Sunday, August 20, 2006

What Trend Research Suggests for the Coming Week's Trade

In a recent post, I raised anew the concern that trading strategies in the S&P 500 Index that are based on trend following (buying strength, selling weakness) have been notably unsuccessful. I attributed this, in part, to the expanded arbitrage trade between the futures and cash, between different futures products, and between futures and ETFs.

Let's see if this phenomenon affects S&P 500 Index ($SPX) expectations for the coming week.

The past five trading sessions have certainly constituted an upward trend, as we've closed higher each day. So let's go back to January, 1990 (N = 4190 trading days) and see what has happened after $SPX has made a five-day high in the previous session and made a five-day low five days prior (N = 729).

Four days later, the market was down by an average -.01% (371 up, 358 down). That is appreciably weaker than the average two-day gain of .14% (2314 up, 1876 down). If we limit the data to 2003-present (N = 153), the results are almost identical, with an average four-day change of -.01% and 76 occasions up, 77 down. I also note that, since 2003, when we've had an upward five-day trend, the returns over the next two sessions have been particularly subnormal, with an average loss of -.08% (72 up, 81 down). That is quite a bit weaker than the average two-day gain of .07% (2239 up, 1951 down) for the entire sample.

Think about what that means: Since 1990 (and during the recent bull market), traders have lost money on average when they have bought the market after a five-day period of upward trending.

Let's add a couple of limiting conditions, however.

First we'll look at five-day bullish trending periods that have also terminated on 20 day price highs (N = 489). That's what we have as of Friday's close. Now we see that the next four days in $SPX average a loss of -.07% (239 up, 250 down)--truly weaker than the average four-day change noted above. Another way of looking at this is that a five-day uptrend within a 20-day uptrend yields even weaker returns than five-day uptrends that do not terminate in 20-day highs (N = 240; average price change .13%; 132 up, 108 down). The more pronounced the market uptrend, the worse the returns going forward.

Finally, consider the VIX and what happens when our five-day uptrend occurs with a VIX below 15 (N = 269). Here again we see the pattern of subnormal returns four days out, with the average change -.01% (140 up, 129 down).

And when we have a five-day uptrend with a low VIX *and* a 20-day high (N = 188; as we saw on Friday)? Four days out, $SPX averages an abysmal -.11% (88 up, 100 down). Since 2003, that four-day return under those conditions (N = 61) has been -.27% (26 up, 35 down)!

Buying strength has lost traders money. Buying persistent strength has lost traders more money. Buying persistent strength since 2003 has lost still more money. I will not act upon this information this week in a mechanical fashion, but you can be sure that I'll be looking for sell setups where buyers lifting offers cannot move the market higher. As noted in my most recent post, I will be posting those intraday observations here during the new mid-morning updates. Also make sure you take a look at the Micropsychology Modeler results to be posted this evening on the Trading Psychology Weblog.

2 comments:

NStone said...

I think your insights into market behaviour are excellent and thank you for your effort. Your market analysis is the only kind in my opinion that will ever be useful in making money in the markets. Thanks

Brett Steenbarger, Ph.D. said...

Hi,

Thanks for your kind words of praise. I do find the historical analysis of market tendencies to be valuable, but I have too much experience with too many successful traders to ever think that my way is the only way, or even the best. How a successful scalper processes market information is so different from how a successful long-term investor operates, yet both can make money.

Success in trading, IMO, boils down to pattern recognition. There are probably as many ways of recognizing patterns as there are ways of processing information. Where I find historical analysis particularly helpful is in tracking occasions in which markets are changing their cycles, moving between trending and non-trending and between volatile and non-volatile.

Trading fixed patterns in a dynamic marketplace invites eventual obsolescence, IMO.

Thanks again--

Brett