Bitcoin Indicator Autopsy
Pi Cycle Top, stock-to-flow, the 2-year moving-average multiplier and the 200-week moving average, each with its complete historical trigger record and what followed.

This page computes its figures in the browser from the daily snapshot, so it needs JavaScript to show them. The underlying data is public and needs no JavaScript to read: akpasz.github.io/btc-data/data/, documented at github.com/akpasz/btc-data. The methods are readable without scripts at /tools/methods.

Research · Validation

Bitcoin Indicator Autopsy

Popular Bitcoin signals, tested. Pi Cycle Top, Stock-to-Flow, the 2-Year Moving Average Multiplier, the 200-Week Moving Average and LPPLS are evaluated using their stated rules. The page records every historical trigger, what followed, the sample size and the current condition. No signal is endorsed; each is examined against its own historical record.

Connecting

Price history loading

Loading price history

Reading the daily snapshot.

Pi Cycle Top

–

LPPL qualified-fit confidence

–

Stock-to-Flow model

–

2-Year MA Multiplier

–

200-Week MA

–

Price and the indicators

Log scale. The moving averages every signal below is built from, drawn on the same history the verdict tables use.

Pi Cycle Top

111-day MA crossing above 2× the 350-day MA (Swift, 2019). A rule intended to identify cycle tops.

Trigger	Price	90d after	365d after	Worst drawdown within 1y

Stock-to-Flow

PlanB (2019): value from scarcity, price ∝ S2F³. Tested against the published specification and a live refit.

Date	Model value (0.4 × S2F³)	Actual price	Actual / model

2-Year MA Multiplier

Buy below the 2-year MA, sell above 5× it (Bitcoin Investor Tool). Both bands tested.

Event	Date	Price	365d after

200-Week Moving Average

A commonly used long-term cycle reference. Every close below it, and what followed.

Crossed below	Price	365d after

Verdicts

One line each, with the sample size that qualifies it.

Signal	Triggers	What the record shows	Current condition

What happened when the rule fired?

A comparison, not a test: the companion to the event study below, run the other way round: each rule is triggered using only information available on the day, and the median return that followed is placed beside the median return after a randomly chosen day over the same history, exactly as on the Technical Signals page. The difference is a historical difference, not an edge. Intervals need at least twenty triggers; none of these rules has that many, so none gets one, and the page says so rather than inferring from a handful.

Rule, as triggered	Triggers	90 days: after the rule	after an eligible day that did not fire	difference	1 year: after the rule	after an eligible day that did not fire	difference	Interval

What the indicators showed before major declines

Events are identified retrospectively from what price did next, by rule rather than from memory: any day from whose daily-average price fell 30% or more within thirty days, and any all-time-high close from which price fell 40% within six months, clustered so one decline counts once. The indicator readings in each row use only information available on that day; the choice of rows does not. This is a description of what was visible before known declines, not a predictive test. The columns are not independent tests: most are transformations of price or of closely related state variables. Today is the last row.

Date	Turned out to be	Price	Valuation	MVRV pct	Mayer	RSI	Pi Cycle	2-yr MA	Fear & Greed	DVOL	LPPL	Then: 30d low (selection for shocks)	180d low (selection for tops)

What this table cannot show yet: leverage. The sharp declines that were not preceded by an expensive valuation may have been visible in leverage rather than in price, but the series that describe leverage — funding, open interest, options skew, CME positioning — have only been recorded on this site since late 2025. They will appear here as columns once enough history exists to compare an event day against its own past. Until then the honest statement is that the family that might have shown fragility before those events is the one with no record of them. Prices are daily averages, so an intraday crash that recovered by the close — 10 October 2025 among them — registers here through the decline that followed it, not on the day itself.

LPPLS: testing for bubble-like price dynamics

The log-periodic power law singularity model (Johansen, Ledoit and Sornette) describes an endogenous bubble: super-exponential growth with accelerating oscillations, ending at a critical time. It is the one model in this family built for cycle tops and it makes no claim on shocks. Primary specification: the published Gerlach, Demos and Sornette (2019) qualification filters, fixed before any evaluation. Sensitivity: a stricter set, reported beside it and never used to select a result. Qualified-fit confidence, house implementation: the share of eight window lengths (120 to 730 observations) whose best fit passes the filters, computed daily from 2013, as of each day on that day's data only. Eight windows is a coarser ensemble than the published multiscale indicator, which scans many nested windows, so this is a house implementation of the published construction rather than a reproduction of it. Confidence is a robustness measure across windows, not a probability of a crash; neither filter set uses a relative-error criterion. Readings are computed through the last completed daily-average observation, so the series carries a one-day lag by design. The series, both filter sets and the random-day baseline are published in lppls.json.

Episode start, first qualifying reading (0.5 or above, 90-day grouping)	Price	Strict set	Estimated critical time	Then: 1-year low	1-year high	Outcome

Lead time, and what happened after the first signal

For each run of days at or above 0.5: when it first fired, how long the price kept rising afterwards, how much it rose before its eventual high within the following year, and what followed that high. The model marks a condition, not a moment; the table shows how long the condition lasted and what it would have cost to treat it as a sell signal on day one. Runs that fired after a top — the early-2014 rows — are not display errors: the model qualified the tail of an advance that had already ended, which is part of its record.

First fired	Price	Run length	Highest daily-average price within 1y	Days to it	Rise to it	Fall from that high within 1y

Method

Everything on this page is computed in your browser from the daily price and supply history in the public snapshot (Blockchain.com feed, 2010 to today). Moving averages are simple; the Pi Cycle crossing requires the previous day on the other side and crossings within 90 days of one another are counted once; stock-to-flow is circulating supply over the trailing 365-day change in supply; forward returns are simple price changes and the worst-drawdown column is the minimum close within a year of the trigger. Nothing is optimised, smoothed or excluded, and the current values of every line are printed in the strip above.

The stock-to-flow test uses PlanB's published approximation (value = 0.4 × S2F³) and, separately, a full-sample refit of ln price on ln S2F; the notes report both, because a refit quietly absorbs part of a model's failure and the honest test is the specification as published.

What this page tells you, and what it does not

Claimed: these are the exact historical records of each rule on this data, reproducible from the snapshot. Not claimed: that any signal works or fails going forward. The sample sizes are three to seven events per rule; that is not statistics, it is a handful of anecdotes with dates, and the verdict column says so. Signals built on halving-cycle structure inherit the diminishing-cycles problem visible in the 2-Year MA sell band, which has not been reached since 2017.

Sources: the Crypto Exponentials daily snapshot (Blockchain.com price and supply). Signal origins: Pi Cycle Top, Philip Swift, 2019; stock-to-flow, PlanB, "Modeling Bitcoin Value with Scarcity," 2019 (and the co-integration critiques of 2020–22); 2-Year MA Multiplier and 200-Week MA heatmap, Philip Swift / LookIntoBitcoin. Nothing here is investment advice.

Loading price history

Data could not be loaded

Price and the indicators

Pi Cycle Top

Stock-to-Flow

2-Year MA Multiplier

200-Week Moving Average

Verdicts

What happened when the rule fired?

What the indicators showed before major declines

LPPLS: testing for bubble-like price dynamics

Lead time, and what happened after the first signal

Method

What this page tells you, and what it does not