TRX's Funding Percentile Read 81. The Rate Was Still Negative

I added a percentile score to my funding-rate tracker a week and a half ago: for each coin, rank today's OI-weighted funding rate against every daily reading collected so far. The idea was simple — "this coin's funding is high" means nothing without knowing high relative to what. This morning I pulled the sheet to sanity-check it and found TRX sitting at the 81st percentile. Sounds like elevated, crowded-long funding. The actual OI-weighted rate underneath it: -0.0042%. Negative. Shorts were getting paid.

My first thought was a bug. It isn't. It's exactly what an 8-day-old percentile does when you read it like a mature one.

What the numbers actually said this morning

Snapshot from the tracker, 13 September, 23:00 UTC, ranked by percentile against each coin's own history so far:

• DOT — 6th percentile, rate -0.0193%. Lowest rank, most negative rate. Consistent.
• ATOM — 19th percentile, rate -0.0145%. Also negative, also low. Also consistent.
• APT — 19th percentile, rate -0.0008%. Same percentile bucket as ATOM. The rate is 18x smaller.
• TRX — 81st percentile, rate -0.0042%. High rank, negative rate. Not consistent.
• BTC — 94th percentile, rate +0.0063%. High rank, positive rate. Consistent.

DOT and BTC behave the way I'd expect a percentile to behave: the rank and the sign agree. ATOM and APT expose the first problem — two coins can land in the exact same percentile bucket while one's raw funding rate is eighteen times the other's. TRX exposes the second, sharper problem. The rank can point positive while the number itself is negative.

Why a young percentile does this

The tracker has 8 days of daily readings per coin so far, building toward a 90-day window. With 8 points, a rank-based percentile only has 8 possible values it can land on — roughly 6, 19, 31, 44, 56, 69, 81, 94. That's the whole scale. It's not measuring "how extreme is today's funding" on a smooth 0-100 line, it's sorting 8 numbers and reporting where today fell in that short list. TRX's funding over its last 8 recorded days must have spent more time even more negative than -0.42bps — so today's reading, while still negative in absolute terms, is the least negative of a bad recent run. High percentile, bad sign, both true, because "high" here means "least negative," not "positive."

APT and ATOM sharing a bucket is the coarseness problem, not the sign problem. With only 8 slots, two coins with very different funding regimes can tie on rank even though their actual carry cost is nowhere close. I'm not throwing the metric out. I'm not trusting it alone until it has enough history that a bucket means something narrower than "one of eight roughly-equal-sized buckets."

What I'm actually using it for right now

Until the window fills — 82 more days at this rate — I'm reading the percentile as a direction-of-change signal only, paired with the raw rate, never instead of it. "TRX's funding rank moved up this week" tells me shorts are getting paid less than they were, which is useful context. It doesn't tell me the sign, and treating a high rank as "go long, funding favors it" would have been exactly backwards this morning. The raw number is still the one that decides whether a position bleeds or gets paid — check that on the live funding rates page before the rank, not after. If you're holding anything that pays or costs funding, run the actual carry through the funding calculator or the funding rate APR calculator — a percentile rank won't tell you what it costs you by Friday, only the rate will.

I've made this mistake before in a different shape — writing off a repeated funding number as noise when it was actually a floor. Same lesson twice now: a derived metric built to make a raw number easier to read can flip the read entirely if you don't know what it's derived from. I'm keeping the percentile column. I'm just not letting it near a trade decision until the sample is bigger than 8.