Position Sizing in Momentum Models

In an ideal investment model, the investor should target a roughly equal amount of portfolio influence with each asset they invest in. In other words, a trade in Bitcoin should have no more or less impact on your portfolio than a trade in Treasury bonds. (Recognizing that some trades will always be much more profitable than others, while many will be a loss.)

The foundation for calculating this impact is based on pre-determined stop losses. Before entering a trade, you need to know where you will exit the trade. That works in two parts.

First, you should have an initial price level at which you will exit the trade. In trading, approximately one-half of all trades will go against you. No signal—or combination of signals—is perfect. Having a pre-determined initial exit price is crucial to limiting losses and sizing your initial trade.

Second, you should have a system in place to determine where to exit trades that move in your favour. This isn't about targeting a profit. A good trend can continue much longer and be much more powerful than you could initially predict. You need to take advantage of these long, powerful trends to push your portfolio forward. Rather, your system should be specific in determining when the trend direction is changing and it is time to take your profits on the trade.

I have published several posts on position sizing using various methods for calculating initial stops:

Position Sizing with Average True Range

Position Sizing with Percent Risk

Position Sizing with Breakouts

Position Sizing with LEAPS Options

Position Sizing with One-Signal Models

Charts can be very helpful in visualizing your trades and finding entry points on a moving average model. Whether you use Price > Moving Average or Moving Average Direction or Moving Average Crossover, the idea is similar and the challenges surrounding position sizing are similar.

This is a simple example of entries and exits with Moving Average Direction with a 50-day simple moving average.

Credit: StockCharts.com, TheRichMoose.com

As the chart shows, entries and exits are easy. Position sizing is not so clear. For the October to April trade, the trend was profitable so a large trade would have been to the trader's advantage. However, the May swing trade was a classic whipsaw and would have seen the investor in at $1,305 and out at $1,275. That's a 30 point or 2.3 percent drop.

With one signal models to calculate momentum, entries, and exits, I would suggest that using methods such as Average True Range or recent price lows (such as the 50-day low if using the 50-day SMA) can be highly effective. These methods incorporate recent market conditions, allow for flexibility in the markets, and can backtest very well.

Percent risk is not as effective across a wide range of markets, but can be very effective for an investor who only invests in small cap stocks, tech stocks, IPOs, or cryptocurrencies. These instruments are less predictable, need small position sizes, and can have unreliable, or even non-existent, historical data.

However, in a model where we measure momentum in many different ways and can measure the strength of that momentum, a different option may be available that is still very robust.

Position Sizing with Multi-Signal Models

Where we use multi-signal models, we can stick with moving average signals and find the points where the momentum direction changes based on the signals we use. If we translate chart data and moving average data to an Excel spreadsheet, we can actually plot the price forward to find these points.

Using this time period example (with LBMA data which is slightly different from the futures data shares above) we can create a simple signal.

Source: TheRichMoose.com, Quandl.com

Above we have the comparable data in Excel format with the 50-day simple moving average in red. (A single model system, but it helps illustrate the idea.) We see the buy and sell signals for the entire trade. In this case we know the Buy signal failed between $1,285 and $1,275. Of course we never know future prices when we enter a position; however, we can easily calculate that fail point by "walking the market forward" to find point where momentum fails and the signal would say sell.

Using a momentum model that tracks many signals (mine tracks momentum in dozens of calculations in several different styles) this type of stop-loss system is quite robust. Price movements have less impact on the trading system across the many signals. Also, by scaling into and out of positions we can reduce the impact of entry and exit signals. There is no "all-in" or "all-out" unless we have a major move.

At the point of entering positions and building positions, we need to think about how much of our portfolio capital we are willing to risk. How much of our money is in stocks, options, futures, ETFs, or volatile currency pairs and how much do we have in the safer stuff—such as short-term bonds. Once again this forces us to think about capital efficiency, use of instruments like options, and futures which trade around the clock.

Comments & Questions

All comments are moderated before being posted for public viewing. Please don't send in multiple comments if yours doesn't appear right away. It can take up to 24 hours before comments are posted.

Comments containing links or "trolling" will not be posted. Comments with profane language or those which reveal personal information will be edited by moderator.

4 Replies to “Position Sizing in Momentum Models”

  1. Yanniel says:

    I’ve been waiting for a post like this. Thanks Daren.

    Question: how do you go about “walking the market forward”? Do you somehow interpolate the historic data in order to guess/predict where the market will go as long as the trend continues?

  2. Daren (Editor) says:

    My model is built on Excel, nothing too fancy. I just plug in price points moving forward week by week for several weeks to find the price levels break the upside momentum. No historical parameters, no fancy math extractions for this process. The fancy (well, grade 12 level) math is in the model itself.

  3. Yanniel says:

    I think I get it…you are using a try and error approach to find that breaking point of momentum.

    I have been thinking to use excel functions TREND and FORECAST to walk the market forward. Under the hood Excel uses the method of least squares to find a line that best fits the price points. You can follow that line into the future and determine the breaking points for momentum.

  4. Daren (Editor) says:

    Yes, trial and error. It works fairly well just because I have a limited universe of things I am monitoring and I only buy or sell a few times a month (on average). I couldn’t figure out how to do it with the TREND function in my model, given that there are so many inputs. It probably will be possible in something like Dual Momentum. I will try play with FORECAST to see how that works.

Comments are closed