Momentum Investing: Ride Winners and Cut Losers. Period.

Momentum Investing: Ride Winners and Cut Losers. Period.

July 16, 2014 Research Insights, Key Research, Momentum Investing Research
Print Friendly
(Last Updated On: January 1, 2017)

Two weeks ago, we posted a simulation study on the performance of cheap and expensive stocks based on various valuation metrics. The dart-throwing monkeys simulations gave us a vivid look of how cheap stocks beats expensive stocks regarding compound annual growth rates (CAGR), standard deviation, and maximum drawdown.

Here is the link to the older post: Never buy expensive stock. Period.

We received over 50 emails asking that we do the same analysis, but on “momentum.”

You Asked; We Listened…

The most basic momentum strategy buys stocks that have performed well in the past. This strategy is very different from a pure value strategy, which exclusively focuses on buying cheap stocks.

CXOAdvisory, GestaltU, Gary Antonnaci,  and Millennial Invest — as well as others — have discussed different angles on momentum. And of course, there is a slew of academic research on the topic.

How Does Our Simulation Work?

For testing purposes, we create 2 samples. The first sample is from 1927 to 1962 and the second sample is from 1963 to 2013. The samples are selected in a way that we can compare the results of the momentum simulations to the value simulations, which run from 1963 to 2013.

We sort stocks  into deciles based on stock performance over the previous 12-month ranking-period returns (months t-12 through t-2, skipping the first month).

We only focus on US mid/large cap to avoid weird micro/small cap outlier effects.

  • Example: If there are 1000 stocks in the universe, stocks 1-100 go in the first decile (High mom/winners), stocks 901-1000 go in the tenth decile (Low mom/losers), and the stocks in between 101 and 900 go in their respective deciles.

Next, each month we draw a random 30 stock portfolio drawn from either the “winners” decile or the “losers” decile.

  • Example: We draw 30 random stocks each month from the top (winners) and bottom (losers) decile from 1927 to 1962. Again, image we have a monkey throwing 30 darts, every month during the 36 year period, to establish, in each month, a new 30 stock portfolio. Once our monkey has thrown his 30 darts in each month, we will then have 432 separate monthly portfolios (12 months * 36 years) and will have made 12,960 (30 stocks * 432 months) individual stock picks. This represents one simulation.

We conduct 1000 simulations for the top (winners) decile and 1000 simulations for the lowest (losers) decile as described above.

We then calculate the performance statistics for each simulated strategy over the designated time period (e.g. 1927 to 1962).

Each simulated strategy represents the returns a high-mom-investing monkey (past-winner buyer) or low-mom-investing monkey (past-losers buyer) would achieve over the sample time period analyzed. We calculate compound annual growth rates (CAGR), standard deviation, and maximum drawdown.

We compile the results in the charts/tables below.

What Do the Returns to Winners and Losers Look Like?

First, let’s look at the distribution of CAGRs. The high mom portfolios generate much higher returns than the low mom, almost 20% higher on average. Clearly, buying winner stocks has generated strong performance and buying losers has been a sucker’s bet.

Momentum investing research_1
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Historically, high mom beats low mom on a CAGR basis–no doubt.How about the Risks?

Let’s look at standard deviations of the portfolios from our dart-throwing monkeys. First, you’ll notice that standard deviations are tightly bound, even across 1000 simulations.

No matter how you cut it, holding baskets of high mom stocks means less volatility. Low momentum stocks exhibit incredible volatility.

Momentum Investing_Ride Winners and Cut Losers
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Momentum strategies are known to crash and burn. This is best captured via drawdowns. The evidence below suggests that high mom stocks protect the downside better than low mom stocks, but let’s be honest–investing in momentum is a wild ride!

Momentum Investing_Ride Winners and Cut Losers_3
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

What are the Results over a Different Time Period?

From the 1927 to 1962 (36 years) period we get a clear picture that past winners keep winning and past losers keep losing. We conduct the same simulation analysis from 1963 to 2013 (50 years).

The results are broadly consistent, but volatility and drawdowns have closed in on one another.

Similar CAGR distributions: High mom stocks are the obvious winners.

