Factor board · conditional distributions · n >= 30 gate

Factor Board

Each metric ByKaranteli records daily is placed in its own historical band (quantiles of its full history). For the band today sits in, the table shows the median BTC move over the next 7 days and how that compares with the median of all days. A large gap is an observation about the past, not a forecast; bands with fewer than 30 matured days show nothing.

Bands that separated from the base raterec · daily 00:50 UTC
3 metrics|edge| >= 0.5 pp over 7 days
built 2026-08-21 02:05 UTC
Metrics with history
14/30
Thin bands (n < 30)
1
Largest gap
Fed balance sheet +1.2 pp
01

All metrics, largest gap first

Edge is the 7-day median BTC move after days in today's band minus the median after all days, in percentage points. Up 7d is the share of those days that closed higher a week later.

MetricTodayBandDaysNext 7dBaseEdgeUp 7dSince
Fed balance sheet$6746BMid 3/593+1.6%+0.4%+1.2 pp58%2017-08Open
Reverse repo (RRP)$0BVery low 1/5466+1.6%+0.5%+1.1 pp59%2017-08Open
BTC funding, daily+0.029%Mid 3/51074-0.2%+0.4%-0.6 pp49%2019-09Open
Coinbase premium-0.018%Low 2/5656+0.0%+0.5%-0.5 pp50%2017-08Open
Fed funds effective rate3.63%Mid 3/5680+0.1%+0.5%-0.4 pp51%2017-08Open
Stablecoin supply$307BVery high 5/5630+0.0%+0.4%-0.4 pp50%2017-11Open
BTC hashrate843 EH/sVery high 5/5620+0.2%+0.5%-0.4 pp52%2017-08Open
BTC difficulty127.5TVery high 5/5620+0.2%+0.5%-0.4 pp52%2017-08Open
Kraken premium-0.033%Low 2/5143-0.0%+0.3%-0.3 pp50%2024-08Open
Spot ETF net flow+$517MHigh 3/3102-0.1%-0.4%+0.3 pp50%2025-05Open
Deribit DVOL (BTC)37.8Very low 1/5389-0.1%+0.1%-0.2 pp49%2021-03Open
Fear & Greed62High 4/5631+0.6%+0.4%+0.2 pp55%2018-02Open
US 10-year yield4.65%Very high 5/5466+0.5%+0.5%-0.0 pp56%2017-08Open
BTC flow toxicity (VPIN)0.159High 3/328thin--0.5%--2026-05Open

Bands are quantiles of the metric's own history (3 bands for short histories, 5 for deep ones). Today's value is part of the band definition but not of the outcome statistics, which use only days with a matured 7-day return.

02

Methodology and honest limits

Data: the metric_daily table (CC0), one value per metric per UTC day, joined with the BTC close 1, 3 and 7 calendar days later. Bands are quantile breaks over the full recorded history; the statistics per band are medians and up-shares, never means, so single outliers do not drive the read.

The n >= 30 gate is the same one every instrument page uses: below it the band shows nothing rather than a number that would not survive a second look. Many metrics have short histories (recording started in 2026), so their bands are 3-way and some remain thin for months.

Ordering by absolute edge is a display choice, not a ranking of predictive value. With 30 metrics and 3-5 bands each, a few large gaps are expected by chance alone; the board exists to make that visible, not to hide it.

03

FAQ

Why BTC as the outcome for every metric?

It is the one outcome every metric can be compared on, and the asset most of these metrics describe. Per-asset outcomes would shrink n and invite cherry-picking.

Why do some metrics show no edge at all?

Because the data says so. A metric whose bands all lead to roughly the base rate is reported exactly like that: no edge either way.

Where is the full per-band table?

On each metric's own page (for example /premium, /etf, /options, /macro, /network) and in the research note on band-conditioned returns.

Related

Band-conditioned returns
The research note behind this table.
Metric context API
Per-metric bands and outcomes, CC0.
Market Indices
Euphoria, Fear & Greed, dominance.
Free Datasets
metric-daily and the rest, CC0.