2. While prices are set at the margin, CTA trading (esp. ex-BW) still represents a single-digit proportion of total trading.
3. Trend is a low signal indicator. #MachineLearning techniques are unlikely to emphasize trend features. More likely they will be combined with other features in Bayesian / decision tree frameworks, but they won’t be dominant trading signals.
4. Trend is very hard to stick with. Long periods - sometimes several years - of waiting followed by abrupt and (definitionally) unpredictable windfall profits. Most people abandon ship before the profit materializes.
#BeFiFails
#BeFiFails
5. Many institutions still consider trend “woo-woo”. They’re reluctant to allocate in the first place, & quick to pull funds when the strategy underperforms.
Most instos are stacked with biz-people at the board level, and MBA/CFAs on the IC:fundamental types. Trend is anathema.
Most instos are stacked with biz-people at the board level, and MBA/CFAs on the IC:fundamental types. Trend is anathema.
6. I fundamentally believe the challenges for trend-following in the recent decade have been largely attributable to a global policy of governments leaning against inconvenient trends.
Stability breeds instability. Trend thrives on large surprises.
Stability breeds instability. Trend thrives on large surprises.
7. Even if (a big IF for the reasons above) the trend premium has diminished, it still has outsized potential value in a portfolio due to its convexity profile (conditional correlation). It has historically been most helpful when everything else is getting smashed.
Overall I'm still a fan of trend-following as a core weapon in the arsenal. We like to combine a variety of macro-edges so we're not overly reliant on any one #alpha source, but trend is a key plank in the alpha stack.