2014-07-14 12_08_28-Microsoft Excel (Product Activation Failed) - MOM_Sim_v01
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

High mom still has relatively lower Standard deviation, but volatility has converged.

2014-07-14 12_15_27-Microsoft Excel (Product Activation Failed) - MOM_Sim_v01
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

High mom stocks still protect the downside better than low mom stocks, but the drawdowns are more similar than in the prior period. There is also a clear message that investing based on momentum doesn’t prevent your stomach from churning–momentum equity investing is RISKY!

2014-07-14 12_18_43-Microsoft Excel (Product Activation Failed) - MOM_Sim_v01
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged, do not reflect management or trading fees, and one cannot invest directly in an index. Additional information regarding the construction of these results is available upon request.

Conclusions

Momentum has historically been a great strategy. Although counter-intuitive to many value investors, buying stocks with rising prices has been a great investment approach–arguably better than value investing.

Moreover, the approach is robust between the 2 samples analyzed. The lesson is clear: Let your winners ride and cut your losers short.

If you’d like to compare the momentum results to the value results, here is a link to the value post: Never buy expensive stock. Period.


Note: This site provides no information on our value investing ETFs or our momentum investing ETFs. Please refer to this site.


Join thousands of other readers and subscribe to our blog.


Please remember that past performance is not an indicator of future results. Please read our full disclaimer. The views and opinions expressed herein are those of the author and do not necessarily reflect the views of Alpha Architect, its affiliates or its employees. This material has been provided to you solely for information and educational purposes and does not constitute an offer or solicitation of an offer or any advice or recommendation to purchase any securities or other financial instruments and may not be construed as such. The factual information set forth herein has been obtained or derived from sources believed by the author and Alpha Architect to be reliable but it is not necessarily all-inclusive and is not guaranteed as to its accuracy and is not to be regarded as a representation or warranty, express or implied, as to the information’s accuracy or completeness, nor should the attached information serve as the basis of any investment decision. No part of this material may be reproduced in any form, or referred to in any other publication, without express written permission from Alpha Architect.


Definitions of common statistics used in our analysis are available here (towards the bottom)




About the Author

Wesley R. Gray, Ph.D.

After serving as a Captain in the United States Marine Corps, Dr. Gray earned a PhD, and worked as a finance professor at Drexel University. Dr. Gray’s interest in bridging the research gap between academia and industry led him to found Alpha Architect, an asset management that delivers affordable active exposures for tax-sensitive investors. Dr. Gray has published four books and a number of academic articles. Wes is a regular contributor to multiple industry outlets, to include the following: Wall Street Journal, Forbes, ETF.com, and the CFA Institute. Dr. Gray earned an MBA and a PhD in finance from the University of Chicago and graduated magna cum laude with a BS from The Wharton School of the University of Pennsylvania.


  • Thanks for posting this simulation. The drawdown for the momentum decile is pretty scary. You may have seen Gary Antonacci’s paper, where he suggests using absolute (trend-following) momentum to reduce the volatility and drawdown of relative (rotation) momentum. Interestingly, drawdown is reduced much more than volatility is, suggesting to me that trend-following reduces tail risk.

    http://www.optimalmomentum.com/RiskPremiaHarvesting.pdf

  • Steven

    Thank you for the interesting study! Could you please explain why you excluded the most recent month when sorting on momentum?

  • Dendrite Research

    Very interesting, well done study. What about the middle deciles? Is there basically a linear relationship between momentum decile and CAGR?
    Can you reduce volatility by doing a paired trade shorting losers and buying winners?

  • Doug

    Very nice work as usual! Do you sleep??

    I bet a 50/50 Value/Momentum portfolio would have some pretty good metrics, as would a 50/50 portfolio with some sort of market momentum filter (i.e. S&P vs. 200 day MA or 4 month abs momentum).

  • Michael Milburn

    re: “(months t-12 through t-2, skipping the first month).” Why is the most recent month’s price change not included when momentum is calculated? It’s not obvious to me why I’d want to exclude the most recent price movement. Thanks,

  • I have an amazing team supporting me in the background. So I get some sleep. Of course, I also have a 2yr old and a 5yr old, so I don’t sleep THAT MUCH….

