Bitcoin vs 200-day average
−18.8% below it
Lower than on 79% of days since Mar 4, 2018; a Mayer Multiple of 0.81
Its close against the average of its last 200.
· Bitcoin's daily closes
Patterns · what moves each coin · 84,557 tests every night
Not every coin against every other: each coin on its own, against its own moves on any day. Bitcoin's and Ether's big days, its sector's, its own chart, its own news and the US releases, each tested on the coin's own history. Past patterns, not financial advice.
Proven drivers
222
on 141 coins: past the cut for every test at once
Earned out of sample
6
kinds of driver and horizon, of 31 proven and 64 likely tested
In play now
40
24 earned calls among them
Coming up
40
on the calendar in the next 30 days
Earned out of sample: Ether fell 4% or more in a day (1d: right 54% of 5,799 calls since 2019, against 49% for that side on any day); its sector fell 5% on average in a day (1d: right 57% of 1,678 calls since 2019, against 49% for that side on any day); Bitcoin fell 3% or more in a day (likely drivers, 1d: right 56% of 18,323 calls since 2020, against 49% for that side on any day); Ether fell 4% or more in a day (likely drivers, 1d: right 55% of 16,560 calls since 2019, against 48% for that side on any day); its sector fell 5% on average in a day (likely drivers, 1d: right 58% of 3,646 calls since 2019, against 49% for that side on any day); it rose three times its usual daily move (likely drivers, 7d: right 63% of 591 calls since 2019, against 51% for that side on any day).
Model map · what moves what
84,557 tests last night
222 proven · 3,886 likely · 949 coins · 65 kinds of event
683 in play now
measured
Proven: past one 10% false-discovery cut over every test of the night, both halves of the coin's history agreeing, 12 cases or more. Likely: the same lean on weaker evidence.
likely links drawn: the 1,500 strongest of 3,886
Tap a node to light its links · arrow keys step through them · Esc clears
Market regime · read every night
12 gauges, each dated · newest reading
Bitcoin vs 200-day average
−18.8% below it
Lower than on 79% of days since Mar 4, 2018; a Mayer Multiple of 0.81
Its close against the average of its last 200.
· Bitcoin's daily closes
Bitcoin vs 50-day average
−24.1% below it
The lowest since Feb 5, 2026
Its close against the average of its last 50.
· Bitcoin's daily closes
Bitcoin's 30-day volatility
45% a year
Lower than on 67% of days since Sep 16, 2017
How far its daily moves swung over 30 days.
· Bitcoin's daily closes
Perp funding, typical coin
+1.8% a year
The median of 246 coins' perpetuals: above zero, longs pay shorts.
LIKELY When a funding extreme high came, GRIFFAIN ended higher a week later more often than on any day · 10 cases in its history · no clear edge out of sample
· Hyperliquid and Binance
Fear & Greed index
12 of 100
Last at an extreme: extreme fear at 13 on Nov 14, 2026
Extreme fear: 15 or under.
LIKELY When an extreme fear reading came, BTC ended lower a week later more often than on any day · 21 cases in its history · no clear edge out of sample
· alternative.me
Altcoin season index
70 of 100
35 of 50 ahead of Bitcoin; what came after each season
How many of the 50 most traded altcoins beat Bitcoin over 90 days, out of 100: 75 or over is an altcoin season, 25 or under a Bitcoin season.
· our daily closes
ETH/BTC vs its 200-day average
0.94 of its average
Under it; what Ether did against Bitcoin after each level
Ether's price in Bitcoin (0.0275) against its own average of the last 200 days: under 0.80 is far under it, 1.20 or more well over.
· our daily closes
Bitcoin implied volatility
52 a year
Last low: 35 on Sep 26, 2026
Deribit's DVOL: the yearly move Bitcoin's options price in.
LIKELY When Bitcoin's options priced a big move, BTC ended higher the next day more often than on any day · 9 cases in its history · not tested out of sample yet
· Deribit
Coinbase premium
+0.004% on Coinbase
Last discount: −0.05% on Oct 30, 2026
Bitcoin's close on Coinbase against Binance's: above zero, US buyers pay more.
LIKELY When Bitcoin sold at a discount on Coinbase, FOREX ended higher 30 days later more often than on any day · 12 cases in its history · no clear edge out of sample
· Coinbase and Binance
Korean premium, typical coin
+0.1% on Upbit
Higher than on 70% of days since Sep 1, 2025
The median of 261 coins on Upbit against the world, both in Bitcoin.
