Sunday, September 22, 2013

Algorithmic Trading System Update 9/21/2013

It's been about 3 to 4 or so months since I've been running my system with real money. It's incredible to see it in action. It's one thing to get a summary of backtests showing variance and profit profiles, but it's a whole other experience to watch the slowly dragged out numbers in live trading. So far, the system has been behaving exactly the way that the backtests have predicted in terms of percentage of winners, frequency of trades, and variance of the PnL. The system is as simple as opening the trade when the spread moves beyond a certain threshold and closing the trade when it comes back to a certain threshold. It looks obvious when examining a long span of historical data, but when looking at the intraday movements of each legs of the pair, I see no reason why this should possibly work. I'm happy that it continues to work, but I feel a disconnect with my understanding of the system.

I've seen some positions show a very nice unrealized profit, just to revert back to smaller profits. This means there's some tweaking that I can do to enhance the profit taking process. I can probably make some alterations to my core algorithm, like using exponential moving averages instead of simple moving averages to better accommodate for faster changes in the spread.

A friend told me that I really only need an amount of equity in the account equal to 4 times the maximum drawdown. My latest set of backtests show a 360,000 profit with a 7,500 maximum drawdown over a period of 7 years when using 10,000 sized legs. This means that I can potentially be making much bigger bet sizes, and multiply my final PnL almost linearly (assuming that the position sizes are still not large enough to cause slippage). I'm pretty happy with my system right now because I can increase the potential PnL without increasing risk. Individually, each pair has a max drawdown between 1500 to 3500. The 40 or so pairs running in a single portfolio however, has a combined max drawdown of only about 7500. I'm realizing that this is due to the system being market neutral and highly diversified at the same time. On average, there about 14 pairs with opened positions, meaning I have a portfolio of 28 legs. It is uncorrelated with the market since half of the positions are shorts.

I've basically been taking a small break, but I will have to start looking for more strategies soon. I'm gradually increasing the sizes of my new legs to prevent a sudden stagger in the weights of the positions in the portfolio. I'm having more confidence each day that my system runs properly and I hopefully can make this my sole source of income in the future.

No comments:

Post a Comment