  |
|
Position Sizing One of the biggest problems with many trading systems is that they risk too much of a traders equity on any given trade. Most professionals agree that you should never risk more than 1% to 3% of your equity on any given trade. This also applies to the amount of risk per sector. For example, risking 2% per trade in highly correlated markets like 2yr bonds, 5yr bonds, 10yr bonds and 30yr bonds is essentially risking 8% in the same trade, far in excess of what's prudent. Overtrading this way can produce what appears to be incredible looking results with returns of 200% or 300% 1000% etc. This is usually just a case of using too much leverage and risking too large a percentage of the account per trade (or sector) and or "cherry picking" an optimal starting date (like right before a series of winning trades). When you run a "Worse Case Analysis" at those high risk levels you see that your risk of ruin climbs dangerously high. A series of bad trades or starting on the wrong day could cause you to lose it all (or have an enormous drawdown) . The bottom line is that when you put a trade on, you should know exactly what percentage of your equity you will lose if your wrong. This percentage should be a very small amount of your available trading capital. This also means you need to know your actual risk when you enter the trade (excluding slippage). Some systems like a moving average crossover reversal systems don't know how much risk they are taking. This is because the system does not know how far the market needs to move to actually trigger an exit etc. We think its dangerous to trade this way and don't recommend it. Another big problem is the lack of market normalization (such as single contract based results). For example, we don’t think it makes sense to trade one contract of natural gas with an average daily volatility of around $2,000 for every one Eurodollar contract with an average daily volatility of around $150. Doing this would imply that natural gas is a more important market than the Eurodollar. We don’t think that's a good idea, if the Eurodollar has a trend we want to give it just as much weight as natural gas (or any other market). In the previous example, you could just simply remove the Eurodollar from the equation and get roughly the same performance. In essence, the results have been inadvertently biased (curve fit) to natural gas even if both markets are traded the same way (an average $150 winning trade in the Eurodollar is not going to offset an average $2000 losing trade in natural gas etc.). The reason you trade a basket of commodities is to be diversified, however, if most of your profits and losses come from a few of the markets in the portfolio your not really diversified. The problem is that going forward, you are going to be dependant on those few markets to perform. Far better to know that any market has the potential to perform at an equal level instead of being dependant on specific markets within that portfolio. We think this type of position sizing and money management is the most robust method. It’s likely that most systems ignore position sizing or apply it illogically because most software packages have been designed to work with single contract based testing. In fact, of the countless back testing products available for sale we are only aware of one software package that can properly do position sizing and money management testing. There are many products that claim to do it (software add-ons, excel spreadsheet macros, "reflection", eye candy type products etc.). However, we have found that almost all of these products don’t do position sizing & money management correctly (there are quite a few reasons for this, contact us for details). We use Bob Spears state-of-the-art testing software Mechanica (which currently sells for $25,000 per copy) for most of our positon sizing based research and testing. Other problems include vendors that only report the smaller drawdown numbers like "closed trade" drawdowns or "average annual" drawdowns etc. There are also problems with position sizing concepts such as as "Optimal F" or "Fixed Ratio". In our opinion, both of these are actually just a dangerous form of hindsight biased curve fitting. Along these lines is another common fallacy that says you should find your "best" single contract based system FIRST and THEN apply position sizing to it. This is not the correct approach; position sizing can completely change the risk-to-reward profiles of single contract based systems. A system that looked great with a smooth equity curve on a single contract basis can look far less attractive when all markets are equally weighted for robustness (as they should be). For all the reason cited above, we make sure our systems are developed and rigorously tested with what we feel is the correct type of position sizing logic and money management software. We believe this increases the robustness and significance of our testing results. This also helps avoid the inadvertent optimizing that can occur with other types of position sizing / money management based testing software.
Feel free to email or contact us
with any questions or comments on this subject. dhoffman@traderstech.net |
||||||||||
|
September and October 2003
articles
|
To see a video
demonstration of us using position sizing in Mechanica®
|
 |
|
 |