Quant investing: making momentum tolerable

For today’ s post and the next few I’ll be going back to my favorite topic, quant investing. In this post I want to explore pure momentum quant portfolios and in particular ways to make pure momentum investing tolerable and implementable to more investors.

Note: for a refresher on momentum and its power (arguably the most powerful factor in investing) see this great paper from AQR. 

You may have noticed that none of the quant portfolios that I have presented on the blog are pure momentum strategies. Only two strategies, trending value and microcap trending value, use momentum to picks stocks at all. In combination with value. The rest of the quant strategies are pure value strategies. Why? Well, I pick my strategies on risk-adjusted returns, Sharpe and Sortino ratios. No pure momentum strategies have made the list of top strategies based on risk adjusted returns in O’Shaughnessy’s What Works On Wall Street (table 28.3 page 625 in the 4th edition for your reference). Pure momentum hasn’t made the list of top strategies because it suffers from periodic momentum crashes which leads to large drawdowns. So, why even consider pure momentum? Most importantly for me is because it tends to be uncorrelated with pure value and can add value as a quant portfolio diversifier. Behaviorally, it can also help to stay in the game when value investing is out of favor as can happen for long periods of time. Do the late 1990’s ring any bells? Finally, there are ways to mitigate some of the largest swings in pure momentum portfolios.

First, lets construct a simple pure momentum portfolio and look at the portfolio statistics. As usual, I’m using Portfolio123 for my portfolio constructions and backtests (these are more complicated portfolio simulations and not just simple stock screens). Let’s take the SP500 index, sort the stocks by 52 week momentum (1 yr return), invest in the top 25 stocks equally weighted, and re-balance the portfolio every month. Here are the results from the beginning of 1999 through the close on Nov 11, 2016.


Can you see the issues with pure momentum now? We get significant market beating returns (8.75% CAGR vs about 5% for the overall index) but with very large drawdowns which leads to low risk-adjusted returns (Sharpe of only 0.42). Most investors would not stick with such a strategy. Also, notice the high turnover of this strategy. Almost the entire portfolio turns over every month making it harder and more costly to implement. Now that we have a baseline let’s look at making some improvements.

Let’s try and mitigate the large drawdowns first. We know stocks perform the worst and have the largest drawdowns during economic recessions. And we’ve already looked at a way to using economic indicators to improve risk adjusted returns. So, lets use the SPY-UI combo indicator and apply it to the pure momentum portfolio above. This indicator is effective in quant portfolios. Below are the results.


That sure helped. Returns increased significantly, from 8.75% to 12.5% annually, and drawdowns dropped from -63% to much more tolerable -35%. Accordingly, Sharpe ratio also increases any almost 50% to 0.63. Also, notice the lower correlation to the index. That’s pretty darn good and it gives you an implementable and tolerable pure momentum strategy. Turnover is still quite high though. There is one more set of tweaks we can make to improve things a bit more. We’re going to limit the universe to Market Leading stocks. This is another O’Shaughnessy concept which I first discussed here. Market leaders are

…defined as non utility stocks in the large stock universe (market cap > average) with shares outstanding greater than average, cash flow greater than average, sales greater than 1.5 times the average.

I think of this as a kind of momentum as well. Large, successful, liquid stocks. To reduce turnover we can also hold on to winners and not sell them until momentum starts to flag. And to make sure it doesn’t flag too much we’ll check once a week instead of every month. Finally lets pull out the complete bag of tweaks, we’ll switch to 6 month momentum since it tends to perform slightly better than 1 year momentum, we’ll concentrate the portfolio even more and limit it to the top 10 holdings, and give the re-balance tolerance wide berth at 30%. Results for this modified pure momentum portfolio are below.


Not too shabby. Higher returns, equivalent drawdowns, much lower turnover, and a higher sharpe ratio. Also, much lower correlation to the index. I would only use 10 holdings if this strategy were part of a larger quant portfolio. As a stand-alone strategy I would probably stick with 25 stocks.

There you go. While pure momentum strategies can be quite effective they can be gut wrenching at the same time. Fortunately, with some tweaks investors can make these strategies tolerable and worthy of inclusion in a portfolio. In a future post we’ll look at what I think is the biggest value in adding pure momentum to your quant arsenal, it’s diversification benefits, especially with respect to value strategies.


Full Disclaimer - Nothing on this site should ever be considered advice, research or the invitation to buy or sell securities. These are my personal opinions only.

15 thoughts on “Quant investing: making momentum tolerable

  1. Great post! Very interesting indeed.

    Could you elaborate a little more on the rules? As I understand the strategy outlined in your posts, it has the following rules:
    * Select stock universe of “market leading stocks”
    * Buy the 10 stocks that has the highest 6 month return (momentum)
    * Use the SPY-UI-indicator to enter and exit the market (i.e. exit when both UI and SPY says sell and buy when either indicator says buy)
    * Sell individual stock when “momentum starts to flag” (this I don’t understand – do you mean MA200?)
    * Give the re-balance tolerance wide berth at 30% (this I don’t understand either)

    1. I’m also confused with the checking period – is it every week for SPY, UI and all indicators or simply when “holding on to winners?”.

      1. Update frequency, in this case weekly, is how often ALL the portfolio buy and sell rules are checked.


    2. Sure. You got the rules mostly right. For flagging momentum I use a Rank < 80 (ranking of stock by 6 month return) in the sell rules. In the position sizing portion of the sim you assign an idea weight to each holding and also a tolerance band, "Max Weight Deviation", it's called in P123. I have that set to 30%. Paul

  2. Hi Paul,

    What level of Portfolio123 do you require for a backtest like this?

