I spent a couple of years using SmartQuant's QuantDeveloper (now owned by QuantHouse) to evaluate the viability of various technical analysis based trading systems. I had great success with tweaking simulations to make in-band solutions work, but when it came to using the developed scenario for out-of-band data, the attempted solutions became woefully inadequate.
Sunday, August 5. 2007
Trading System Design Thoughts: Price - Volume - Time
In reading various books and blogs, I could see that people, when trading using traditional technical analysis tools, would spend much of their time on the look-out for new stocks with a potential for a large trend, whether the trend be up or down. I asked my self why does one need to jump from stock to stock to find trades? In effect, those traders are looking for directional volatility. As a corrollory, it would appear that they are unable to make money when markets go sideways (ie, don't trend one way or the other).
The people looking for trends will always have market scanners running in order to find the 'hot stock' for the day. Depending upon the sensitiviity of the scanner, much of the trend to be found could have already been run, with very little left to go. One really needs to be in on the ground floor, but those opportunities are few and far between.
When looking through Amazon book lists for traders, all one really sees are books based upon chart analysis, technical indicators, and stock selection. During my initial research into trading, I did in fact obtain a number of those books. But as already mentioned, I became dissillusioned with what they had to say. I couldn't really put my finger on the answer to the question of why. Some good, solid, statistically validated answers became apparent once I obtained "Evidence Based Technical Analysis" by David Aronson. He basically proved what I had finally learned: a lot of published techniques are only so many words on paper. (I think I ranted about this once before, come to think of it).
While looking at equity trading, I also did a bunch of research into options trading. Good options traders know all about volatility, and how to make use of volatility in selecting an appropriate options trading strategy. Because of the wide variety of options strategies, and my inexperience with making money in this realm, I decided to back off of options, and move back to equities.
As a side note, it is interesting to note that the authors (Chacko, Jurek, and Stafford) of a paper entitled "The Price of Immediacy", in a recent issue of Journal of Finance, "show that limit orders are American options", which is a nice segue into equities. (The article number is 4458.pdf).
During the transition back to equities, I came across J. Welles Wilder Jr.'s book called "New Concepts in Technical Trading Systems". He appears to be the one who introduced the Average True Range, which is a mechanism for measuring volatility.
With a better understanding of volatility, I set out to use this knowledge in trading equities. I created a stock screener to use end of day data to find equities with good daily volatility. From an absolute volatility perspective, GOOG always landed on the top of the selection list. But one needs to be well financed to trade there as it is currently in the $500 range. ICE turned out to be a good runner up with it being in the $150 range with good daily liquidity.
I havn't assimiliated all it's nuances yet, but Joseph E. Murphy, Jr.'s book "Stock Market Probability" has much to say on statistics and probability as it relates to stock movement. Although it covers mostly long term trading, it may be useful for intraday movements.
Content of "Bollinger on Bollinger Bands" by John Bollinger assisted much in terms of understanding and measuring volatility.
In relation to Bollinger Bands, I developed a peak detection tool to determine how often an equity changes trend direction in any given day. The relationship is that peaks will typcially relate to Bollinger Band edges, and point out new edges, so to speak. Since the peak detection tool provides peak determination in a real-time delayed fashion (yes, I know I could explain that better, but it sounded more interesting that way), it can't be used directly as a trading tool, but it does yield some interesting statistics in terms of average peak-to-trough runs and their average duration. On a volatile equity, one gets lots of peaks, some bigger than others. I've found that I should be able to focus on one or two stocks regularily, and begin to learn it's idiocyncracies, and as a result trade it profitably, even though it may, from a daily chart reader's perspective, be going sideways. It may be trading side ways over a period of days, but it will have lots of intra-day ups and downs.
This, in effect, is what Market Makers do: act as sources of liquidity to traders. They play the market on both sides simultaneously. They enter the market at the beginning of the day directionless, and attempt to end the day directionaless, that is either with no portfolio, or with a portfolio with evenly matched short and longs. In Option Maker's parlance, this is called ending the day with a zero delta.
You'd think that a book by the title of "The Market Maker's Edge", which in this case is written by Josh Lukeman, provide some details about market making and how to trade in that manner. Instead, the book has a decidedly technical analysis bent, with not enough on the higher frequecy perspective on trading. "The Nasdaq Trader's Toolkit" by M. Rogan LaBier does a much better job of introducing one to Level II data, and what is happening on the markets. But the book dates itself through screen shots using fractions rather than the current decimalized system.
As a book not necessarily devoted to Level II analysis, I did find "Mastering the Trade" by John F. Carter to be extremely helpful in finding out about various market relationships, including what to look for before the market opens. It also suggested ways to make use of the trin and tick indicators while the market is open. The book "The ARMS Index (TRIN)" by Richard W. Arms, Jr. provides much background on how this works, and is a very useful tool for helping in determining short term (intra day) market movement.
