Building a Winning $TICK System Part 1

Download Word Document, PDF, and Performance Report

Part 1 Why $TICK Can Work
$TICK is the aggregate of stocks ticking up and ticking down in a realtime “snapshot” of all NYSE stocks. If 500 more stocks last ticked up than down on the NYSE, $TICK will be 500. The $TICK and $TIKQ (for NASDAQ) indicators are among the only pieces of short term technical data about entire exchanges of stocks publicly available on realtime data streams. While most system developers are concentrating on price information, here is a piece of short-term, aggregated information about thousands of stocks that move major indices.

While “fundamental” refers to the underlying condition (usually valuation) of a company or market and “technical” refers to price transformations, $TICK is neither. It is a mechanical structure of the exchange, describing a unique directional “state” of a huge group of securities. If this could be done on the company level, imagine a constant “voting machine” that computes internal corporate sentiment by the actions of its employees. Nearly impossible, right? But it certainly would be related in some way to the performance of the company. In the analogy, the exchange is the corporation and $TICK provides a “voting machine.”

The reason $TICK can work better than traditional indicators is that most indicators are price transformations. Price transformations try to model patterns and in turn statistical edges directly from price changes. This usually does not work exceedingly well without curve fitting, and certainly requires rich diversification among price-based strategies to maintain low volatility characteristics. As a side note, the one major price based strategy that seems to be exploited successfully is the fact that the distribution of price changes have “fat tails.” In other words, price changes that would be unlikely based on a “normal” set of random price changes are (relatively) likely in the stock, commodity, and currency markets. Trend following systems exploit this piece of fundamental information by getting aggressive when these out-of-line movements occur. Think about that for a second. Trend following does not work because prices contain all there is to know about the market or because price leads news. It works because the systems are exploiting a fundamental piece of information about the distribution of price changes!

The key to getting a giant edge with technical analysis is to “zoom in” on the market’s environment and influences whenever possible. Contrary to popular belief, fundamentals are valuable to technical analysis because they can reveal clues that price alone can’t. The best traders in the world agree that technicals should be used in tandem with fundamentals. I personally have always thought fundamentals can be incredibly technical, too!

Many traders are duped into taking “sides” of the fundamentalists or technicians. This campy attitude often influences a trader’s rejection of an entire input type. For example, some technicians claim that everything is priced into the markets and therefore price is the only information necessary to gain an edge. If only it were that easy! As mentioned earlier, price alone is the least likely input to yield statistical edge because price patterns can only influence price to a certain degree. There are basically two ways to use price alone to gain a statistical edge. One is to exploit the “fat tails” characteristic of financial markets, capturing those few times when big moves occur. A sub-problem of this strategy is to maximize those gains with position sizing, and maximum diversification is necessary. The second way is to use ten sor hundreds of individual adaptive systems that are constantly optimizing and learning which price transformations are currently showing the strongest edges. The first method requires ample diversification, and the second method requires complex programming and computing power. These are major tradeoffs for only using price data!

The point is by remaining open to both technical and fundamental information such as $TICK substructure, it’s easier (but not easy) to gain an edge with classical research methods and fewer resources.

 

Part 2 Norman Packard Interview Excerpt
I’m not the only one who thinks price data can only go so far, especially for the undercapitalized. Norman Packard was member of a team of physicists that brought chaos theory to science. He and his colleagues searched for needles in haystacks, finding order in the random “systems” of nature such as the structure and evolution of snowflakes. Packard and Doyne Farmer left academic jobs in the early 1990s to apply their theories to the world of finance. They were not altogether surprised when chaos theory failed to work on financial price data, but they quickly adapted…

The following is a 1995-1996 in an excerpt of an interview by Joe Kolman available at http://www.derivativesstrategy.com/magazine/archive/1995-1996/1295qa.asp.

Kolman: What are the main limitations of such directional betting?

Packard: Our approach basically uses elements from both of these areas, technical and fundamental. But we're sig- nificantly different from the traditional technical and fundamental trading paradigms. First of all, the technical trader usually looks at just the price stream of the instrument that he's trading. He tries to extract patterns from the price stream, which takes the form of charts. There are various ways of analyzing charts and making diagrams derived from the charts, and extrapolations from the diagrams about where the price is going to go.

Kolman: Candlesticks or-

Packard:-or support levels or flying wedges or Fibonacci sequences or Elliott waves. All of these things are basically tools that allow a technical trader to try to extract some kind of structure from the price stream to tell where the price is going to go in the future. Now we actually look at the price stream ourselves. We're not above looking at just about all of the technical variables. We call them technical transformations. And we find that some of them do have predictive value and some of them don't. But very few have enough predictive value on their own. But when combined with many, many others, we can sometimes build a model.

But in addition to these technical transformations, we generally have other kinds of information that we use as inputs for a model that would come under the heading of fundamental information.

 

Part 3 Exploiting $TICK With TradeStation®
Although $TICK is based on NYSE stocks, the S&P 500 has a significant concentration of stocks from that exchange. Many S&P stocks belong to the NASDAQ, however, so the S&P 500 index’s relevancy to $TICK is not perfect.

Commissions are deliberately not deducted from this test. The first object in testing an “input” is to identify whether an edge exists at all, and that’s what we are doing here. The second is to decide whether the method is tradable or if there is room for improvement, which I will leave to the reader to decide.