    Thanks for your posts!


    1. Designer level to run portfolio sims back to 1999. With the Screener level you can run stock screens back to 1999 but portfolio simulations only back to Dec 2008.


  3. The only way for me to make momentum tolerable is to limit drawdowns … I would choke on a 35.38% drawdown. The majority of self directed investors must start with the rate of return and consider drawdowns as an afterthought. The only way I can remain in the market and sleep well at night is to limit my drawdowns to my comfort level which is in the 10%-15%+- range. This applies whether I am investing in hedge funds, managed portfolios, or non-discretionary trading systems such as TAA.

    The first metric I look for in considering an investment is the maximum daily drawdown. If I can’t find that, I minimally calculate the maximum monthly drawdown. It is only when I am comfortable with the drawdown that I consider the rate of return. The challenge comes when one realizes that the Financial Services Industry as a whole does a very poor job in controlling (and reporting) drawdowns. I believe that the SEC should mandate reporting of maximum drawdown history for all Bear Markets and Corrections since fund inception, but that’s isn’t going to happen.

    I’ve been a believer in Tactical Asset Allocation (versus Strategic Asset Allocation) for years. I’ve read widely and deeply however the size of drawdowns in systems published published by TAA gurus far exceed my comfort level. That of course is not unexpected because high value work is generally held close to the vest.

    Adam Butler has published some fine theoretical work at gestaltu.com which opened my eyes to controlling drawdowns by controlling volatility. While Adam’s best work is also proprietary; his work kick started my own volatility research and model building. I now routinely toss in the bit bucket, any TAA system with a maximum drawdown exceeding 10% during the last Bear Market.

    1. Earl, thanks for your comment. I couldn’t agree with you more on momentum and drawdowns. And value is even worse. Of curse, that’s one of the reasons they continue to outperform. I’m a big fan of TAA strategies and have been invested in them quite heavily for a while as you can read about on the blog. Also, a fan of Butler’s work and have written about it as well. Also, agree that drawdowns are a key metric, especially for investors in the withdrawal phase of their lives (like me).

      Having said that it is not valid to compare one equity quant strategy to a portfolio approach like TAA. Quant strategies need to be integrated into an overall portfolio. (Also, the drawdowns in these sims are overstated because I couldn’t integrate the bond component when the system is out of the market.) When combined in a portfolio, the diversification offsets some of the larger drawdowns of the individual strategies. For example, drawdown can be limited to about 20% (I have a post coming on this). Then just like traditional portfolio building the overall portfolio risk/return profile can be tweaked by the addition of bonds. Obviously, this is way more complicated than canned TAA strategies, so why do it?

      For me, three reasons, one – higher risk adjusted returns, i.e. for the same level of portfolio drawdown I can get higher returns. Second, better portfolio diversification which leads to better long term risk adjusted returns. In bull markets, like since 2009, quant based portfolios will do much better than TAA strategies. Thirdly, I think this approach is more durable and lasting that many of the TAA approaches.


  4. The quants you’ve followed have been rebalanced annually with monthly reviews for acquisitions. etc. Now we are looking at a lot more management and apparent effort. Am I viewing this right?

    Also in your post of Nov 22, 2014, which I think of as your Diworsification post, the money pic looked like the best quant size was 10-15 stocks. Yet, you’ve continued to use 25 stocks in each of the quants followed. I’m missing something here. Please help me understand.

    1. Hey Don, great questions.

      Yes, you are viewing this right. With traditional momentum, like the first example, annual rebalancing doesn’t work. Performance is awful. So, another downside of pure momentum is the more frequent re-balancing. But, as you can mitigate this somewhat by implementing sell rules which only sell when certain things happen – in the latter examples I showed I uses a momentum ranking below 80 as a sell rule. This dramatically drops portfolio turnover. Yes, it requires weekly ‘checking’. But with modern tools like P123 it can be all automated. I get an email Sunday night with all the buy/sells I need to make for my quant portfolios. I don’ have to do anything else except make the trades on Monday. I usually enter orders Sunday night to execute at the close of Monday’s session.

      As for the final question, I use 25 stocks because that is the O’Shaughnessy standard. The minimum portfolio he uses for all his examples. It better allows me to compare performance to the past. I think that is the best choice if implementing only one or two quant portfolios or if someone is just getting started. If running multiple quant portfolios, then 25 stocks is too many. I run 10-15 stock quant portfolios depending on the strategy. Currently I run 5 quant portfolios with a total of about 55 stocks. Much more than that would be unmanageable for me.


  5. Great post. One question, does Portfolio123 deal with survivorship bias? As in, say we are in 2001 in the backtest, will the momentum strategy select from the universe of stocks in the SP500 *at that date*? Or does it introduce bias by only selecting from *today’s* SP500, but the value they were in 2001?

    1. Hey Ryan, this is a key issue for any data set and must be dealt with. P123 uses what many consider to be the best data set in the business, the Compustat point in time data base which handles these issues and more. It was one of the main factors in me choosing P123 as a tool. You can read all about the data sets used P123 and why here.


  6. My sense of momentum portfolios is that the “alpha” is in part a function of shifting beta exposures. Specifically, during extended bull markets the best performing stocks have high beta and thus the momentum portfolio gradually becomes a high beta portfolio. When the top passes, the stocks get slaughtered but then the portfolio starts readjusting into low beta stocks and then the portfolio starts to do better during a bear market. But then the bottom passes and the momentum portfolio underperforms again because it has too many low beta stocks during the new bull market.

    Nick de Peyster

Comments are closed.