Skip to content

Moderators of Daily Follow Through MR: Volatility (Part 1 of 3)

August 24, 2009

Note: Detailed backtest reports for the Donchian Channel series can be found on the website built by Corey Rittenhouse: Check it out!

Today we will take a look at one of the different factors that affect the performance of a standard daily follow-through mean-reversion strategy. As a refresher, the strategy assumes that you buy after down days, and go short after up days. Since early 2000, mean-reversion following up or down days has been the dominant paradigm in financial markets–especially for the broad market indexes. This is also a very good proxy for how well other short-term MR indicators such as RSI2 or DV2 will perform under similar circumstances.

Historical volatility will be addressed in this post, and tommorow we will take a look at implied volatility as well. Finally, in subsequent posts, i will combine these factors to have a better understanding of their impact on classic daily follow-through MR. Broadly, this series of posts will also cover the impact of volume as well as the impact of trend strength or R-squared. I will combine all of these factors to create a matrix that helps better predict when to bet on mean-reversion and when to bet on continuation from today’s closing direction in the market. This will help disentangle the sources of MR returns, and also broaden our understanding of how today’s market is functioning.

The first test I ran was on the impact of historical volatility on the daily follow-through MR.  I used 3000 bars of data on the SPY (S&P500) as a test sample ending last week. I classified high volatility as being in the 75th percentile, and low volatility as being in the 25th percentile, both with a lookback of 1-year. Average volatility was anything that fell in between those two levels. For comparison, i used both 10-day and 100-day historical volatility. Here are the results:

  CAGR Sharpe % correct avg daily ret
High 10d Volatility 7.32% 0.45 52.60% 0.126%
High 100d Volatility 5.28% 0.32 51.00% 0.087%
Average Volatility 0.62% 0.03% 50.30% 0.013%
Low 10d Volatility 2.25% 0.29 51.00% 0.036%
Low 100d Volatility 1.47% 0.18 51.20% 0.025%

It is immediately obvious that high volatility is far more favorable for the daily follow-through MR strategy. In contrast, the worst environment for daily follow through is the average volatility environment, followed by environments with low volatility. 10-day volatility was a far more superior moderator of returns than the 100-day lookback.  It pays to wait for short-term volatility to bet on mean-reversion, and at present historical volatility is in the 17th percentile. When volatility is low, the return to following the long term trend (1-year average) is superior to daily follow through MR. This suggests that continuation is more probable than reversion on balance, even though the market is overbought on a variety of metrics.

In the second test, i compared what happened in the contingency when the 10-day volatility was lower or higher in relation to the 100-day volatility over the past year in different volatility regimes. I took the 10-day minus the 100-day and looked at the percentile ranking of the difference over the past year (technically above/below average is actually the median).

  CAGR Sharpe % correct avg daily ret
High Vol + Below Avg 10d/100d





High Vol + Above Avg 10d/100d





Low Vol + Above Avg 10d/100d





Low Vol + Below Avg 10d/100d





It seems apparent that in high volatility environments, follow-through returns are more consistent when the 10-day is lower in relation to the 100-day–ie you want to buy high volatility environments when volatility is slowing down. In contrast, in low volatility environments you want to buy when volatility is speeding up. Logically this makes sense, as it is risky to fade market direction into high and accelerating volatility, and also it is risky to fade extremely quiet markets (low volatility, low 10d/100d vol)  which are prone to major breakouts/breakdowns.

In conclusion, historical volatility and the relationship between short-term and long-term volatility is an important moderator of daily follow through MR returns. A simple contigency matrix can help traders better understand when it is likely to work, and when it is not reliable.

14 Comments leave one →
  1. August 24, 2009 12:24 am

    Hi David,

    I like your work. I have a question about your last sentence. Even though average volatility is the weakest environment for MR, your data seems to show that MR is still a better strategy than follow through in all environments. Why do you say that continuation is more probable than reversion?

  2. david varadi permalink*
    August 24, 2009 12:55 am

    blue, technically you are correct……however i muddled in my knowledge of another moderator in that sentence–that following the direction of the long term trend ie if the market is above the 200 or 252 day average is superior to MR when volatility is low. That will be covered in subsequent posts.


  3. August 24, 2009 10:15 am

    This is a well thought out and important article… I will be linking. Thanks David

  4. david varadi permalink*
    August 24, 2009 11:03 am

    thanks jeff

  5. kostas permalink
    August 24, 2009 11:46 pm

    Hi DV,
    In the last table, the CAGR of High+Above is better than that of High+Below but the Avg Daily Rets are going the opposite way. Why isn’t CAGR that points to the stronger regime but you chose the Avg Daily Ret?

    • david varadi permalink*
      August 24, 2009 11:56 pm

      good question kostas, i chose the avg daily ret because the sample sizes were substantially different. avg daily ret and sharpe ratio are indifferent to sample size provide there are enough observations. that is why i post both numbers. Often, the samples for many tests will have uneven numbers in each contingency.


  6. August 25, 2009 6:58 pm

    Powerful stuff David. Thank you.

    • david varadi permalink*
      August 26, 2009 1:10 pm

      thanks wood

  7. August 26, 2009 3:11 am

    I tried implimenting this, it seems a MR strategy only works in case of developed economies. For India and China, the strategy fell flat! This may be because the markets in these economies are more trending.

    There was a marked improvement in result when you wait for 2 days of +ve returns to go short and vice versa.


  1. Regime Switching System Using Volatility Forecast « Quantum Financier

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: