Risk Control

Position Sizing Filters

These days, almost all trading systems have entry filters that hope to reduce sour trades and whipsaws. They enhance performance. Most entry filters for TF systems these days require that the signal be in line with the long term trend.

Today, i tried searching up on the subject of position sizing filters but found nothing. Not only do i think there is a limited research on position sizing, but there is little to no research on position sizing filters.

Position sizing filters are different from position sizing models. Position sizing models tells you how “much” you should buy or sell (ie number of contracts). These can range from martingale models to anti-martingale models. On the other hand, position sizing filters takes a step further to increase or decrease leverage based on some other external factors. These factors can include volatility, overbought/oversold conditions, or any other criteria in your imagination. Will they improve results? Well, thats where you should do more testing.



Smoothing Equity Curve- Reducing Drawdown

This subject is by far the most important one in fund management. A lot of fund managers and traders alike fixate on trying to develop new ways to produce an equity curve that is as close to a 45 degree line as possible. I was once told by an experienced fund manager that “success in our business=smooth equity curve.” Thats where you will fund hot money following into your management. You can have the highest return, but if that comes with a drawdown thats really high (ie 50-80%), not many investors will invest in you.

Although my ultimate goal is to manage a hedge fund with quantitative strategies that pursues absolute return, i want to achieve absolute results with minimal drawdown/risk. There are different ways to do that. 

You can continue on to search for better entry techniques…and waste your time there. The other important and useful way is to apply proper money management. 

Money Management

Portfolio Heat: capping the % of open positions so that in case all positions go sour you will only loose x% of your portfolio

Sector Exposure: allocate a maximum % of your capital to each sector limiting highly correlated assets to increase your risk

Position Sizing: how much should you buy/sell whenever you receive a signal from your system

Volatility Control: high volatility may cause wide equity swings; finding a way to limit that will help in terms of lowering portfolio volatility

Portfolio Selection and Diversification: trading as many markets as possible and forming one portfolio where assets are least correlated will increase the odds of catching the trends and making up for the losses

The above are a few pointers i have come across that have shown value in reducing drawdowns.

Money Management Techniques

There are a lot of ways to implement money management (MM) to reduce risk. Some even believe that finding a system that is able to be right 100% of the time means that they can all together forget the MM.

From my own little research I have come up with the following ways to reduce risk…

1. exit strategies (hard stops, trailing stops, etc)
2. trade filtering (trade in direction of trend)
3. position sizing (fixed fractional, optimal f, kelly criterion, etc)
4. diversifying across markets (modern portfolio theory, or ranking on relative strength)
5. diversifying across systems (uncorrelated systems)

Opinions differ in terms of the importance of the above techniques but they all to a certain degree reduce risk significantly if paired correctly with a system.

It is to my realization lately that I have been focusing on too much on the entry side of the market. My excuse being that I am in search of a trading system that will complement my TF system. I have decided to take a step back and improve my existing TF system. The current TF system is a breakout model paired with a common trend filter. The system is good in that it is not curve fitted, but the MAR ratio (CAGR/MAX DD) currently 1 on the dot, can be improved if I focused more on the risk management side.

I have been fortunate to be able to talk to Mr. Bob Spear regarding trading and achieving success in system development. He has many years of experience in the subject and currently a professional money manager. He has time after time suggested that the position sizing is really what I should be gunning for in terms of improving my systems.

Giving more thoughts on the subject in general, I felt that position sizing alone really seems to be more important than many of the risk control measures numbered above. The reason being that it is a holistic approach to controlling risk system wide with the ultimate aim to bet less on losers and bet more on winners / bet more aggressively on strong equity growth compared to periods of mininal growth / betting no more when strategy exposure exceeds x% of equity…etc. When I mean holistic I mean that it achieves risk control by incorporating what the other numbered techniques (above) also try to do.

Having thought about it for a few weeks, I have found that position sizing is most commonly a function of risk. While most traders tend to implement a common % capital risk for each trade, I went on to ask whether it is possible to make position sizing as a function of more variables, ie equity growth rate.

Idea: bet more aggressively when account equity gets bigger and less when its small.

The idea is not new and I came across it in a archive called “purebytes” ( I strongly suggest traders read stuff on there; its got amazing material). Anyways, the above idea seems troubling as your bet sizing does not adjust downwards on equity drawdowns cutting deeply in to your hard earn profits.

Heres my solution….have an equity growth look back period of let say 6 months. If it is greater than some percentage you increase your bet size by half of the mean average return of the last 12 months. On the other hand if the last 6 month return is negative adjust bet size downwards by the mean average return of the last 12 months. With this your bet-sizing algorithm, you bet more while the equity growth is upwards and less when its downwards. This algo can be an addition to a volatility based sizing or others like % capital; its all depends on your creativity.

This is a new idea i think? Enjoy…