Saturday, August 12, 2006

Trading With the NYSE TICK - Part Three

In the first post in this series, I showed how the NYSE TICK fits into a common intraday trading setup. The second in the series illustrated how an ongoing assessment of the distribution of NYSE TICK values during the day provides a useful tracking of trader sentiment. In this final installment in the series, we'll take a multi-day look at the value of the NYSE TICK as a trading tool.

For this study, I use the Adjusted TICK statistic reported daily on the Trading Psychology Weblog. This is a daily average of the moment-to-moment TICK readings, adjusted to create a zero mean. As a result, positive numbers mean that there has been net buying sentiment on the day; negative numbers indicate the reverse. The correlation between this daily Adjusted TICK reading and daily price change in SPY has been over .70, suggesting that about half of all variance in price can be attributed to trader activity at the bid vs. offer across the universe of NYSE stocks.

Since July, 2003 (N = 774 trading days), I found 63 occasions in which the S&P 500 Index (SPY) was up by more than 1% on the day. Four days later, SPY was up on average by .07% (35 up, 28 down). That represents no bullish edge whatsoever relative to the average four-day SPY gain of .14% (432 up, 342 down) for the entire sample.

Now, however, let's factor the NYSE TICK into the mix. When SPY has been up by more than 1% *and* the NYSE TICK has been strong (N = 31), the next four days in SPY average a gain of .39% (22 up, 9 down). When SPY has been up by more than 1% and the TICK has been weak (N = 32), the next four days in SPY have averaged a loss of -.24% (13 up, 19 down). Clearly, the TICK makes a difference: a single day's TICK reading has bullish or bearish implications four days out.

How about when the S&P 500 (SPY) is weak on the day? We had 67 occasions in which SPY has been down by more than 1% in a single day since July, 2003. When the SPY was weak and the NYSE TICK was relatively strong (N = 34), the next four days in SPY average a gain of .39% (21 up, 13 down). When SPY was weak *and* the TICK was weak (N = 33), the next four days in SPY averaged a gain of only .03% (16 up, 17 down). Once again, we see that a single day's TICK reading exerts an influence several days out.

Now let's look at the TICK over multiple days. When the adjusted NYSE TICK averages more than +500 over a four-day period (N = 36), the *next* four days in SPY average a gain of .39% (23 up, 13 down). That is much stronger than the average four-day gain of .14%, as noted above. When the TICK averages less than -500 over a four-day period (N = 34), the next four days in SPY average a gain of .48% (23 up, 11 down)--again much stronger than average.

What we see is that very positive trader sentiment over a several day period tends to generate strength over the next several days, but very negative sentiment tends to lead to reversals. The bottom line is that the moment-to-moment lifting of offers and hitting of bids among traders does make a difference--even on a multi-day time frame.

4 comments:

yinTrader said...

Hi Brett

Your observations of trading with the NYSE tick v SPX tick come close to predicting trends for the next 4 days, generally.

Coupled with the S&P 500 Market Probability of the Almanac, we have some good indicators to go by to take a trade in SPX.

What is your take on this Probability for SPX?

Brett Steenbarger, Ph.D. said...

Hi,

Thanks for the note and for your interest. I am currently in the process of developing a multivariate model for the S&P 500 that utilizes several market psychology variables--including a version of the NYSE TICK--as inputs. Further information is on the Trader Performance page of my personal site (www.brettsteenbarger.com). Eventually the outputs from the model will be posted daily to the Trading Psychology Weblog.

Brett

John Wheatcroft said...

When you say "the next four days in SPY average a gain of .48%" is there any difference between day 1 and day 4 in the reading i.e. does price seem to ascend into the fourth day more often than not or descend into the fourth day? I am interested in looking at that factor as a subset of the finding.

For example - if on day 1 (of 4) the return is less than current average (.14) will it be greater than .48 on day 4? Or if there is an immediate jump on day 1 to .56 (for example) does it then descend?

Secondly, if you look at the average of the 4 day return as your primary results factor then what was the 4 day return on the day the tick reading suggested an above average return to be expected? If that was below average (.14) wouldn't that provide a stronger case for an above average return over the next 4 days?

I hope you understand my questions. If they aren't pertinent to the study then don't bother with them.

Brett Steenbarger, Ph.D. said...

Hi John,

Thanks; you're asking excellent questions. Everything you're mentioning is a subanalysis of results that I will commonly carry out before utilizing a pattern in my own trading. For instance, a pattern may have a favorable four-day outlook, but no edge one-day out. This would affect my bigger picture thinking as an intraday trader, but might not influence my trades that upcoming session.

Your other example points to the importance of controlling for hidden variables in the patterns. I do try to screen for that, particularly the overlap between price change and indicator readings.

When I begin posting my Micropsychology Modeler results on my personal site, I will include many of these considerations, including returns at both next day levels and further out. In that sense, the patterns I identify here are really a starting point for analysis and research, not mechanical trading signals.

Brett