A 97th-percentile ETH week carried by one 15.63% day tests whether a signal survived
Should the research team keep its ETH signal unchanged after a week whose 23.19% return came largely from a single 15.63% day, or suspend it until a calmer week reproduces the edge?
The market facts are real; the decision maker is illustrative.
The week and the position
In the week of 1 April 2019, ETH rose 23.19%, a 97th-percentile return against 59 weeks of history, with a 27.4% range and 110.8% annualised realised volatility. Dollar volume ran 2.27 times the prior 12 weeks. The best day, 2 April, added 15.63%; the worst, 3 April, lost 2.02%, yet that Wednesday also held the week's largest hourly move, -5.5% at 22:00. The close of 174.21 sat 90.6% up the weekly range but still 79.25% below the 52-week high, and ETH trailed BTC by 2.84 percentage points. The prior four weeks moved only +3.65, +2.44, -1.86 and +3.86 per cent. An illustrative team decides whether the signal earned its place.
Judgment
Keep the signal at its current status and size, and attribute the week before drawing a conclusion. Recompute the week's result with 2 April removed. If the edge survives that day's removal, the week supports the signal; if it does not, this week is evidence of exposure to single-day jumps, not of a repeatable edge. Accept that a 97th-percentile week may not recur soon, since the prior four weeks were mild. Do not raise the signal's allocation on an outlier week.
What would change it
If removing 2 April turns the week's signal return negative, suspend the signal pending a week whose largest daily move is under 5%. If the edge reappears in a subsequent week with no single day above 5%, the signal can be promoted.
What sets this case apart
This is a validation question, not a directional one: the issue is whether one anomalous day supplied the evidence, regardless of the week's sign.