Michael Covel: Trend Following

Michael Covel is an author and seller of trend following trading systems for individual traders. I borrowed this book at the public library, I enjoyed reading it but it is optimized for motivation and engagement rather than conveying actual information.

Michael Covel Image credit: wikicommons

• He mainly gives life-advice and motivation talks about agency and self-reliance, this corresponds to the first edition of the book, which is about 300 pages.
• After that, there are 300 pages of podcast transcripts and articles, followed by a list of the hundreds of podcast he recorded without context or abstract
• He interviewed Nobel Prize Harry Markowitz early in his podcast. After that, he could invite and interview many other Nobel Prize winners, and had access to successful trend follower traders such as Ed Seykota,
• He interviewed successful of the top 10 momentum traders and CTA, but the interviews were done after the traders have had their successful run. This begs a question of survivorship bias that is never addressed. Only Seykota points to testing his ideas rather than applying a recipe.
• He does not disclose his own track record as a trader, despite being in this trend following gig for more than 20 years
• He is a former GMU (George Mason University) graduate living in Thailand, he has a libertarian bend. He lumps together MPT (Modern Portfolio Theory), EMH (Efficient Market Hypothesis), and the Welfare State.
• He claims that Sharpe ratios are a useless metric because trend following PNL is skewed, and yet he finds that the average return is a useful metric. He does not discuss the use of volatility for position sizing.
• The technical content of the book can be summarized to the following: a trading strategy has entry signals, position sizing, and exit signals. He does not describe ATR or volatility.

Some podcast transcripts are instructive. An interesting episode is that Ed Seykota explains that as an intern, he met Mr Donchian who was working for a broker to propose trading signals for clients. Donchian had the idea of running maximum signals but never systematically tested his strategy. When Ed used the massive computer center of the broker to run simulations, he found a strategy based on running maximum that worked.

The interesting part was that there was an immediate conflict of interest, as the strategy generated more profits but annual or less trading signals and therefore less commission for the broker. Ed could not work for a broker due to the misalignment.

The articles are generally more informative than the book, although they look more like marketing than methodology:

• A Multicentenial View of Trend Following
• Two Centruries of Trend Following by Y. Lempérière, C. Deremble, P. Seager, M. Potters, J. P. Bouchaud 2014, monthly signal on $s_t=\frac{P_t-Ewm(5)}{\sigma(5)}$ where $\sigma(5)$ is the ewma of absolute price move. Sharpe ratio of 0.8.
• Trend Following, Quality not Quantity Anthony Todd, Martin Lueck, 2016. Evaluates 13 trend following strategies on 146 asset vs their current one. They all have roughly 1 sharpe ratio. There is no diversification benefit mixing trend follow strategy. The paper does not specify sharpe ratio by strategy, trading costs, trading frequency, or compare theoretical results with their actual.
• Evaluating Trading Strategies Campbell Harvery and Yan Liu, 2014, for a single test, $t-stat = SR x \sqrt{nb years}$ the best test out of $n$ sees its t-stat divided by $n$ according to the Bonferroni correction. Authors suggest using BHY correction to ensure p-values are significant.
• Black Box Trend Following - Lifting the Veil Nigol Koulajian Paul Czkwianianc 2010. Point out at crossover strategies from 10-100 to 10-200 all outperform big CTA, but the latter use the longer lookback to reduce trading frequency. The shorter lookback periods allow for smaller drawdowns.
• Risk Management Ed Seykota 2005. Quotes Kelly and otherwise advises MC for position sizing.

Harnessing Grok: Trend Following Strategies

Trend following, often synonymous with time-series momentum, involves taking long positions in assets with positive past returns and short positions in those with negative returns, typically scaled for volatility to maintain consistent risk. This differs from cross-sectional momentum by focusing on an asset's own historical performance rather than relative ranking among peers. Common implementations use moving average crossovers (e.g., exponentially weighted moving averages or EWMA) or simple past return signs over a lookback period, with positions rebalanced periodically.

Common Parameters

• Lookback Periods: Typically range from 1 to 12 months for return calculations or trend detection. Shorter periods (1-3 months or weekly/daily equivalents) capture "fast" trends, while longer ones (6-12 months) target "slow" trends. For EWMA crossovers, turnover rates define speed: fast (1-week), medium (6-week), slow (13-week).
• Holding Periods: Often 1 month for monthly rebalancing, but can extend to 3-12 months with overlapping portfolios. In alternative setups, holding is half the lookback (e.g., 1-week hold for 2-week lookback).
• Other Adjustments: Positions are frequently scaled by inverse volatility (e.g., targeting 40% annualized or using 60-day rolling estimates like Yang-Zhang). Trading signals may use the sign of returns or t-statistics for statistical significance (e.g., t > 2 for long, t < -2 for short).

Typical Sharpe Ratios

Sharpe ratios (risk-adjusted returns, often gross before fees/costs) vary by asset class, period, and speed, but generally range from 0.8 to 1.2 for diversified portfolios (e.g., futures across equities, bonds, commodities, currencies). Performance has declined over time, especially for faster strategies, with gross values around 1.0-1.25 historically but lower post-2000. Net ratios after fees (e.g., 2/20 structure) drop to 0.7-0.9.

