill in the test I only used price. No dividend or cash allocations were made.
May 22nd, 2008 at 9:30 am
Then it’s an incomplete test, isn’t it? Dividends won’t make too much difference, neither will taxes, if the system trades infrequently, since most gains will be “capital gains” in a slow-moving trend system.
Very few timing systems on the major stock indices outperform buy+hold over long period on a “cash to the mattress” basis, when CAGR is the only metric. A proper comparison for any in/out system should include the returns of the competing investment, in this case, cash in a money market account, 30- or 90-day T’s, etc.
It should also measure the volatility-adjusted results, since that’s the primary purpose of timing for many participants, avoiding volatility. Kelly Criterion is optimal position sizing from a CAGR standpoint, but I doubt many people would view its drawdown characteristics as “optimal” with their own money at risk.
That may be why Dr. Koch hasn’t found trend-following systems to work; he hasn’t tested them properly (accounting for alternate returns), and hasn’t applied proper metrics for comparing results. Very few people are interested in CAGR only, and many that say they are, won’t be anymore, after a big account drawdown.
There’s a certain arrogance in saying “I haven’t been able to see this work, ergo, it doesn’t” - especially when many other people HAVE seen it work. I used to be guilty of that sin, and have been trying to root it out for the past few years.
Aside from that, 50/200 simple moving average isn’t optimal in my testing. I found the general area of 80-120 and 160-200, with exponential moving averages, to be a bit better. The main benefit is avoiding prolonged bear markets.
FWIW, I agree that the fear concept is very useful, although it’s far from the only useful one.
May 22nd, 2008 at 9:37 am
Absolutely incomplete. Remember the article was based upon the original article I pointed to. Obviously a very simple test. My point really was to show how the opposite in it’s simplest form kept up with the “safe” concept.
Maybe you have the capability to incorporate dividends and cash holding?
Thanks for the comments.
May 22nd, 2008 at 9:55 am
Simple stuff like this you can do in a spreadsheet. Or whatever software you’re using, if you track the ins and outs, you can approximate it. Slow systems like the one shown will generally be out for months at a time, long enough to buy some T’s and hold to expiry.
Software like http://www.tradingblox.com/ (I’m not a user, but I like their forum) includes adjustments for divvies and splits on stocks, accounting for margin interest and risk-free rate, etc. It’s probably the best route for testing more complicated trend-following systems on a basket of leveraged issues, which is where probably most of the trend-followers make their money.
Another note: many of the classical indicators DO work, just not in their default parameters. 14-day RSI is whack, but 2-day RSI shows “fear” on the timeframes Koch mentioned. Using a percentage-based MACD on a longer parameter than default, one could assemble a relative strength portfolio on a basket of index constituents and outperform the index pretty handily, with relatively low turnover. Just because you haven’t made it work YOURSELF, doesn’t mean it WON’T work …
May 22nd, 2008 at 10:50 am
Bill I am not sure what your getting at. I trade many systems. Trend, counter trend, and everything in between.
And many timeframes from intraday to weekly bars.
http://dayvejohnson.blogspot.com/2005/10/hope-you-didnt-get-sucked-in.html
http://thetradingdigest.com/blog/2007/03/27/are-you-diversified-no-really-really-diversified/
Maybe those will give you a bit more insight to how I approach the market.
Thanks again.
May 22nd, 2008 at 11:05 am
It seems from Dr. Koch’s comments on this thread, and on the Wealth-Lab thread that you link to in another post, that he believes trend following systems don’t work, when clearly they DO work for some people, just not in his testing.
That’s what I was getting at; nothing specifically about you, but since the comments from Koch were here, I addressed them here. Sorry for any confusion that may have caused.
May 22nd, 2008 at 5:00 pm
Bill that is always where this discussion leads. Defining “works”. There are scores of academic papers on the subject. By trying to take the argument of it doesn’t work leads me into a whole myriad of issues we must first define and quantify. Then we have to agree on a reasonable test, then we have to answer a zillion questions of data integrity, then I have to get the noobies up to speed. I honestly don’t have the time or desire to even go there. Suffice it to say each trader has his tools and in the end only performance matters. I’ll leave the does or does not work to the academics.