Ripple Effect: Finding Patterns in the Chaos

• Posted by DynamicHedge
• on September 5th, 2013

Events occur every day that alter the sentiment of the investing public. Every large natural disaster, unexpected central bank action, or a closely scrutinized company shocking the analyst community is a reminder of what makes the market so difficult to model. Most traders and investors spend their time trying to predict these significant events, and it’s a shame because they’re just too random. The good news is that unpredictable events create the conditions in which predictable opportunities exist. The reason? Big events act like a tanker ship cutting through the sea and leaving a wide frothy wake of open PnL gains and losses behind them. Why is this relevant? People behave in very predictable ways when faced with large unexpected winnings or big unrealized losses after a market moving event. In the end, events are difficult to model and predict, but the human reaction to those events are less difficult to predict. With the right tools, it’s actually kinda easy to see the patterns.

Let me explain in an example.

Last week the market was down more than -1.25% in a single session (August 27, 2013). The headlines read: U.S Stocks Drop On Syria Tensions. Let’s stop and consider that the reason the market moves doesn’t really matter — only the fact that it did move. People sell stocks because they anticipate losing money — they either want to keep the money they’ve made, or they don’t want to lose any more money. So, price performance action itself is more important than the actual news. Back to the example, losing -1.25% in a day is a somewhat rare occurrence and is usually enough to get people stopped out of positions or at least change the short-term sentiment. When combined with a market that has already suffered a recent downtrend, the conditions become much more unique and severe for sentiment. These were the conditions that the market was in on the week when the “war jitters” appeared. The essential question then is how have market participants behaved after these conditions occurred?

Here are the settings I punched into my system. NOTE: The conditions for a downtrend are at least nine day under a 20-day moving average.

 

 

Here’s what a random sample from May 2012 looks like to make sure we are matching apples to apples. NOTE: The grey shading is an area which matches the moving average condition, the blue area shows the condition which matches both criteria:

 

Now what we want to do is to analyze five days after the blue highlighted area over the last four years. Shown below is the same chart, but the end of the dark grey area represents the end of the analysis period. We are not looking for a long-term hold, just trying to harness the aftershocks of a severe sentiment event.

 

Good. So now we’ve set the stage for a particular condition. Let’s take a look at the statistics for this particular setup (after the -1.25 loss has already occurred).

 

So, now we know that this trade has a positive expectancy and the results indicate that the profits from the trade taken on the long side are statistically significantly different from zero. Big. Freaking. Deal. In trading, statistics alone don’t mean much. There is a statistic for everything. Sure, they can provide a framework of what to expect and how big the max draw down has been, but our brains don’t work well with numbers. FACT: a -3% open loss after a sustained downtrend feels a lot more sinister than it looks a sitting all innocent-like in a spreadsheet table.

We need more information about what to expect from a trade so that we can anticipate required action points in various situations. Statistics are two-dimensional, and traders need something more than numbers to put our big brain to use and add context to the situation. That’s right that big super computer between your ears that gets such a bad rap for being prone to bias is actually a pretty powerful ally when you feed good data into it. For that, we developed something called “Alpha Curves”.

Alpha Curves are graphical representations of the archetypal patterns the market creates. They represent a special blend of uniqueness and repeatability uncovered by our pattern recognition algorithms. These are the paths the market has traced out over and over as it created the statistical outcomes observed in a stats table. Results are path dependent, and Alpha Curves illuminate that path. My trading partner and I invented these little puppies and they will offer more insight into market behavior than any statistic ever could. Feel free to unleash the power of your brain and interpret these patterns because the algorithms have cleaned the majority of the bias out of the data.

Here are the Alpha Curves:

 

Without reading any further take a look at the shape of the curves and understand they are ordered by pattern dominance in the data. NOTE: The shaded area represents in-sample data (the down -1.25% day).

The patterns should be obvious at first glance or they don’t exist. First impression is all the patterns are positive. Second impression is the top two patterns digest the loss for a couple of days and then rally. The bottom two patterns rally quick and then retest lows before finally bouncing hard. The interpretation I get from all the patterns together is that most of the gains come near the end of the 5-day period. What action do I take? Wait — because you need more information.  If the market rallies I will take one action, if the market tanks I will take another. If this, then that.

Now let’s look at what actually happened.

 

Looks an awful lot like a market that bounced hard, retested the lows and made big gains at the end around the 5-day mark (there was also a holiday in there). Take a lok at Alpha Curve pattern 3. Not perfect but a helluva lot better than guessing.

If you’re interested in this type of thing I’m testing this software with a select group of traders and investors. We have a bit of a waiting list at this point, but if you want in, send a message to info@dynamichedge.com and I’ll get you sorted out.

Another good example of how to interpret Alpha Curves is my post on $MSFT: Ballmerisms and MSFT Gaps

blog comments powered by Disqus