Strategy Type/Speed	Lookback/Holding Example	Assets/Period	Sharpe Ratio (Gross)	Notes/Source
Fast Trend	2-week lookback / 1-week hold; 1-week turnover (EWMA)	20 futures (1984-2013)	0.87	Declines significantly post-2004; alternative past-return version: 1.38 (1984-1993) to -0.01 (2004-2013).
Medium Trend	1-quarter lookback / 33-day hold; 6-week turnover (EWMA)	20 futures (1984-2013)	1.12	More stable; alternative: 1.19 (1984-1993) to 0.59 (2004-2013).
Slow Trend	1-year lookback / 6-month hold; 13-week turnover (EWMA)	20 futures (1984-2013)	0.81	Least decline; alternative: 1.26 (1984-1993) to 0.60 (2004-2013).
Monthly TSMOM	12-month lookback / 1-month hold	71 futures (1978-2012)	1.21-1.27	Representative: M^{12,1} at 1.21; net after fees ~0.81. Higher in recessions.
Weekly TSMOM	8-week lookback / 1-week hold	71 futures (1978-2012)	1.23-1.29	Representative: W^{8,1} at 1.25; net ~0.85. Low correlation to monthly (~22%).
Daily TSMOM	15-day lookback / 1-day hold	71 futures (1978-2012)	1.22-1.27	Representative: D^{15,1} at 1.25; net ~0.86. Highest turnover.
Diversified TSMOM	12-month lookback / 1-month hold	58 futures (1985-2009)	~1.0-1.2	Positive across all instruments; by class: commodities ~0.4-0.6, equities ~0.3-0.5, fixed income ~0.5-0.9, currencies ~0.2-0.4.
Volatility-Adjusted TSMOM	12-month lookback / 1-month hold	75 futures (1975-2013)	0.82-1.04	Baseline 0.83; improved with YZ estimator and 3-month window to 1.04; TREND signal reduces turnover by 66% with similar Sharpe (~0.99).

To arrive at these Sharpe ratios, calculate as annualized excess return divided by annualized volatility (e.g., for a monthly strategy with mean excess return μ and standard deviation σ, Sharpe = (μ 12) / (σ √12)). Historical data shows positive skewness for faster strategies but tension with Sharpe due to occasional large drawdowns.

Cross-Sectional Momentum Strategies

Cross-sectional momentum ranks assets by past performance, going long the top performers (winners) and short the bottom (losers) to form zero-cost portfolios. This is relative ranking, unlike trend following's absolute signals. Classic work (e.g., Jegadeesh and Titman, 1993) focuses on stocks, but extensions apply to other assets.

Common Parameters

• Lookback Periods: 3-12 months for ranking past excess returns (often 6 or 12 months; 1-month gap to avoid reversals).
• Holding Periods: 1-12 months, with monthly rebalancing and overlapping portfolios; commonly 1-3 months.
• Other Adjustments: Top/bottom 10% (deciles) for winners/losers; equal or value-weighting; volatility scaling optional. Optimized versions use mean-CVaR with time-varying risk aversion (e.g., based on VIX or market volatility) to avoid crashes.

Typical Sharpe Ratios

Sharpe ratios are generally lower than trend following (0.06-0.8), reflecting higher volatility and crash risks (e.g., -70%+ drawdowns in 1932/2009). Optimized versions improve to ~0.5-0.8. Performance is stronger for shorter lookbacks in some markets but varies by period (e.g., higher pre-1990).

Strategy Type	Lookback/Holding Example	Assets/Period	Sharpe Ratio	Notes/Source
Standard XSMOM	12-month lookback / 1-month hold	Swedish stocks (1999-2022)	0.06	Annualized return 2.03%, volatility 32.6%; better for short lookbacks (e.g., 3-month: mean monthly return 0.41%).
Standard XSMOM	3-month lookback / 1-month hold	Swedish stocks (1999-2022)	~0.15 (implied)	Higher returns than 12-month but insignificant; alpha 0.0148 in Carhart model.
Baseline Heuristic XSMOM	9-12 month lookback / 1-month hold	US equities (1926-2020)	0.51	Excess return ~12-15%; prone to crashes (-91% in 1932); post-1990: 0.39.
Volatility-Adjusted Mean-CVaR XSMOM	9-month lookback / 1-month hold	US equities (1926-2020)	0.73-0.78	Best with 6-month mean/3-month volatility windows; avoids crashes; UP ratio higher than baseline.
Reward-Risk Mean-CVaR XSMOM	12-month lookback / 1-month hold	US equities (1926-2020)	0.26-0.50	Underperforms baseline due to mean estimation issues; best with 12-month windows.
VIX-Adjusted Mean-CVaR XSMOM	9-month lookback / 1-month hold	US equities (1990-2020)	0.47	Outperforms baseline (0.39) post-1990; lower excess return but crash-resistant.

Sharpe calculation follows the same formula as above. Studies note cross-sectional momentum complements trend following, with combinations boosting overall Sharpe (e.g., by 45-66% in diversified portfolios).