Sorry for not posting this past week, but I have been struggling with some thoughts and plans for this blog. I have a very ambitious plan that involves laying out a foundation for technical analysis, sharing the tools I have found useful over the years, and also showing how these can be used in real-time. However, I want to be sure that once I start this project, I can commit to carrying it through on a regular schedule. In addition, I am trying to balance a planned course of attack on a deep subject with maintaining the ability to post on spontaneous and interesting topics as they arise… many conflicting interests to balance here, but I am close to figuring out how to do this. More on this soon, but today I want to share a small research project I did this week that reveals an important, but maybe not well-known, market tendency.
I spent several years focusing on and only trading the S&P futures. In that time, I did a lot of research on basic market tendencies, and (hopefully) also gained some intuition about the way this market moves. (As an interesting aside, I say I have only been trading stocks for the past 4 years, but maybe nearly 10 years of trading the S&P itself prepared me for this in some way.) Right after the market opens, I will often announce to the traders sitting near me something like “the first ticks are up, but that’s usually wrong.” This is an illustration of a tendency that is (as far as I know) specific to the stock market, that I like to call first ticks – wrong ticks.
Let’s drill down into this using the full history of the ES futures, which is 3,195 trading days beginning on 9/12/1997. When we try to quantify market tendencies, it makes sense to look at the long history, but also to focus in on smaller time windows to see if something has changed. Today’s stock market is very different from 1997, so let’s also examine the most recent trading year, extending back to 6/1/2009 which is 256 trading days. On one hand, it would seem to be very remarkable if we found tendencies that remained stable through years of bubbles and busts, through the structural changes of decimalization, the rise of electronic trading, HFT’s, algos, etc, but you might be surprised. Most of the tools I use work equally well on intraday charts of AAPL, daily stock prices from the 1700s or weekly grain prices from the 1300s, but that is a topic for another post. Throughout this current analysis, I will present stats first for the long history, then for the more recent window immediately following in parentheses. So, in this format, we are examining 3,195 (256) trading days extending back to 9/12/1997 (6/1/2009).
The ES futures reflect the price of the cash S&P index, plus an amortized risk-free rate, minus the dividend yield on the S&P 500. In normal conditions, this is a premium of a few points that converges to the cash value as the futures expire (quarterly), but futures currently trade at a small discount since the risk-free rate is so small. The minimum tick size is .25 of a cash S&P point, which corresponds to roughly .05 on the SPY. (I find the SPY to be a very poor market proxy due to the extreme level of noise.) First question: how many days have an immediate directional drive off the open, as opposed to trading in a noisy range? We can get a rough answer to this by looking at one minute bars and keeping all of the bars that have the open within one tick of the high or low. The assumption would be that one tick is noise, and the market went into an immediate directional drive off the opening print. We find that this happens 65.9% (64.8%) of all trading days in our sample, which might at first seem to be surprisingly large. (It is not at all surprising if you understand the Arcsine Law of Brownian motion and how it applies to random walks, but again, that’s another (even more nerdy) topic.)
So, we know that 2/3 of the time we will see a directional drive off the opening print. The second question to ask is, what should we do about it? Does it make sense if the broad market explodes off the open to pay offers and chase strength? Should we fade it? Do nothing? To really answer this question we would need to do an in-depth research project on tick-level data, but let’s try to tease something out with a back-of-the-envelope calculation sticking to our one minute bars. Let’s first look at how many times the market reverses, and trades on the other side of the opening tick, within the first 30 minutes. We find this happens 84.3% (80.1%) of the time, which is a very interesting statistic. Furthermore, we find that the average adverse excursion (how far the “fade trade” would move against us) is only .47 (.51) S&P points, and the largest move against us would have been 5.0 (3.0) full S&P points on 10/24/2008. Note that this statistic does not tell us what the maximum directional move of the open was; it tells us specifically what the largest directional move for a failed drive off the open is. If you faded every single move off the open you would certainly experience more pain than this, but it also tells us that any move that drives more than a few points is not likely to reverse in the first half hour. This is also valuable information.
We can also ask how quickly the market is likely to reverse, if it is going to reverse at all. We find that that, on average, the first directional drive fails and reverses to the other side of the opening print in 4.3 (3.3) minutes. The median reversal time is 2 (2) minutes, and the ninth decile is 12 (7) minutes. For the purposes of this test, we considered any drive that did not reverse within the first 30 minutes a failure of this tendency, so there certainly were reversals many hours later that we are simply counting as failures to reverse.
We should also probably consider whether there is longer-term significance to this tendency, and one easy way to do this is to simply look at the relationship of the 10:00 print to the opening. We find that a fade trade after a directional drive off the open would be a winner from the opening print (which is pessimistic since if you are fading you should be positioned better than the opening print) 57.9% (63.9%) of the time, with an average win of .75 (1.0), and a median of .95 (1.50). Understand that we are only using this test to illustrate a market tendency. If we were trying to build a rule-based trading system we would need to do a lot more work, as there is certainly a better exit than just getting out at the randomly-chosen 10:00 print.
I know that was a lot to digest, but let’s simplify it: We need to use the S&P futures, rather than the SPY, because there is a little less noise in the futures. About two-thirds of all trading days will have a first one minute bar that is strongly directional, which we define as the opening print being within 1 tick of the high or low of that one minute bar. When we see this condition, we know that the first drive is quite likely to exhaust itself and cross to the other side of the open within the first 10 minutes or so. This is one of many little quirks or tendencies in the market that can be quantified. Taken together, many of these start to help a trader build some intuition and insight into how price action unfolds from a technical perspective.