The Escalator and the Elevator
Why long trades and short trades require independent calibration, and what the data reveals when you test it.
Episode 7 ended with a question. Three forces explain the decline in simple trend-following returns: regime suppression, short-side impairment, and the horizon gradient. But the second force, the apparent death of short-side alpha, was measured using symmetric models that apply the same lookback to both long and short signals. Bull markets last three to five times longer than bear markets. Bear markets deliver intensity per unit time that is fifty to one hundred percent greater. Prices take the escalator up and the elevator down. If the two directions operate at different frequencies, the symmetric model is structurally misaligned with at least one side at all times.
This episode investigates whether the short side is genuinely dead or invisible to a symmetric lens.
The Mechanism
The asymmetry between upward and downward price moves is one of the most robust findings in empirical finance. Lunde and Timmermann, studying two centuries of US stock data, found that bull markets averaged approximately forty-six months in duration while bear markets averaged fourteen months. The median drift in bear markets was fifty to one hundred percent larger in absolute terms than during bull markets. The escalator and the elevator.
This is not an anomaly within the framework established in Episodes 1 through 6. It is a direct consequence of Postulate 1: the spectral state is the aggregate of agent impact. The agent population is not symmetric. The global capital base is overwhelmingly long-biased. Passive index funds, pension funds, sovereign wealth funds, endowments, corporate buybacks, and retail investors through retirement accounts are all structurally long. The ratio of structurally long capital to structurally short capital in equity markets is on the order of twenty to one or higher.
During uptrends, convergent impact builds gradually against the trend. Value investors trim overweight positions. Rebalancers sell winners. Profit-takers exit. The two forces are roughly balanced in magnitude because the long-biased population is comfortable holding. Selling pressure builds slowly. The escalator.
During downtrends, the entire long-biased population becomes a potential source of divergent impact simultaneously. Margin calls force leveraged longs to sell. Stop-losses trigger across systematic strategies. Redemptions force portfolio managers to liquidate. Risk management systems across thousands of institutions hit their drawdown limits at roughly the same time. Meanwhile, the convergent force, bargain hunters, is tiny relative to the wall of forced selling. The elevator.
A symmetric trend model, using the same lookback for both long and short signals, is calibrated to a single frequency. If bull trends operate at low frequency and bear trends operate at high frequency, the symmetric model is permanently misaligned with at least one side.
The Test
We tested this directly. We ran a Donchian breakout ensemble, the same architecture used in Episode 7, across all sixty-eight markets with independent lookbacks for long and short signals. Long entry on the break of a slow channel. Short entry on the break of a fast channel. Both with a flat state: no position until a material breakout occurs. The symmetric ensemble uses seven lookbacks from fifty to five hundred days, each with an exit at half the entry channel. The asymmetric variant pairs each long lookback with a shorter entry channel for shorts.
We ran a grid search across long lookbacks from one hundred to five hundred days and short lookbacks from ten to one hundred days.
Figure 8.1 MAR surface across asymmetric lookback combinations. The optimal point is far from the symmetric diagonal. Shorter short lookbacks dominate across virtually all long lookback pairings.
Figure 8.1 makes the structure visible. The hottest zone on the surface sits in the bottom rows, where short lookbacks of ten to thirty days pair with long lookbacks of two hundred to three hundred days. The optimal single pair, marked on the chart, is two hundred and fifty days long and ten days short, a ratio of twenty-five to one. At the portfolio level, this single asymmetric pair delivers MAR broadly in line with the full symmetric seven-lookback ensemble. The more important finding is the surface itself: shorter short lookbacks consistently dominate across the grid, regardless of which long lookback they are paired with.
What the Asymmetry Reveals
The finding is not what we expected. The asymmetric model does not resurrect short-side standalone alpha. The short side remains near zero in the post-2020 period regardless of calibration. What the asymmetric model does is reduce short-side bleeding during adverse regimes and marginally improve long-side calibration by aligning it with the escalator’s lower natural frequency.
Figure 8.2 Left: short-side cumulative returns, symmetric versus asymmetric. Right: short-side Sharpe by regime. Asymmetric calibration reduces the QE-era loss.
The flat state in Donchian breakout models already provides natural protection that the always-in moving average lacks. When a short trade hits the exit channel quickly, as elevator moves do, the model goes flat rather than reversing. The three-state architecture, long, short, or flat, is itself an implicit asymmetric mechanism. This is why the improvement from explicit asymmetric calibration is meaningful but not transformative at the portfolio level: the symmetric Donchian ensemble already handles much of the frequency mismatch through the flat state.
The Revised Conclusion
The short side is not dead. It is structurally impaired by the asymmetric agent population, and that impairment is real and likely permanent in the post-QE world. But a symmetric model overstates the damage. The short side operates at a higher frequency than the long side, and a model that respects this, using faster lookbacks for short signals, recovers measurable improvement at the portfolio level even though standalone short-side alpha remains negligible.
