Skip to main content

Market & Strategy Risk

The main failure mode​

Trend-following loses in choppy, mean-reverting markets. When price oscillates around a level rather than moving in a direction, the signal flips repeatedly. The program buys after each move up and sells after each move down, paying the spread and the impact each time, and none of the positions run long enough to pay for the ones that did not.

This is the ordinary cost of the strategy, not an edge case. The payoff shape is many small losses funded by a few large gains. The 2010s were largely range-bound and the CTA category performed poorly across the decade.

Expect extended flat-to-negative periods. They are not evidence that something has broken.

Drawdown​

The deepest drawdown in the simulated period from April 2025 to June 2026 was 9.3%, over 92 days.

⚠️ That is one simulated window, not a limit. A strategy targeting 25% annualised volatility can draw down well beyond 9.3% without anything having gone wrong.

Model risk​

The strategy is calibrated on historical data. Trend persistence is documented over 50 years and 100+ markets, but crypto microstructure is young and evolving. The effects being harvested could weaken as these markets mature and as more capital chases them.

Simulated performance flatters​

All performance before 27 May 2026 is simulated on historical execution data, net of modelled commission, slippage and funding. Modelled frictions are not real frictions. Live trading reveals costs a backtest does not contain: queue position, adverse selection, and the price impact of being the marginal participant in a thin market.

Live execution has tracked the backtest within approximately 7 basis points per day since inception. That is a sample of weeks.

Concentration of outcome​

The portfolio is diversified by risk contribution, not by outcome. In the simulated window one sector, Metals, produced 37% of the return while Energy detracted.

That is the construction working as designed. It also means a given year's result may be dominated by a small number of sleeves, so twelve months of data is not evidence that the diversification is delivering, in either direction.

Volatility targeting cuts both ways​

Scaling positions down as volatility rises bounds the risk profile. It also means the program is smallest exactly when a violent move is underway, so in a sharp sustained rally it participates less than a static-weight portfolio would.

Leverage​

The public vault is fully collateralised. The strategy does not gear the portfolio to manufacture return. Perpetual positions still carry margin mechanics, covered in Venue, Counterparty & Curator Risk.

What would make the thesis wrong​

  • Trends stop persisting in crypto as the market matures and the flow that created them is arbitraged away.
  • Correlations between sectors rise toward the levels seen inside crypto-only portfolios, removing the diversification the construction depends on.
  • Tokenized RWA markets stay too thin to carry a meaningful risk allocation, removing the three sleeves that supply most of the low correlation.
  • Live execution costs come in materially above modelled costs.