*01/18/2022 UPDATE: The study MACD Study (Part 2): Period Settings has been deprecated. An overhaul and re-post is pending
I. Introduction & Overview
II. Currencies (FOREX)
III. S&P 500 Futures
IV. Commodities
V. Bonds
VI. Conclusion
I. Introduction & Overview
"I’m convinced that one of the biggest lies of traditional technical analysis is that you can apply any tool to any market or any timeframe with the same results" - Adam H Grimes
The purpose of this third part of the MACD study series is to compare the indicator's performance over various asset classes. So far, all parts of the study series have been performed on equities. The following asset classes and symbols will be tested:
- Currencies (FOREX): EUR/USD Pair
- S&P 500 Futures: E-mini S&P 500 (/ES)
- Commodities: WTI Crude Oil Futures (/CL)
- Bonds: iShares Core U.S. Aggregate Bond ETF (AGG)
The MACD under default settings (EMA, 12, 26, 9) will be tested across all of these asset classes, followed by the same process used to find the most optimal period settings for each as described in part 2 of the study series. Since all asset classes performed stronger using a Simple Moving Average (SMA) in their calculation, their most optimal period settings will use be calculated with an SMA (see MACD Study (Part 1): EMA vs SMA for more details).
Futures and commodities backtests will utilize continuous front-month contract data obtained from Quandl and will begin on 01/01/2003. The currency pairs and bond backtests will begin on 01/01/2005 due to difficulties obtaining data further back. All tests will be conducted on the daily timeframe under both the Any Crossing and Below Zero Line test conditions outlined in part 1 of the study series.
II. Currencies (FOREX)
The MACD under default settings on the EUR/USD currency pair yielded the following results:
The reported means, medians, and lose frequencies suggest that the MACD under default settings does not yield favorable results for currency pairs trading. While attempting to find the most optimal period settings, the top three setting combinations with the highest expected value had the same issue with repeated crossing and sample size inflating as outlined in part 2 of the study series. The fourth highest settings combination (43, 63, 11) yielded a low expected value of 46.90% and the following summary statistics:
It is significant that the win frequency is less than 47%, the sample sizes are low, and the standard deviations are relatively larger than those found in equities. This could suggest that the MACD may not be an appropriate indicator for trading currency pairs. Perhaps this is because, unlike equities, currency pairs are not subject to mean reversion.
III. S&P 500 Futures
The MACD under default settings on the E-mini S&P 500 (/ES) yielded the following results:
While the reported means are positive, the lose frequencies and large standard deviations suggest that the MACD under default settings may not yield favorable results for S&P 500 futures trading. While attempting to find the most optimal period settings, the primary setting combination with the highest expected value had the same issue with repeated crossing and sample size inflating. The second highest setting combination (46, 50, 6) yielded an expected value of 169.99% and the following summary statistics:
The new (46, 50, 6) setting performs much higher than the default settings, yielding results similar to the equities' most optimal (41, 46, 8) setting found in part 2 of the study series. Both of the most optimal S&P futures and equities settings yield almost identical expected values, and similar win frequencies and standard deviations. However, the Below Zero Line results for S&P 500 futures outperform those in equities in every statistic.
IV. Commodities
The MACD under default settings on the WTI Crude Oil Futures (/CL) yielded the following results:
The low to negative reported means, lose frequencies, and massive standard deviations suggest that the MACD under default settings does not yield favorable results for commodities trading. After finding the most optimal period settings, the setting (19, 36, 40) yielded a large expected value of 487.31% and the following summary statistics:
Keep in mind that the max return of 339.51% is a heavy outlier. The massive summary statistics, low sample size, and longer period setting values suggest that it performs best by capitalizing on large reversals over long periods of time. However, because the most optimal setting of (19, 36, 40) is slower to react, it may be beneficial to disregard it as a precise entry trigger and instead use recent crossovers as a tool of confirmation.
Users of the MACD on commodities should do so at their own discretion. Commodities may not mean revert in the same way and for the same reasons that equities do, but rather by the physical (or political) circumstances of their individual markets. Changes in these circumstances drive large, encapsulating moves which the MACD (SMA, 19, 36, 40) appears to capitalize on. Commodity traders would give precedence to the information driving these large price moves before considering any lagging technical indicator. Therefore, commodity traders may not regard the MACD as a valuable indicator.
V. Bonds
The MACD under default settings on the iShares Core U.S. Aggregate Bond ETF (AGG) yielded the following results:
The reported means, medians, and lose frequencies suggest that the MACD under default settings does not yield favorable results for bond trading. While attempting to find the most optimal period settings, the majority of the top combinations with the highest expected value had the same issue with repeated crossing and sample size inflating. This is because the backtest on the AGG failed to yield a significantly high mean. A script was run to remove any combination where the difference between the Long MA and the Short MA was less than 3. The most optimal setting combination (50, 56, 7) yielded a low expected value of 23.4% and the following summary statistics:
The low means and barely sufficient win frequencies suggest that the MACD may not be an appropriate indicator for trading bonds. The minimal increase in performance from the Any Crossing condition to the Below Zero Line condition reveal that this may be due to a lack of mean reversion present in the bond market, or at least the MACD's failure to capitalize on it.
VI. Conclusion
The MACD period settings found on currencies (FOREX), commodities, and bonds do not produce comparable results to those found on equities and S&P 500 futures. It is also significant that the MACD crossover strategy under default settings produces lesser to negative results on all of these asset classes besides equities. Commodity traders may disregard the MACD as a valuable indicator, even though the most optimal setting yielded a high expected value of 487.31%, because the core, physical circumstances are the underlying drivers for the large price swings which the MACD capitalizes on in the commodity market. Therefore, it is concluded that the MACD indicator is appropriate for equities and S&P 500 futures but does not produce favorable results on other asset classes.
Traditional trading education teaches the MACD crossover strategy under default settings on various asset classes. This shows that much statistical work is required to evaluate the concepts and strategies taught in traditional trading education.
Comments