The short side’s value was never its edge. It was always the portfolio construction: the crisis alpha from Episode 3, the negative bear-market correlation, the convex tail payoff during sustained declines. Asymmetric calibration does not change what the short side provides. It reduces the cost of providing it.
This constitutes a refinement of the framework from Episodes 1 through 6. The frequency of positive feedback is direction-dependent. Due to the asymmetric composition of the agent population, uptrends operate at low frequency and downtrends operate at high frequency. A single lookback is structurally suboptimal. Divergent strategies should treat the two directions as independent signal processes with independent calibration.
Four Forces
With this finding, the three forces from Episode 7 become four, properly ordered by importance.
First, regime. Cyclical and dominant. QE and ZIRP suppressed the variance ratio. When macro conditions deliver genuine trends, the edge returns. This force can reverse.
Second, short-side impairment. Structural but overstated by symmetric models. Central bank backstops and passive flows have severely impaired short-side standalone alpha. Asymmetric calibration reduces the damage but does not eliminate it. Short-side portfolio utility remains intact and essential.
Third, the horizon gradient. The edge has redistributed from shorter to longer horizons. Crowding at visible frequencies. The ensemble case is stronger than ever.
Fourth, the escalator/elevator asymmetry. Bull and bear markets operate at different frequencies. Independent calibration for long and short signals improves portfolio-level performance. This is a frequency mismatch, not a missing edge.
The Future of Trend
The case for divergent strategies is stronger and more nuanced than the monotonic decline narrative suggests.
The feedback structure is regime-dependent, not dead. When macro conditions provide genuine trends, every divergent strategy recovers. The current environment, with higher rates, fiscal dominance, and geopolitical fragmentation, may provide more raw material than the lost decade did.
The edge has relocated, not disappeared. From short horizons to long ones. From symmetric to asymmetric calibration. From uniform signals to those calibrated to the fingerprint. From single-strategy implementations to ensembles that exploit declining cross-horizon correlation.
The feedback structure provides the roadmap. The variance ratio tells you when raw material is available. The asset-class fingerprint tells you where. The escalator/elevator asymmetry tells you that long and short signals require independent calibration. A strategy that reads all four dimensions will extract what a uniform static strategy cannot.
Static, symmetric trend-following is dying. Regime-aware, horizon-diversified, asymmetrically calibrated trend-following is where the edge has gone.
Next
Episode 9 brings the series to a close. It synthesises the full arc: the feedback structure beneath the zero, the three states, the fingerprint, the paradox, the escalator and the elevator, and what it all means for how we understand markets. The zero is the most important number in finance. Beneath it lies a machine.
Endnotes
Methodology
- Escalator/elevator asymmetry: Donchian breakout ensemble with three-state (long/short/flat) architecture across 68 CSI ratio-adjusted continuous futures contracts, 8 asset classes, September 1984 to January 2026. Symmetric ensemble: 7 lookbacks (50, 100, 150, 200, 300, 400, 500 days), exit at half entry channel. Full-sample: MAR 0.616, Sharpe 1.206. Long-side Sharpe 1.143, short-side Sharpe 0.368.
- Asymmetric grid search: long lookbacks 100–500d in 50d steps, short lookbacks 10–100d in 10d steps. Best single pair: 250L/10S, MAR 0.615, Sharpe 1.241. Lookback ratio 25:1. Shorter short lookbacks (10–30d) dominate across virtually all long lookback pairings.
- All returns vol-scaled to 10% annualised target using 63-day rolling volatility, capped at 3.0x. Equal-weight portfolio across all available contracts and lookbacks. No transaction costs, no slippage.
- Empirical asymmetry references: Lunde and Timmermann (2004), Duration Dependence in Stock Prices, Journal of Business and Economic Statistics.
Data and Figures
- Same dataset as Episodes 1 through 7. 68 contracts, 8 asset classes.
- Figure 8.1: MAR surface across asymmetric lookback grid.
- Figure 8.2: Short-side cumulative returns and regime Sharpe, symmetric vs asymmetric.
This research series is drawn from The Fractals of Finance: Determinism, Adaptation and the Geometry of Markets
The book explores the full architecture of feedback, fat tails, and fractal structure in financial markets, and what it means for how we trade, invest, and understand risk.
Available now on Amazon in paperback, hardcover, and Kindle.
Want a practical field manual for trading trends and capturing outliers?
The Aussie Turtles Trend Following Guide: A Field Manual for Hunting Outliers adapts the timeless principles of the original Turtle traders into a systematic, rules-based approach for modern markets. Co-authored with Adam Havryliv.
Available now on Amazon in paperback, hardcover, and Kindle.