    Yes, mixing the 2 creates some amazing results, historically. There has never been a backtest I didn’t like, and that is especially true for the one you mentioned

  • Yes, basically a monotonic relationship.

    Not without some serious heartburn. We have a paper that discusses why L/S is a difficult game: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2226689

    If you looked at the L/S as an alternative risk buried in a much broader portfolio it would be an interesting play. As a stand alone strategy it would be very painful at some points in history.

  • This is fairly standard practice. The basic reason is that there is monthly reversion in returns. So what was up last month goes down the next month, on average. If you are trading momentum, you want to cut that effect out of the picture.

  • Definitely. Being long and strong ANY equity strategy is going to involve some face-ripping drawdowns at some point.

    One can definitely toy with timing models and Gary’s is a great example of a simple robust way to potentially deal with the drama of equity investing.

    Wish I could say I found the holy grail of how to make high returns without a lot of risk. One problem: the data suggests there are no robust free lunches. D’oh!

  • Florent

    Thank you for this very interesting study. Could you please explain what is the difference between volatility and standard deviation ? Thanks.

  • Hi Florent,
    In the context above, volatility = standard deviation.

  • Doug, here are some links you may be interested in:

    “The cheapest 10% of stocks by P/E have historically delivered 16.3% per year, and the top momentum stocks have delivered 14.4%, but a combination of the two has yielded 18.5% per year (bottom right corner of the table below). What’s more, the volatility of this combined strategy comes way down: from 24.4% for raw momentum to 18.9% for value + momentum.”
    http://www.millennialinvest.com/blog/2014/5/12/two-ways-to-improve-the-momentum-strategy

    Global Tactical Cross-Asset Allocation: Applying Value and Momentum Across Asset Classes
    http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1079975

    Fact, Fiction and Momentum Investing
    “What if the expected return on momentum were truly zero? Suppose, despite all of the evidence to the contrary and our strong belief it’s positive, momentum had a zero expected return going forward. Would it still be a valuable investment tool? The answer is clearly, though perhaps surprisingly, yes. The reason is because of momentum’s tremendous diversification benefits when combined with value….
    “Even in the extreme case where we assume a zero return for momentum, the optimal portfolio still places a significant positive weight on momentum… Since value is a good strategy and momentum is -0.4 correlated with it, one should expect momentum to lose money based only on that information. Yet, the fact that it does not lose but in this assumed case breaks even makes it a valuable hedge.”
    http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2435323

  • Aaron

    Have you seen Meb Faber’s reproduction of some SocGen research about l/s momentum? The historical problem has been the shorts ripping after a severe market downturn. So for every 10 percent decline in the market you take off 20 percent of the shorts. If the market falls 50 percent, then you wouldn’t have any more shorts. The backtest was very promising and didn’t have the 80+ percent drawdowns like most l/s momentum strats.

  • Jack Vogel, PhD

    We have not seen this, and I could not find the post. Would you mind sending us the link so we can look at it?

  • Aaron
  • Jack Vogel, PhD

    Yeah, that is a neat strategy that Meb outlines. He does some good work!

  • Tony

    Do the results of this study contradict at all with the results of the “Never Buy Expensive Stocks..” study? I would imagine a material portion of the momentum stocks are also growth/expensive stocks.

  • Jack Vogel, PhD

    Tony, that is a great question. A quick study on our end shows that the overlap between the “expensive” stock and “high momentum” stock portfolio was 27.84% from 1963 – 2013. So while there is some overlap, growth stock investing is not the same as momentum investing.

  • Tom Carlson

    Good article. Thanks for sharing. Quick question: did the strategy you modeled here have negative alpha over the last three years based on your data sets?

  • Tom Carlson

    Also, have you done a similar study (as the ones on value and momentum) on low volatility stocks?

  • CAGR 2010-2013
    STD 2010-2013
    MaxDD 2010-2013

    Not much difference. High mom has slightly higher CAGR, but higher vol (note: figure below is mislabeled and should say 2010-2013)

  • figures associated with last comment

  • will add this to our R&D workflow and post.