From visual observation I always thought abnormally high or low $TICK readings are more likely to precede market moves in the opposite direction. This may be true on a very short time frame and is a topic for exploration, but on a manageable 15-minute time frame $TICK is overwhelmingly indicative of future market movement in the same direction.

There are an infinite number of ways to define what is a high and low $TICK reading, but some are certainly better than others. Ahh, some of the most forgettable moments in your college statistics class have come back to make you cry! Actually, it’s not difficult with a vague understanding of the Z-score. I choose to use the Z-score probability to announce when any indictor reading is out of the ordinary. The Z-score measures how likely a point is in a sample or population. The sample in this case is the number of $TICK bars from the past. There are 26 15-minute periods in one S&P day, so it seems reasonable to look back on 52 15-minute periods, or the past two days, for “normal” $TICK behavior. The “point” to be tested is the High or Low of the current $TICK bar, depending on whether the test is for a long signal or short signal.

First, define variables and inputs. zProbLength is the number of bars to look back in defining what is “normal” behavior. zProbabilityHigh/Low store the current bar’s Z-score probabilities.

Code:

Inputs: zProbLength(52);
Vars: zProbabilityHigh(0), zProbabilityLow(0);

Now generate signals. Here, two Z-score probabilities are calculated. One is the probability that the High of the current $TICK bar occurred. When this probability is low, buy. The second is the probability that the Low of the current $TICK bar occurred. When this probability is low, sell short. A high result from the function call means the data point was very high. A low result indicates a very low reading. Note this is unlike the traditional Z-score probability which ranges from -0.50 to +0.50.

Code:

zProbabilityHigh = ZProb(High of Data2, zProbLength);
zProbabilityLow = ZProb(Low of Data2, zProbLength);

If the High of this $TICK bar is unusually high, buy the next bar on open.

Code:

If zProbabilityHigh < 0.02 then
     Buy at next bar on open;

If the Low of this $TICK bar is unusually low, sell the next bar on open.

Code:

If zProbabilityLow > 0.98 then
     Sell Short at next bar on open; 

That’s it! The strategy is in the market all the time – invariably long or short based on the most recent extreme $TICK reading that opposed the position before it.

Check the .MHT file for performance information on the E-mini since 2/12/2001 (since intraday $TICK data has been available) and including $12.50 commissions. This strategy is wide open for improvements!



Presented as research only. The past does not equal the future!

 

Adaline,

How are u calculating the Z score probability?

 

AMEN

 

trend456, built-in TradeStation function

 

My feeling is that the question whether the TICK leads the eminis or the eminis lead the TICK is nontrivial and depends maily on exactly how fast the respective exchanges are, how good the connections of various participants are to the exchanges, and generally on how "they do business" at the exchanges. Globex has not changed much, but it has been going through slower times and faster times in the last couple of years. NYSE is a whole 'nother story. It seems that trading culture, and especially electronic quote distribution, at the NYSE is re-invented every half year or so. Just ask those who scalp NYSE stocks how things were two years ago, 18 months ago, 12 months ago, 6 months ago, and how they are now.

Now if quotes at the NYSE in just one specific market condition are suddenly updated a couple of seconds faster, the TICK might seem to "gain on" the eminis in these situations. Or if your quote provider changes something in their configuration, you might suddenly get globex or TICK quotes a few seconds faster (or slower).

I remember a time about a year ago when all you had to do to earn a living trading NYSE stocks was to wait for the eminis to jump a point or two, get a TMBR auto-execution off of a slow NYSE quote and get out a minute later. Of course, it wasn't long before TMBR adjusted its algorithm...

 

Lobster,

I see what you're saying, but would that not be most appropriate in a 1-3 minute time frame involving $TICK? I gather you're talking about very short term reactions.

The one I describe is finding trends lasting hours to days in the same direction... move the signal back or forward a few 15-minute bars and the performance does not change much.

Something related though... do this test on $INX (S&P500 Index) and the performance is significantly better. But you can't trade an index value of course. The contracts themselves are more efficient than the index feeds. Someone out there has surely proven this.

 

I was thinking about that, as you are of course correct in your assumption that I was talking about very short term phenomena. You probably have a good point, but we should be careful about exactly how many points per trade a system like yours would average. Let us say, for example, that you get an average (including losing trades) of 3 ES points per trade. This might be all but annihilated in a few seconds, since the ES often jump a few points in the same minute that the TICK reaches an extreme value. Therefore, you might do the exact same trades, only 10 seconds later than you "should" and thereby turn this day- to swingtrading system from a winner into a loser. This is obviously wild speculation and might be totally untrue. But the thing to keep in mind is that NYSE quotes are sometimes several minutes delayed, especially in situations where the TICK is extremely high or low, while the ES often have a very large range in those very minutes.

 

We're talking about slippage now, yes? You're absolutely right, but note the signal is at the high or low of the 15-minute ($TICK) bar, but the entry does not happen until the close of that bar. Invariably the signal point is going to be before -- up to 15 minute before -- the entry/reverse. Whether this helps or hurts performance is actually still unknown, but logically a late entry would hurt performance, so that is inclusive in the backtest.

One telling method is to have the backtest fill at the low of the bar for shorts and high of the bar for longs. I'll try that and post it soon!