TrendCatcher2 Revisited

The primary indicator behind the intraday TrendCatcher strategy is TICKscore, an indicator that utilizes intraday NYSE TICK readings to identify when institutional investors are active participants. Over the past few months, with the explosion in volatility, I (and my subscribers) have been less than impressed with the performance of the strategy. It posted its first back-to-back losing months on record, and losing trades have outnumbered winners over the past six months. That’s significantly worse than its historical performance. At first, I posted a column last week in which I filtered the strategy’s performance through my own short-term bias. This improved the results significantly, but it wasn’t really what I was looking for as the strategy was always intended to be purely mechanical. Once the trades are filtered through my own bias, it becomes a quasi-discretionary approach. That’s not necessarily a bad thing, but I wasn’t satisfied. After all, there’s no point in following a strategy if it doesn’t work on its own. I originally assumed the strategy needed another input. It’s based on only two inputs – breadth & TICKscore, so a third input (such as relative volume – a topic recently covered by Brett Steenbarger in his January 27th column) seemed like a good starting point. While compiling the data necessary to research that indicator, I decided to conduct a thorough review of the historical TICKscore data we already have on file. After all, if this primary element of the strategy does not work on its own, there’s little reason to craft a strategy around it.

To gauge the effectiveness of the TICKscore indicator on its own, I stripped the original strategy of the breadth input as well as all rules regarding the entry, stop, etc. I then simulated trading the S&P based solely on this one indicator. One of the first things I noticed is that in general, if TICKscore started the day higher, it usually continued to move higher throughout the day, and if it started lower, a persistent move lower usually followed. In other words, whichever direction institutions initially leaned didn’t normally change during the day. That was a major wake-up call. Rather than wait for confirmation of heavy institutional participation, which was the premise behind the original TrendCatcher strategy, it appeared the better idea was to simply note, as quickly as possible, whether big money was buying or selling and to latch on for the ride.

A second point was also evident. A good trend day sees institutions solely on one side of the market, buying or selling. If there’s evidence that they’re shifting sides during the day, it’s best to simply exit the position immediately. Remember that the goal of the strategy is to catch trend days, and the odds that we’re experiencing a trend day drop significantly when institutions are both buyers and sellers. With those two points in mind, I began a series of tests utilizing just the TICKscore indicator to trade the S&P, entering as soon as it was evident which direction institutions were leaning and exiting at the first sign that their conviction was wavering. The results were surprisingly effective.

This stripped-down version of the TrendCatcher strategy outperformed the original strategy on all counts over the past six months. Particularly encouraging was the fact that the strategy profited in 54 out of 91 trades, or 59% of the time. In the same time frame, the original TrendCatcher profited on just 38 out of 84 trades, or 45% of the time. Total gain over the six-month period was also significantly better for the stripped-down TrendCatcher2 (see equity curve), coming in at 35% vs 15% for the original TrendCatcher. The equity curve also highlights the small drawdown, particularly when you compare it with the original strategy’s equity curve. Average profit and average loss both declined, given that this strategy exits much quicker – average profit came in at 1.1% vs an average loss of 0.7%. As a means of comparing it with the original strategy’s track record, here’s a rundown of the performance during the past six months based on S&P points gained/lost… August +33, September +104, October +103, November +49, December +14, January +24, February -5. The table below lists every trade.

While I fully plan on reviewing and posting the entire four-year track record as quickly as possible, I have enough confidence in the results that I’ll be posting TrendCatcher2 signals in lieu of the original signals going forward.

(for those sick of these TrendCatcher posts, I promise this is the last for awhile.)

Categories : Archived EOD Commentary, End-of-Day Commentary

 Print This Post