  • Going forward, how long do you give a stock before you declare it a winner or a loser? do you use the 12 month period?

  • Every month we rerank the entire universe on 2-12 momentum and take another draw of 30 names. In general, stocks in month T-1 that are ranked high/low on 2-12 momentum are somewhat similar to the stocks in T, but not always.

  • Mick

    Hi Wes, long time listener first time caller; thanks for you & your
    teams great work and loved your book – I’m learning lots 🙂 A couple of
    questions:

    1. Is there any way of including trading costs in the simulations?
    2. What effects would trading costs have on performance? (esp re CAGR)

    Re 1927 to 1962: 12,960 individual stock picks (30 stocks *

    432 months) x 2 (as you would have to buy AND sell 30 stocks each
    month) = 25,920 trades over 36 years. At a flat fee of say $5 per
    trade, that’s $3,600 pa and ~$130k over 36 years… depending on the
    size of your portfolio and the flat fee assumption, that may seem a bit
    expensive?

    If it’s not flat fee but commission, would your data
    allow you to assume a commission % per trade and calculate trading costs
    based on monthly value of the portfolio?

    Basically trying to figure out what post-trading pre-tax returns would be…

  • 1. Sure, the simple method is deduct some baseline estimate from the stated returns. For example, 2%/year. The more complicated method is to estimate impact, b/a spread, and commissions, figure out total trades, and identify a more accurate figure.

    2. In these monthly rebalanced simulations, for momentum stocks, you’d probably have 30-40% turnover on average. Let’s say your total trading costs are 10bps, or maybe 3-4bps on the total portfolio value each month. 4bps*12months =48bps. Historically, trading costs could be 10x that (i.e., lot cheaper to trade today than it was back in the day).

    The bottomline is that trading costs will deduct overall returns across all strategies, but these simulations are presented in a “relative” format. Under the assumption that a low-momentum and a high-momentum strategy generate roughly the same turnover and transaction costs, the relative spread will be the same after adjusting for trading costs.

    Going forward, I’d say the all-in trading costs (impact costs, bid/ask spread, commission) on mid/large cap momentum is probably 200bps a year on a portfolio of 500mm. Maybe you are at 100bps on a $1 to $10mm allocation. Who knows.

  • Tony Yin

    great work thanks for sharing.
    I’m still not clear on how winners and losers are declared.
    So after reranking each month, stocks that stays in the group will still be on paper profit/loss and the ones that are dropped will be considered positions closed and then declared winners or losers ?

  • Hi Tony,

    Sorry for confusion. Winner=highest performance over past 12 months (skipping the most recent month) and Loser = lowest performance over past 12 months (skipping the most recent month).

    Each month you sort universe based on “momentum,” which is the past 12 month return (skipping the most recent month), and divvy universe into 2 broad winner and loser buckets. The simulation then takes 30 random from each. Repeat each month over the entire period. Repeat that simulation procedure a 1000 times. Plot the results. You can think of it as a 1000 dart-throwing monkeys that either throw darts at a “winner” dartboard or a “loser” dartboard.

  • Tony Yin

    Dr. Gray, thank you for the prompt response and detailed explanation.
    And to my understanding, the results were limited for “given the best and worst performance after numerous dart throwing, how were the results comparing to each other”? How many stocks were included as highest and lowest performers per 12 months period ? and btw, may I ask what type of method was used to define the momentum stocks being tested here?

    Thank you very much for being so patient about answering questions

    Tony

  • Hi Tony,

    Let’s say you have a 1,000 stocks in the universe. Each month, we will calculate the cumulative 2-12 month return on each of the 1,000 stocks. We then rank them from 1-1,000 and divvy them up into buckets. Let’s say we do decile buckets. That would imply that there are 100 stocks in each “bucket.” Next, we’d have the monkey choose 30 of the 100 stocks, randomly.

    Does that make sense?

  • Tony Yin

    Thanks I understand now.
    I originally thought you pick the 30 out of 100 top decile stocks then hold them for some period and record the results (price difference) over the holding period AFTER picking rather the results are recorded along with picking.

    Tony

  • james

    Do you have the data on a say long 150% high mom and short 50% low mom port or a similar long/short strategy?