We made a fundamental error in calculating the returns in our last article, and many thanks to one of our readers who was on his toes and caught it. Some of our calculations were based on dollars per trade and others on dollars per share—clearly a case of pineapples and pomegranates, along with doing the calculations much too late at night, so let us clarify the situation:
1) Rather than using a made up example, we will show everything running through one of our models, using the Russell 2000 ETF, IWM. In each case we are using $10,000 per trade and not compounding the results. All examples are holding trades overnight—that is, they are not day trading examples.
2) We will use $0.01 of slippage on each side and $0.005 in commissions per share. The slippage would be close to zero if we were using limit orders, setting the limit a penny or two above/below the price that triggered the signal, but we want to show the impact of a modest slippage assumption so are including it in the calculations.
3) The trading system we are using will have a sweet spot—they all do—and if we go much above or below that range, results will start to deteriorate even without slippage. I don’t know where that will be with the Russell, so we will discover it together. This was going to be the subject of another article down the road, but let’s take the time to look at it here as long as we are filling out the table.
|Signal Generator Periodicity||Expectancy||After Friction||Impact From Friction||Trades Per Day||Net $$ Per Day||Net $$ Per year||Return On Capital|
4) A few hours later after burning a lot of electrons, we now know the answer—the sweet spot is somewhere south of a daily timeframe and north of 5 minutes. Going below 10 minutes for example was pretty horrible—it wouldn’t take a lot of work to see exactly where things fall off the cliff—ditto for going above hourly–but it really isn’t necessary since the “perfect” timeframe for today will move around a little over time, and we don’t want to curve fit so tightly that there is a cliff just above or just below—that is a pretty sure way to not make much money, or more likely to lose it. In this case, we would want to stay north of 10 minutes since the cliff is near, and we would probably look a little harder between daily and hourly signals to see what we could learn.
5) Take a look at the impact of friction. It is nominal until we get down to the 10 minute level, and then it balloons, so it will also be a major factor in falling off a cliff somewhere below the 10 minute timeframe.
6) Unless you were going to just trade IWM—nothing wrong with that idea by the way—we would go through all the potential candidates and look for patterns in common, such as where do the timeframes group; what are the best parameters for each time frame; can we get all the assets into a single set of metrics, or at least just a handful so we minimize the risk of curve fitting; and so forth.
7) The key message is that we want as little curve fitting as possible even if it means we sub optimize every single asset, if by doing so we can have more confidence that we will achieve similar results in the future on a portfolio basis even though every single asset may vary from the back tests when looked at on a standalone basis.
8) A general comment about trading lower order signals—the lower the stock price, the more shares you have to trade to make a reasonable dollar return, so friction can mount. When we are trading intraday signals, our average stock price is in the neighborhood of $50 per share which minimizes friction. We do trade stocks as low as in the teens, but they have to be very exceptional in our selection process to make the cut—we’ll save that discussion for later.
And next time we really will look at how you can lose all your money even with a trading system that has a high expectancy
Author: Rick Martin
For more information or answers to your questions, email Rick at [email protected]
Hypothetical computer simulated performance results are believed to be accurately presented. However, they are not guaranteed as to accuracy or completeness and are subject to change without any notice. Hypothetical or simulated performance results have certain inherent limitations. Unlike an actual performance record, simulated results do not represent actual trading. Since, also, the trades have not actually been executed; the results may have been under or over compensated for the impact, if any, of certain market factors such as liquidity. Simulated trading programs in general are also subject to the fact that they are designed with the benefit of hindsight. No representation is being made that any portfolio will, or is likely to achieve profits or losses similar to those shown. All investments and trades carry risks.
There are no relevant positions in IWM.