Finding and Filtering Trading Ideas
In my last article, I said that essentially every trader can become an algorithmic trader, no math or programming skills needed. If you are already trading profitably as a discretionary trader, the transition is even easier. But what if we haven’t started trading at all, and want to jump right into algorithmic trading -- where should we start?
Both algorithmic and discretionary trading start at the same place: a hunch, an idea. It is all well and good if you already have this hunch. But even if you don’t, there is a wealth of ideas out there waiting for you to cherry-pick. Right here at Benzinga, there are hundreds of experts publishing their ideas. All you need to do is to find one that suits your budget, risk tolerance, and ability. My blog, epchan.blogspot.com, also contains a wealth of ideas. Books that I think are particularly useful in this respect include: Short Term Trading Strategies That Work by Larry Connors and Cesar Alvarez, Option Tradingby Euan Sinclair, and of course my own Quantitative Trading: How to Build Your Own Algorithmic Trading Business.
Often the problem is not that there are not enough ideas or suggestions out there in the public, but that there are too many, and one doesn’t know how to pick the ones that are truly promising. Here is my 5-point guide for choosing a promising strategy for further backtest and improvement:
- Pick only those strategies that you can practically implement within your personal constraints.
If you picked a day-trading strategy that requires constant monitoring throughout the trading day, but you have another full-time job to attend to, this would obviously not be a suitable strategy for you. Or if the strategy requires you to trade a portfolio of 200 stocks to be consistently profitable, and you don’t have the capital to buy them all, this strategy is also not suitable for you.
2. Pick only strategies that are simple and have reasonable economic rationale.
If a strategy has a 20-step setup for a trade, and 10 different adjustable parameters, you shouldn’t bother with it. It can probably fit to any historical data to generate amazing backtest profits, but will have no predictive power whatsoever. Technically, this is called “data-snooping bias”. Also, if a strategy has a clear economic rationale, such as the January effect1, then we can be more confident that its past success is not just a fluke.
3. Pick strategies that have high Sharpe ratio.
Sharpe ratio is a measure of risk-adjusted returns: mathematically it is the average return divided by the standard deviation of returns. Sharpe ratio measures how consistently profitable a strategy is. Maybe we have a strategy that returned 50% in the last year, but if that return came from just one lucky trade that made 50% in one day while making zero dollars the rest of the year, it will have low Sharpe ratio, and we shouldn’t expect the strategy to be so lucky in the future.
4. Did the authors of the strategy include transaction costs in their backtests?
Including transaction costs in a backtest is particularly important when the strategy trades intraday and holds positions for mere hours or shorter. A careful estimate of trading costs, which may include commissions, slippage, market impact, or opportunity costs are crucial to determine if a strategy is really profitable in the backtest. Don’t bother with strategies that look great because the authors did not include some of these costs.
5. Does the strategy appear to be losing steam in recent years?
Many strategies look great in the distant past, but because of their increasing popularity among traders, or because of market structure changes, they are quite unprofitable nowadays. Famous examples include pair trading of stocks, or index arbitrage. The former became much less profitable after the decimalization of stock prices which led to decreasing bid-ask spreads, and the latter can now be executed profitably only by high frequency traders. If a strategy looks like it has decreasing returns in recent years, chances are it will be even less profitable once you start to trade it.
Even if a proposed strategy passes all these tests, you should still take it with a grain of salt, for it is far easier to generate paper profits than real profits, due to myriad subtleties in the transition from backtest to live trading. They should be read just as inspirations. In the next article in this series, I will talk about the process of backtesting them carefully for validation.
Ernest Chan is a hedge fund manager and the author of "Quantitative Trading: How to Build Your Own Algorithmic Trading Business" (Wiley, 2009). Find out more about him at www.epchan.com.
The following article is from one of our external contributors. It does not represent the opinion of Benzinga and has not been edited.