Copper Trend System After a Whipsaw Quarter
A commodity trading advisor runs a long/short system on COMEX copper futures using a 20-day/100-day EMA crossover. Over the past three months the system has produced seven round trips and a net loss of 4.8%. ADX(14) is 14 and falling; the 100-day EMA is flat; open interest has declined 12% over six weeks; the front spread moved from $0.012 contango to $0.006 backwardation. Annualized realized volatility is 28%, versus a six-month average of 18%. The fund risks 1.5% of equity per trade and is up 6% year to date. The risk committee asks whether to keep trading the system unchanged, cut position size, or add a regime filter. The portfolio manager worries that overriding a backtested system after a drawdown is curve-fitting, but also notes that the system's average win/loss ratio has fallen from 1.9 to 1.2.
1. Diagnose whether copper is in a trendless regime or a transitional one. Use the ADX, EMA slope, open interest, spread, and volatility data, and explain what each indicator does and does not tell you.
Read the worked answer
The evidence favors a trendless, whipsaw regime rather than a confirmed new trend. ADX(14) at 14 and falling is below the common 20 threshold and indicates weak directional movement; it measures trend strength, not direction. The flat 100-day EMA confirms no sustained intermediate trend, and seven round trips with a falling win/loss ratio are consistent with range conditions. Declining open interest (-12% over six weeks) suggests fading participation and less conviction behind moves; it does not by itself predict direction. The move from contango to backwardation can reflect physical tightness or near-term supply stress, but a $0.006 backwardation is modest and is not a standalone long signal. Realized volatility at 28% versus 18% raises noise and the probability of stop-outs. A trend-following crossover is designed for persistence; in this regime its edge is weak. The correct diagnosis is 'trendless with elevated noise,' not 'broken system.'
2. Recommend whether to keep the system unchanged, reduce size, or add a regime filter. Give a specific implementation that respects the 1.5% per-trade risk limit.
Read the worked answer
Do not abandon the system after a 4.8% drawdown, but do not trade it at full risk in a low-ADX, high-volatility regime. A reasonable compromise is to keep the system logic but add a regime overlay: (1) reduce per-trade risk from 1.5% to 0.75-1.0% while ADX(14) is below 20 and the 100-day EMA slope is flat; (2) require a minimum separation between the 20- and 100-day EMAs or a close beyond the recent 20-day range to take a new signal; (3) use ATR-based sizing so the higher 28% volatility automatically cuts contracts. For example, if the stop is 2 ATR and ATR has risen roughly 55% (28/18), the position size should fall by about one-third even at constant risk. Keep a written rule for restoring 1.5% risk after ADX exceeds 20 and the EMA slope turns up or down. This preserves the system's long-run edge while avoiding curve-fitting to the last seven trades.
3. What confirmation would justify re-risking to full size, and what would tell you the regime filter is failing?
Read the worked answer
Re-risk when the market demonstrates persistence, not just a single breakout. Confirmation would include ADX(14) rising above 20, the 100-day EMA slope turning positive or negative, open interest expanding on the directional move, and the 20-day EMA holding above/below the 100-day EMA for at least 10-15 sessions. A widening backwardation with rising open interest would add physical-market confirmation for a long bias. Re-risk in steps, e.g., 0.75%, then 1.0%, then 1.5%, as these conditions persist. The filter is failing if it suppresses profitable trends for a full cycle: track the counterfactual system's signals and compare hit rate and average win/loss. If filtered signals underperform unfiltered signals over a pre-specified sample (e.g., 30 trades), revise the filter. Also monitor whipsaw frequency and slippage; if round trips remain high after ADX recovers, the problem may be execution or contract roll, not regime.