Bitcoin at a glance
Every reading the site publishes, placed in its own history, with what is firing today and what the evidence adds up to. Ten seconds for the dashboard; the questions beneath it are the same numbers with the reasoning attached. Nothing here is a recommendation.
Anything that is price against something else — the fall from the high, the multiple of what holders paid, the cycle multiple, the gold ratio — follows the live price and updates as you watch. The valuation state, crowding, ETF holdings and the bubble model are computed once a day by the pipeline and are marked with the day they describe.
Live price
Daily readings through Oct 14, 2026 — the last day with a complete close — published —
all sources refreshed
Bitcoin
$83,094
+8.2% over 30 days · 15.2% above its 200-day average
Where it sits in its 90-day range
The site’s reading
Neither cheap nor dear
Our valuation composite reads 42 of 100, its “average” band. It has fired too few times to know whether that band has ever meant anything.
Valuation composite42of 100, low is cheap
cheapdear
MVRV, the Mayer multiple and the power law, each as a percentile of its own history and averaged equally. It is a description of where price sits, not a target. Power law-0.23 dex26th percentile of its own history
cheapdear
Distance from the long-run trend, in decades of price: -0.23 dex is about 41% below the trend line. The percentile says how that compares with every other day. Like the Metcalfe fit, the answer moves with where the fit begins; the page shows it from several start dates. Metcalfe premium+6.4%56th percentile of its own history
cheapdear
Price against what network size alone would justify, on the fit from 2017, which has the lowest out-of-sample error. Fitted from 2011, as published elsewhere, the same model says -40.1%, valuing the network at $138,792 rather than $78,094; that fit fails the out-of-sample test. Since 2017 a model of price on time alone predicts about as well. MVRV1.54×37th percentile of its own history
cheapdear
Price divided by what the average holder paid, so 1.50 means the average coin is up 50%. Below 1 it is under water. Crowding87thpercentile of its own history, high is crowded
cheapdear
Funding, open interest and implied volatility combined, expressed only as where today sits in its own history since 2017. Fear and Greed62of 100 · 67th percentile of its history
cheapdear
Alternative.me’s 0–100 sentiment index. Both extremes were tested as rules and neither beat its baseline. This cycle1.30×since halving day
cheapdear
Price against the price on halving day. At the same day of past cycles: 2012 19.02×, 2016 5.82×, 2020 2.31× Since the halving907days
cheapdear
The halving was 20 April 2024. Four cycles is too few to call this a pattern.
Firing today 3 of 36
Of the rules this site scores. A rule firing is a reading, not a recommendation — the record beside each says whether it has ever meant anything.
What the evidence says
36rules scored
2fired often enough to judge
0beat doing nothing
At 36 rules, about 3.6 would clear a 90% interval by chance if none of them worked. None has. The median rule has fired 5 times since 2013.
Every rule this site scores 36 tested, 2 judgeable
enough history to judge
too few times to judge
one of ours, same test
firing today
One cell per rule, sized the same whatever it claims, shaded by how much is known about it — never by whether it points up or down. 2 of 36 have fired often enough for the evidence to say anything at all. The rest are not wrong; there is simply not enough history to tell them from chance. A tally that counts them as votes is counting 34 things it cannot measure.
For scale: on any day since 2013, a 40% fall followed within a year 35.1% of the time and a doubling 52.6%. Every figure above is placed against those base rates, not against zero.
What does one bitcoin cost?
$83,09433% below its highest ever close of $124,777
today
$61,641 five years ago$124,777 high
How far does it fall?
-33%from its high, so far in this cycle
This is the first thing worth knowing, because it decides what everything below is worth to you. Every cycle on record has contained a fall of at least three quarters.
2011–12-93%
2013–15-85%
2017–18-83%
2021–22-77%
this cycle-33%
If a fall of half would force you to sell, that is a fact about your position rather than about the market, and no reading on this page changes it.
Is it cheap or expensive right now?
Averagecheaper than 58% of all days since 2013
42
very cheapaveragevery expensive
Three ways of valuing bitcoin — against what holders paid, against its long-run trend in time, and against the size of its network — each ranked against every day since 2013, then averaged. What has followed from each state.
What did the people holding it pay?
$53,855the average coin last changed hands here — price is 1.54× that
Every coin valued at the price it last moved on the blockchain. Above one, the network as a whole is in profit. Past cycle tops came at three to four times. Realised value.
Where is it in the four-year cycle?
1.3×higher than on halving day, 907 days ago
Bitcoin’s new supply halves roughly every four years. Measuring each cycle from its halving day puts them on the same footing, and each has multiplied less than the one before.
2012 cycle19.0×
2016 cycle5.8×
2020 cycle2.3×
this cycle1.3×
All measured at the same day count. Four cycles is the entire history, far too few to call it a rule. Cycle monitor.
Is the market crowded?
