Tuesday, May 15, 2007

Thoughts on a Weak NYSE TICK and Strong Patterns

Monday's trading saw significant selling in the broad market. My cumulative adjusted NYSE TICK reading for the day was the lowest since March 13th. The "stealth correction" that I had noted in the Weblog broadened into more noticeable weakness.

Recall that I utilize the NYSE TICK as a sentiment measure reflecting the willingness of traders to execute at the market offer vs. bid across the broad range of listed stocks. I watch for shifts in the distribution of NYSE TICK readings to alert me to changes in the sentiment of large market participants. Such a shift alerted us to the downside breakout from a narrow range on Monday, and it tells us that sentiment is turning more negative on a short-term basis.

But what does it tell us if the Dow Jones Industrial Average (DIA) rises on a day in which we have significant selling sentiment?

I went back to the start of 2005 (N = 589 trading days) and found 121 occasions in which we hit an NYSE TICK reading of -1000 or less during the trading day. On only 28 of those instances did we see a flat or rising Dow on the day. Five days later, the Dow was up by an average of .92% (22 up, 6 down). That is considerably stronger than the average five-day Dow gain of .20% (346 up, 243 down) across the entire sample.

In sum, when the Dow has held up in the face of selling pressure, we've tended to see further strength in the near term. Of course, I'll need to see a drying up of negative TICK readings intraday and a holding of the Monday lows to act on this pattern. It's the kind of pattern that would lead me to book some profits intraday, but also leave a piece of the position on to catch a larger move.

The best setups are those that combine a multi-day edge with intraday confirmation; clear, conceptually grounded stops (at the Monday lows); and equally clear profit targets (the S&P highs). As any baseball batter knows, if you only swing at good pitches and take enough quality cuts at the ball, you'll eventually make contact and get your hits. With the market weak in the pre-opening action as I write, we may not get that good pitch to hit from this pattern.

But that might lead to another pattern. And an even better pitch.

Sometimes laying off the high outside stuff is the best thing a batter can do.

8 comments:

Johan said...

"I went back to the start of 2005 (N = 589 trading days) and found 121 occasions in which we hit an NYSE TICK reading of -1000 or less during the trading day. On only 28 of those instances did we see a flat or rising Dow on the day. Five days later, the Dow was up by an average of .92% (22 up, 6 down). That is considerably stronger than the average five-day Dow gain of .20% (346 up, 243 down) across the entire sample."

Did you look for different periods before coming up with 5 days or did you chose it by random? If you go through a whole lot of different periods like checking after 1,2,3,4,5,10,20 days, then you will surely find at least one anomaly. That is with is called curve-fitting and will not give any predictive value.

This question is not only regarding this post but all other posts you make and other reaserachers such as "marketsentiment" and "markettells".

The quality of the statistics goes up quite a bit if you put in more than one time period. For instance to check if the index has outperformed both in 3 days and 5 days. Or so I believe, I am not a statistician.

I have seen those kind of stats a few times and think they must have better predictive value.

Any statistician here who can help me out?

Also, thanks for the great posts Brett!

Regards,
Johan from Sweden

Brett Steenbarger, Ph.D. said...

Hi Johan,

I agree that there's a problem with making multiple comparisons in the data and then reporting on the one in 20 that might be "significant" by chance. That's why testing and reporting different time periods as you suggest later in your comment might be problematic.

I don't select the time period by random, and I don't test "a whole lot of different periods" either. I eyeball the data and select a period that seems appropriate. I'm not conducting formal statistical significance tests, but I also don't want to go on data fishing expeditions.

My goal is simply to show how the market has recently behaved under a given set of conditions. Thanks for the opportunity to clarify--

Brett

Ziad said...

i noticed the TICK distribution becoming quite negative between 10:30 and 12:00 EST, but i also noticed that the ES was actually flat and posting slightly higher lows. I took that to mean inefficiency, as 1.5 hours of negative tick reading couldnt push it lower. However, what it truly was indicating was weakness to come. Where did i go wrong?

Brett Steenbarger, Ph.D. said...

Hi Ziad,

I see how you interpreted that. When we got a new high in the Dow and not the other indices, and when the TICK distribution started shifting downward after around 9:15 AM CT, it didn't seem as if the rally was going to broaden out and extend. There was a lot more selling in that 10 AM CT period than normally should have been the case in an upward directional move.

Brett

Omar said...

Hi Brett,

I am still unsure of how the adjusted TICK is formulated. I understand it is the average TICK price per minute minus the 20 day TICK average cumulated to form an end of day index.

Could you kindly clarify a few things to me,
1. Is the 20 day average a minute average ?
2. Is the adjusted TICK's sole purpose to evaluate sentiment at the end of the trading day and when compared to previous days provide ideas for the following day? If so do you track The NYSE TICK intra day via the raw TICK OHLC chart or through an average e.g the emerging average.

I have gained so much knowledge since discovering your blog site not so long ago. One of the first things i did, based on your advice, was study Market Profile and AMT.

Thanks for all your help,

Omar

Brett Steenbarger, Ph.D. said...

Hi Omar,

The Adjusted TICK is calculated each minute during the trading day. I track shifts in the distribution of the Adjusted TICK values to identify emerging buying or selling sentiment. Today's trading (Tuesday, 5/15) was a good example of a downward shift in the distribution as the session wore on. I find the cumulative line of the Adjusted TICK--shifts in its slope and direction--to be helpful on an intraday basis.

Brett

db said...

There was a theory about the Dow rising when the rest of the market was selling off. That is there are funds required to be full invested, and that they move to larger cap stocks when they sense a down market move approaching because large caps tend to drop less percentage wise in a bear market than smaller caps.

Admittedly, a lot to take on faith in that statement.

Brett Steenbarger, Ph.D. said...

Hi DB,

I agree; the focus on large caps generally reflects a defensive strategy--

Brett