On using statistics to make money in the markets

Using statistics to make money in the markets

It can be as easy or as difficult depending on where you get your information and what approach you take. 

To summarize simply:

1. Take your starting capital, allocate 20% for trading and the rest as a buffer.
2. Research and rank strategies based on three criteria: 
1. Number of repeatable trials
2. Expected profit per unit of risk taken, taking fees into account
3. Standard deviation of expected profit
3. Maximize the number of trials and the ratio of expected profit to standard deviation (usually called the Sharpe ratio)
4. Take the best performing strategy, determine number of trials to execute to get a statistically significant live market result, call the number of trials NT. Go for strategies where you are able to execute the required number of trials in less than 6 months.
5. Execute the strategy in the market NT times. You will come to one of two outcomes.
1. If your profit is near to your expected profit then Congratulations! You are now using statistics to make money in the markets! You can plot your trades against your expected profits to see where you have deviated from your strategy. The difference of the live execution to the theoretical price is called slippage and the value can be 1-5% of your total profit. Now continue running at the same size.
2. If your profit is not close to your expected profit this is what the capital buffer is for. Evaluate the data and underlying assumptions of your strategy.
6.  Rinse, repeat and profit.

Done right, the execution part of the process will feel easy, even boring. If you are not bored, that is a warning signal that your strategy is not predictable or repeatable enough or that you need to reduce your risk allocation.


Key points:

1. The emphasis is on well-defined, highly-repeatable strategies. The rules should be defined such that a computer can perform the task of identifying the trades.
• The repeatability criteria rejects a lot of well-accepted strategies, for example buying the S&P 500 stock index when the PE10 is below the historical average is a strategy that is historically tested but can not be repeated enough to prove its validity due to its long time-frame.
• The strategy must be well-defined, for example ranking the Graham NCAV of the universe of stocks in the Russell 3000, buying the lowest 10 stocks and holding two years, and adding every 6 months would be an example of a well-defined strategy.
• A shorter-term example would be to identify all instances where a stock has gone down 20% in one day, and evaluate the expected return if you bought it and held for one day. This strategy can be checked out using the very helpful EOD screener at paststat.com in the section "Change from Open".
2. Execution - you must execute the strategy according to plan after you have specified the parameters. This should be fairly easy as long as you avoid making personal decisions and stick strictly to the plan.
3. An aside is to not make any trades outside of your defined strategy. It helps if you take a detached but professional approach to the process of trading, rather than as a personal endeavor.

 Have fun, be bored and profit from your knowledge of statistics in the markets.