Here is an update of an indicator that I wrote about a while back. Suppose you create a channel defined by a moving average of the high prices for the past 10 trading days and a moving average of the low prices from the past 8 days. When you close above the upper band of the channel, you buy and go with the strength. When you close below the lower band of the channel, you sell and go with weakness. The position is closed when you re-enter the channel; no other stops.
According to Barchart. com, such a "system" would have produced eye-opening results for the S&P 500 Index (SPY) from March, 2004 to the present.
There would have been 248 completed trades, averaging about six days in holding time.
A total of 68 of the trades would have made money. 180 would have lost money.
The net number of points lost in SPY would have been over 100.
If someone had merely bought and held since March, 2004, the return would have been about zero. No significant gain or loss.
If they had followed this channel system, they would have lost all their money.
It is human nature to extrapolate the future from the recent past. This is one reason why traders who enter trades impulsively, needing action, invariably lose money.
The patterns that markets follow are not necessarily the ones that come to us intuitively. That is why knowing historical odds of markets moving up or down under various conditions can provide useful decision support.
Yesterday was a pretty strong day in the stock market, as we made a nice bottom this week. It was logical to want to go home long over the extended weekend. Maybe that will work; I see, however, that Vertical Solutions' backtested trading system went home short.
Buying strength and selling weakness, over time, has lost people money. Indeed, I see where MarketSci identifies "mean reversion" as "the most effective directional trade".
There are limits to human discretion: The best trades are not always the ones that feel best.
.
Subscribe to:
Post Comments (Atom)
Posts
5 comments:
I ran this system in Amibroker using the following code and found it to be a pretty consistent loser rather than winner.
Buy = Close > MA(H,10); //Buy if close above a 10MA of high prices
BuyPrice = Close; //buy the close
Sell = Close < MA(L,8); //Sell if close below a 8MA of low prices
SellPrice = Close; //sell the close
Short = Sell; //Supports shorts
Cover = Buy; // Support covers
Since Jan 3, 2005, there have been 127 trades, of which 35% were winners. Profit was -39%
Maybe I'm doing something wrong but this doesn't look that good to me. :)
Hi Damian,
Yes, your test pretty well follows the results that I report. As I mentioned in the initial post, the system is so bad that it's promising... ;-)
Brett
Since the system results were so dismal, my first thought is that it would be interesting to just do the reverse. e.g. go short when there is a breakout above the channel and go long when there is a breakdown below the channel.
You mention that selling strength and buying weakness gives a historical edge. Sounds like this test proves it again.
I do a lot of backtesting and many of my findings are counter intuitive, but not all.
I had to read the post twice (in different days) to finally come to grips with it. Aha!! now I follow what you're saying..
It is so counter-intuitive that it was staring at me and I couldn't figure out why you were so excited about a losing system. ;)
Gustavo
Reversing the system provides a system that wins 68% of the time - that's the way to go!
Post a Comment