So, after having said all that, I've come to realize that 'it' is really all about short term (intra day) market movement. Can one make money from all the gyrations of the market? It comes down to statistics and probability: how often are trades within a range and how often and when do they do a range extension?
It comes down to evaluating price, volume, and time.
In using Interactive Brokers Trader Workstation interface, in particular with the BookTrader interface (otherwise known as the ladder interface), one can see the latest price, bid, and ask. When subscribed to Level II, the content of the Limit Order Book is also available. By clicking on the bid or ask column at a price level, one can quickly place Limit Order bids and asks in order to bracket price movement. As price moves, the Profit/Loss of the cumulative position is updated in another column. In addition, a tick histogram is available for determining popular price levels. I find the book trader easier to work with rather than the traditional side by side bid/ask Limit Order book.
About the time I found out how that works, and how effective it is for active trading, I came across a few threads in Elite Trader which discussed this as a 'Non Linear Trading' method. One contributor explained how he used two accounts to work both sides (the buy side and the sell side) of the market at once.
Since IB isn't/wasn't all that much into customer service or special requests, I scouted elsewhere for a broker who would be willing to set something like this up. Genesis Trading turned out to be easy to work with in this regard. They were able to set me up with two trading accounts that draw off the same fund account. The only drawback with them is that they are mostly equities, they don't do the miniDow (YM), which I've been paying attention to in one fashion or another recently.
As Genesis doesn't seem to offer the equivalent of IB's BookTrader for monitoring Price - Volume - Time, I did a quick prototype in SmartQuant's QuantDeveloper. Unfortuneately, Genesis' API is somewhat lean when tied to a .NET framework. Gensis, instead, has a robust C++ framework. And since I found the .NET libraries a bit slow, I'm currently involved in rewriting my prototype in Microsoft VC++ 2007. It 'feels' faster, and 'closer to the metal'. C# is good for building systems quickly, but one loses the feeling of 'getting dirty' when working with it.
During trial runs on the C# version, I found I was getting caught up in following the tick rather than keeping track of the big picture in order to bracket trade ranges and follow range extensions. I found I needed to see the 'forest for the trees'.
My Peak Detection module was supposed to help with that, but not as much as I hoped. I came across a technique known as the Market Profile. The Market Profile breaks a day into 30 minute slices. The trading range in each time slice is marked with a letter of the alphabet and then 'draped' over the predecessor time frames. This allows one to find where most of the market action is occuring. By bracketing the 70% range, it should be possible to pick up a bunch of good trades with relatively little effort.
There are two recent books, both by Dalton/Jones/Dalton. The older one is "Mind over Markets" and should be viewed first, as it introduces the concept. The newer, recently released one is "Markets in Profile", which builds further on the theory. My plan is build and process Market Profiles in real time so as to maintain a 'big picture' view of the trading day.
There are also significant online resources for Market Profile. Much of the initial research was performed by J. Peter Steidlmayer while at the Chicago Board of Trade. The CBOT has a good Market Profile resource area including a free downloadable handbook in the educational resources area. Cisco Futures has a tutorial on Value Based Power Trading, which shares some of the material from the CBOT manual. The tutorial can also be downloaded as a .pdf. They have more links at Value Based Trading Research page.
In one of these references, I came across a remark to the effect that people were having a problem with using the Market Profile for building multiple day strategies. Given that market research indicates that any day is a 50%/50% chance of going up or down, I can see why people would have this problem. I think this is another reason to not try holding multi-day positions. Each day should be treated separately. This becomes readily apparent when doing end of day recaps, and realizing that each day moved due to some different market stimulous.
At the CBOT site, there are two good introductory articles by Jack Broz: Trade by the Book - A Guide to Reading Order Flow and Reading Order Flow. The first uses the Limit Book side by side format, while the second shows the ladder format.
The ladder format is used by many trader applications, Ninja Trader and Button Trader are ones that come to mind immediately. However, by the look of them, they don't appear to handle two simultaneous trading accounts. Hence, my motivation for coming up with my own application.
Which brings me to the present. My trading software is almost tradable, as in I'll be able to place and cancel Bid/Ask limit orders in a ladder format quite soon. There is a bunch of supporting infrastructure to implement, but the hard bit has mostly been accomplished. I hope to provide a screen capture of it in operation soon.
The goal of TradeFrame, the name I've given the software, is to provide good perspectives on price - volume - time. At each price level, accumulated volumes and ticks are presented. It is able to provide limit order book depth. And through auxilliary charts, it will provide market statistics such as tick and trin.
Then, as time goes by, I hope to try adding in semi-automation. The ultimate goal will be to fully automate the process, but can only be done once I've got a good handle on the manual process.