LIKELY When Korean buyers paid up for it on Upbit, POKT ended higher the next day more often than on any day · 10 cases in its history · no clear edge out of sample
· Upbit
Stablecoins, 7-day change
−$3.41B −1.3% in a week
Lower than on 99% of days since Dec 6, 2017
The supply of the 6 biggest stablecoins the collector keeps, against 7 days before.
LIKELY When a big stablecoin print came, VIDT ended higher 30 days later more often than on any day · 22 cases in its history · no clear edge out of sample
· DefiLlama
Spot ETF flows
0 in · 3 out big days
Latest: Bitcoin ETFs: $582M out in a day (Nov 20, 2026)
One light a day for the last 30: lit, US spot Bitcoin and Ether ETFs took in far more than usual; a ring, they paid out far more.
LIKELY When a big spot ETF outflow day came, BTC ended lower the next day more often than on any day · 18 cases in its history · no clear edge out of sample
· SoSoValue
§ 01 · In play now40 open
as of · its latest case is still inside its window
§ 02 · Coming up40 dated
as of · its next date is on the calendar
§ 03 · The strongestMEASURED
as of · any coin: proven first, then likely, the biggest effect first
Source: every coin's own daily closes, its events and the market-wide releases on record · fetched
§ 04 · The whole market
as of · the series we collect, tested on Bitcoin, Ether and the altcoins
Each line is one of the outside series we collect every night (rates, the dollar, stablecoins, flows, positioning, attention), its jumps, drops, one-year highs or lows tested the same way as a coin's drivers, against the market's moves on any day. Proven ones pass the cut for every test at once; likely ones lean the same way on weaker evidence. Proven first, then the biggest effect. None of them is tested out of sample yet, so none carries a chance or makes a call.
§ 05 · How it is tested
as of · methodology
A driver is judged against the coin's own moves on any day, never against other coins, so a coin that simply rose all year is not credited to whatever happened along the way.
For each coin with a year of closes, every candidate cause is tested on that coin alone: Bitcoin rising or falling 3% in a day, Ether 4%, the 100 biggest altcoins or the coin's own sector 4 to 5% on average, the coin's own chart (a crash day, a pump day, a volume spike, a new 90-day high or low, a new one-year low, a seven-day streak, a cross of its 200-day average), its own news (listings, delistings, unlocks, funding extremes, the Korean premium and more) and the market-wide releases (the Fed, CPI and the other US releases, options expiries, implied volatility, ETF flows).
A case is read from the coin's close on the day it happened, or from the close before for news dated to a day, to its close the next day, a week later and 30 days later; a run of the same cause counts once per window, so cases never overlap. Each coin's moves are capped at its own 2.5% and 97.5% marks, so one giant pump cannot make a driver.
Every test of the night goes through one Benjamini-Hochberg cut at 10%. Proven means past that cut, both halves of the history leaning the same way, the coin ending higher more or less often than on any day by a clear margin, and 12 cases or more. Likely means the evidence is weaker. On random prices with twenty thousand random events, nothing comes out proven. In play means the latest case's window is still open; coming up means its next date is on the calendar.
Proven on the whole history is not the same as right the next time. So every night the discovery is also re-run as it stood at the start of each calendar quarter, on what came before it only, and the quarter after it is called with the drivers it found: each call takes its driver's side and is scored once its window is over, the proven drivers and the likely ones counted apart. A kind of driver earns a call on the site only when, counting every day once, its side was right more often than that same side won on any day of the coin's quarter, in both halves of the quarters walked, at a Newey-West t of 1.645 or more, on 100 or more scored calls over 60 or more separate days, and the average move on the call's side was above 0%, overall and in both halves; a likely driver's call says it is one. The chance a driver's own cases gave is not shown: out of sample, every proven kind's 1-day calls together stated 63% for their side and were right 54% of the time.
Every proven kind together, out of sample (as of Nov 21, 2026): 1d: right 54% of 19,032 calls since 2019, against 49% for that side on any day; 7d: right 52% of 145 calls since 2021, against 50% for that side on any day; 30d: right 80% of 5 calls since 2023, against 61% for that side on any day.
Every likely kind together, walked apart from the proven ones: 1d: right 54% of 86,443 calls since 2019, against 49% for that side on any day; 7d: right 51% of 13,628 calls since 2019, against 52% for that side on any day; 30d: right 43% of 1,278 calls since 2020, against 45% for that side on any day.