Clarke Capital Management

* 1993 Futures Magazine "Top Trader"<br> * 2000 Futures Magazine "Top Trader"<br> * 2007 Futures magazine "Top Trader"<br> <br> Personal History: <br> In 1989, I decided that I would investigate whether I could apply my computer software development knowledge and previous trading experience in generating computerized systems with which I could make a living trading futures. The development of my current trading philosophy has derived from the research into methods and strategies as well as my actual trading experiences since that investigation began. <br><br> Prior to futures trading I had a successful career arbitraging equity options. My strategy was to be a disciplined buyer of volatility sensitive options when they were extremely cheap and hedge them with overvalued options or stock. This was a successful strategy providing you had discipline, could stand many losing months in a row while waiting for an overall “pop” or expansion of premium, get very cheap commissions and could be extensively diversified. In the late 1980s several things changed in the nature of the equity options business to make it less attractive, among which were a overall contraction of options volume as well as the contraction of clearing member firms desiring to have individuals such as myself trading firm accounts. These factors as well as the institution of NYSE trading collars, which I felt would reduce the potential for large “volatility-expanding” moves, led me to start developing a new career as a futures trader and eventually a CTA. <br><br> From the beginning of my investigation it became evident that the most direct way to make money and the one most compatible with my strengths, was to be a position trader using computer models to develop the entry and exit points. I purchased the System Writer Plus package from Omega Research, but soon found that this software was inadequate and I would never be able to develop anything worthwhile with it. I decided to develop my own testing platform. The platform was designed to test multiple markets together for extended periods with more sophisticated techniques than could be implemented in a straight forward way with System Writer. My earliest systems although profitable, over-traded the markets and required too much risk to achieve their results. <br><br> Constant re-evaluation of techniques and strategy, as well as the development of the software tools necessary for research and trading characterized the period from 1989 through late 1992. Although research and development have continued at an active pace, much of the focus since late 1992 has been on enhancing the strategy that formed around this time. <br><br> Model Development: <br><br> When developing a model, our philosophy is to test it using a large pool of commodity interests (approximately 105) some containing data that dates back to 1945. In order for the model to be accepted into CCM’s portfolio of available models, it must trade all 105 markets using the same rules and parameters, and the results should indicate excellent performance characteristics for the vast majority (at least 90%) of the markets AND for the group of markets as a whole. <br><br> No model is accepted unless it shows stability of performance during tests involving the shifting of parameters and altering of the rules. Much effort has been expended in developing tools to assist in this effort to assure robustness of the models. I consider the software we have developed in this area to be one of our edges in the markets. <br><br> With regards to out-of-sample testing CCM no longer performs this test. I noticed that if the length of the out-of-sample period was longer that the average length of time that the particular model might be expected to have a mediocre or poor performance, then the effective parameters and rules were very similar to the test sample. <br><br> CCM’s general trading strategy is to use several independent models having a time frame focus varying from intermediate to long-term to very long-term, trading a broadly diversified group of futures. The models, in their aggregate trading, have a complimentary effect with each other by virtue of the fact that they are not all getting into, or out of the same positions at the same time or at the same exit price. <br><br> The models in our various programs (when traded as a group) show significant reductions in the length of drawdown periods as well as a smoothing of the equity curve - as compared to any individual model that we use. <br><br> I decided early on to use a multi-model approach because during testing I discovered that during the ideal "trading" state, many of the models produced infrequent entry signals. Rather than having to take large positions from these infrequent signals, (for the large amounts of funds that I planned to have under control), it seemed more favorable to distribute and diversify the decision making process among the many models. We would rather use this process (multi-model approach) because it allows the "scaling in and out" of positions on a systematic basis which is a considerably more preferable and efficient route (not to mention less risk) as opposed to taking on a full commitment with a single model signal. <br><br> In many cases one model is putting on additional entry signals after the first few entry signals from other models are already solidly profitable. This activity also represents the effect of having "another opinion" on any given market from multiple sources. <br><br> All of the models have various filters which may override entry signals. The most basic of these is a choppiness filter which prevents entry when the markets are too choppy. The second filter (and something I feel in unique to CCM) is what I call the “Fuzzy-Logic Trend Filter”. All of our models benefit from this one. The analysis system which initially developed this filter isolates significant patterns of trend over a broad data range. This analysis system also makes sure that any trend pattern it finds is robust, and the filter must perform in all markets. Although all of our models can stand alone and perform well without this filter, we find their performance greatly enhanced by the use of it. <br><br> The “Fuzzy-Logic Trend Filter” finds not only patterns which are trend oriented, but also contra-trend and no-trend situations. <br><br> Risk Control: <br><br> All models have an initial stop loss consisting of either a volatility based hard stop, a dollar based hard stop or usually both, choosing whichever is closer. Currently our “quicker” models have stop losses around $300-$600. About a third of our models fall into this category. Another third of our models have stop losses in the $500-$1200 range. The final group range from about $700-$1500 and in some isolated cases as high as $2000. All models have some sort of trailing stop exit. They can be based on time, level of profits to protect or a combination of both. Also, where existing positions are concerned, all models have a stop unique to itself, based upon entry in the opposite direction. This is probably the least likely way a model in our strategy would exit. On the shorter-focus models we use trailing stops which give some "wiggle" room as long as the position keeps increasing its open profits. Once this stops, however, tightening of the stops takes place quickly with the rapidity of the tightening being inversely related to the accumulated profits in the position. This is because of my belief that: The more a position has demonstrated its ability to win, the more room it should be given to keep winning. <br><br> The longer term models use a combination of chart based slowly tightening trailing stops and time-oriented trailing stops. Two of the models have special limit exits in addition to all the previously mentioned stops, and these go into effect after extended profits have developed beyond a certain level. <br><br> Concerning risk control, CCM’s internal systems calculate (daily) the risk for all open positions, as well as risk for any new positions which might be put on. This is calculated as: The difference between the closing price and the closest adverse direction stop. Stop signals for all models are always taken. <br><br> Trading Philosophy: <br><br> With regards to entry signals, we generally try to take all signals, especially when they occur over a series of days at ever increasing “favorable direction” prices, i.e. the previous entries are already profitable. An example of a situation where we might omit (and not execute) a particular entry signal perhaps would be: If we get a lot of models issuing signals to enter all at the same time at relatively similar prices for a market (or group of related markets), or if we feel that additional exposure is not warranted because of currently existing exposure in related markets. This is especially important in the European bond markets and the world’s bond markets in general as there is often a high degree of correlation in their moves. Other markets such as currencies can also join in with the bonds or other markets in correlating moves. We don’t have a fixed ratio of risk to equity which triggers the desire to reduce exposure. It is something that is closely monitored however, and evaluated subjectively based on the current situation and exposure. We generally try to keep Margin to Equity in the high 20% to low 40% area although in some of the programs, it has gone briefly as high as 65%.