Crowded87th percentile of its own history — traders pay 5.3% a year to be long
87
washed outcrowded and levered
How much borrowed money is in the market and how eager it is. Crowded today. Insurance against a fall is priced unusually cheaply, which says what protection costs rather than what happens next. Flows and positioning.
How much do the ETFs hold?
889,769bitcoin — about 4.4% of every coin that exists, from 5 issuers
Counted only from issuers publishing a daily figure a machine can read: five of them. Seven more report to the regulator each quarter. This is not the total held by all US funds, and the flows page names the ones missing. Coverage.
What is it worth in gold?
19.2 oz48% below its December 2024 peak of 37.1 ounces (monthly average)
Gold beat bitcoin in 2025 and is ahead so far in 2026; bitcoin had never trailed gold in two consecutive calendar years. Rules for switching between the two are tested on the relative value page, and most of them lose.
Is this a bubble?
Nothe bubble model finds no accelerating pattern on any of its eight windows
There is a model built to detect the runaway acceleration that preceded some past crashes. It read zero before the October 2025 top, which is why the autopsy page publishes its whole record and not just today’s reading.
Why believe any of this?
You can check it
Every figure comes from a public daily snapshot of free sources, computed in your browser, with the workings on the page it came from. Every mistake found so far is listed with its date and what changed. Methods and corrections.
What the tests say
The part most sites leave out. Each line is a result from testing the claim against its own history, not an opinion about it.
Do the popular chart signals work?Relative strength, moving-average crossings, MACD and the rest, each compared with buying on a randomly chosen day.30 signals tested, 2 have fired often enough to judge, none beats doing nothingBitcoin rose after most days in this period, so “price went up after my signal” proves nothing on its own. Each rule is measured against what an ordinary day did instead.
Does valuation warn before a crash?The ten largest declines since 2013, chosen by rule, with every indicator replayed as it read the day before.It flagged 4 of 5 cycle tops, and none of the 4 sudden crashesThe slow tops came after price got expensive, so valuation saw them. The sudden crashes started from ordinary valuations, because a cascade of forced selling is not the same thing as being overpriced.
Does money printing move bitcoin?Central bank rates, money supply, net liquidity, credit spreads and equity volatility, tested at every lead from zero to twenty-six weeks.No series beat what random chance produces on data like this
Does the bubble model predict crashes?The 561 days it has fired since 2013, against every other eligible day.A big fall followed 28.0% of them, and 35.0% of all daysFewer, not more. On this record the model tends to fire during the accelerating part of a rise, and those rises mostly kept going.
Does switching between bitcoin and gold beat holding?Momentum and mean-reversion rules at three lookback lengths each.Rarely, and changing the lookback flips the answer
Do the signals work in other markets?Six of the rules run across bitcoin, ether, Solana, the Nasdaq and the S&P 500, each against its own base rate.Overbought RSI predicts nothing anywhere; one rule, the golden cross, beats the baseline thinlyDoes it work anywhere?
What happens with no signal at all?Every day since 2013 as a starting point: how often a 40% fall or a doubling followed within a year.A 40% fall on 35.1% of days; a doubling on 52.6%Base rates — the denominator under every test above
Can I test my own rule?Ten indicators, four operators, two outcomes, scored through the same engine as every result on this page — with a count of how many rules you have tried.Test your own rule
Where is the money, and what will it be worth?Realised cap, who holds the coins from filings and addresses, and a surface separating how much arrives from where the cycle is when you look.The model that turns inflows into a price forecast overshot at 18 of the last 18 originsCapital flows — no number, on purpose
Does network size explain price better than time alone?The Metcalfe network model against a plain trend in time, refitted every year.Better in 5 of 7 years, which is not clearly better
What we can and cannot say yet
- Adding
- Moderate. Valuation reads average, a state whose forward record carries no reliable direction; price is in a drawdown, and drawdown days were followed by a median +112% at one year against +14% for advances — on a label that is late and unstable; a hash-ribbon recovery fired 49 days ago; across twenty completed episodes the median year after was +105%, 85% positive, with three named losers.
- Reducing
- None. Valuation reads average, not the very-expensive state, which is the only one with evidence of poor forward outcomes; options skew at +2.77 points shows no fear premium; the LPPL bubble model finds no qualifying fit: the advance is not super-exponential; no tested sell rule on the technical-signals page shows an edge, and the textbook “overbought” reading ran the opposite way.
- Hedging
- None. Skew at 2.77 points, neither cheap nor expensive.
- Rotating
- None. Rotation rules tested on the relative-value page win at one lookback and lose at the others; mean reversion mostly loses, and its few wins depend on the lookback and the start date. A rule that depends on its parameter is not evidence.
No tested evidence currently establishes an advantage in adding, reducing or rotating. Which of these applies to you depends on your horizon, how much you hold, tax, and whether a fall of half would force you to sell.