Saturday, August 25, 2007

Staying on the Right Side of a Stock Market Trend: The Power Measure

A reader asked me to comment on Friday's market action, which started a bit weak on low volume and gradually morphed into a rally.

I thought I'd use the occasion to illustrate a broader point that pertains to Friday's trade, but can also be used across various markets and time frames.

In the chart above, we see the 5-minute closes for the ES futures (pink) and a 20-bar Power Measure.

The Power Measure is a moving correlation between the size of a bar (its high-low percentage differential) and the price change of that bar (open to close). The correlation is computed with a moving lookback period, which in this case is 20 five-minute periods, or 100 minutes.

What we're looking for with the Power Measure is what I call trendiness: the relationship between volatility (bar size) and directional price movement. In a trend, we see a correlation (positive or negative) that is becoming more extreme over time and that stays in positive or negative territory for an extended period. In a range bound market, the Power Measure will behave as a kind of overbought/oversold indicator and can be used as an aid in fading range extremes.

A very important relationship: In longer-term bull swings, dips in the Power Measure will occur at successive price highs. In longer-term bear swings, peaks in the Power Measure will occur at successive price lows. An extreme reading in the Power Measure that is not associated with a corresponding new high or low in price is often a great opportunity for a fade. That occurred around Point 200 on the chart, which occurred around 1:00 PM CT on 8/22: we hit a negative extreme to the downside in the Power Measure, but price was holding above the low registered at the prior Power dip.

Early in the day Friday, we looked as though we were topping out in the Power Measure, but notice that the rally on the housing news pushed us right back into positive territory (red arrow). If you look at my Twitter comment at the time, I mentioned that this switched us into a buy-the-dip mode. The reason for that should be clear from the Power Measure: there simply were not enough sellers to sustain a negative relationship between price change and volatility. (We could see that from the absence of very weak NYSE TICK figures as well). To put it in simple terms, the big bars on the chart belonged more to buyers than sellers. That's what makes a trend.

One advantage of the Power Measure is that it can be computed for any time frame and any market in which you have open-high-low-close data. I use it, not so much as a timing tool, as a general reference to tell me whether we are gaining or losing trendiness at various time frames. As the link below explains, I generally am loathe to buy markets that are at high positive readings or to sell markets that are at extreme negative readings unless we're relatively early in a breakout mode. Otherwise, it pays to let the sellers or buyers take their turn and show their inability to move the market to successive fresh lows or highs.

Note: All calculations and charting is within Excel. I don't know of any commercial data services that calculate a Power Measure.


Power Measure and Countertrend Bias


dan said...

Excellent post (as always)! Could you go into more detail regarding how the Power Measure is calculated? Are you calculating (close - open) / (high - low) of the bar and then calculating a correlation coefficient of all such calculations over the previous 20 periods? I assume you are only using the pit session for calculation purposes.

Thanks again for your mind expanding material.


Brett Steenbarger, Ph.D. said...

Hi Dan,

Thanks for the note. I'm simply correlating the size of bars (high-low) with the price changes for those bars (open-close). Both are expressed as percentages and the correlation is for a moving lookback period. I just updated my Trader Performance page with another chart to illustrate:

I am currently working on a new iteration of this indicator, but it is still under development/research.


dan said...

I'm confused how to express the range in terms of a percentage. Is this just the high divided by the low (high/low)? Thanks for your help.


Brett Steenbarger, Ph.D. said...

Hi Dan,

I'm looking at the difference in price between the high and low price as a proportion of the low price.


frankwang60 said...


Let me try to understand this. You are correlating (open-close)/close and (high-low)/close?