A 94th-percentile week also carries 0.113% eight-hour funding
Should the research team treat this week as confirmation that the signal works, or as an outlier week to be excluded from the track record while sizing stays capped?
The market facts are real; the decision maker is illustrative.
The week and the position
In the week of 15 February 2021, BTC rose 18.17%, from 48,580.47 to 57,408.57, closing 1.33% below its 52-week high with a return at the 94th percentile of 157 weeks of history. Three of the prior four weeks were up. Volume ran 1.38 times the prior 12-week average, the range spanned 28.05%, and the largest hourly move was only 2.44%. Eight-hour perpetual funding averaged 0.113%, peaking at 0.1563%. ETH lagged BTC by 10.75 percentage points at a 0.73 hourly correlation. An illustrative research team runs a trend signal and decides how this week enters its evidence.
Judgment
A week in the 94th percentile of 157 carries almost no independent information about a signal's average edge, and positive funding means long positions paid short positions: 0.113% per eight hours compounds towards roughly 2.4% over the week, a real cost against any trend gain. Keep the signal running at its existing, pre-week size, record the week as valid but weight it as a single outlier observation, and separate the price return from the funding paid so the edge is not overstated.
What would change it
If the team's written evaluation rule treats high-percentile weeks as valid signal evidence, excluding it is inconsistent and the size cap should be lifted. If subsequent weeks' eight-hour funding readings settle near this week's 0.0406% minimum, the carry cost objection weakens and sizing can be reviewed.
What sets this case apart
This is not about whether the signal made money; it is about whether one week in the top decile of 157 can validate anything, when the cost of holding longs was itself extreme.