STH / LTH SOPR

Two SOPR series share the linear right axis (0.6 to 1.5 typical). The sage line is STH-SOPR — the reactive cohort of coins last moved under 155 days ago. The amber line is LTH-SOPR — the structural cohort of coins held 155 days or longer. BTC price runs muted on the log left axis for orientation, and the horizontal reference at 1.000 is the breakeven line: above it, the cohort is spending coins at an average profit; below it, at a loss. The two lines almost never agree — and the gap between them is the entire point of splitting the metric.

Today STH sits at 0.997 and LTH at 0.993, the Mutual loss regime. Read together they answer a question the aggregate network SOPR cannot: who is taking the profit or loss — the cohort chasing the trend, or the cohort that sets cycle extremes.

The SOPR definition is one line:

SOPR(t) = meani ∈ spends(t)( price_spenti / price_acquiredi )

For every spent output on day t, divide the price at which it was spent by the price at which it was last moved on-chain. Average over every spend on the day. The denominator is the realized-cap contribution that disappears when the coin moves; the numerator is the spend-side value. A result of 1.000 means every coin moved at exact breakeven.

The cohort split. STH-SOPR restricts the average to outputs younger than 155 days at the moment of spend; LTH-SOPR restricts to outputs 155 days or older. The 155-day boundary is not arbitrary: Schultze-Kraft & Heeg’s 2020 cohort paper identifies it empirically as the age at which a coin’s spending probability stabilises — the point where reactive short-term behaviour gives way to structural long-term holding. The same boundary drives the STH/LTH realized price and RHODL pages.

Why split at all. A reading of STH = 1.05 with LTH = 5.0 says “new buyers barely profitable, old hands distributing heavily into the rally” — structurally the opposite of STH = 1.20 with LTH = 0.95 (“new buyers taking profit, old hands realising loss into a recovery”). The aggregate network SOPR averages those two stories into mush. Coverage starts in mid-2010; reads from the first three years carry a near-zero-cost-basis distortion documented in the methodology page.

The signal is in the crossings, not the absolute level. STH-SOPR crossing 1.0 from below after an extended sub-unit stretch has historically marked cycle restarts: the reactive cohort has stopped realising losses and the bid is back. STH-SOPR crossing 1.0 from above signals early-cycle stress or the start of distribution — cross-reference price to disambiguate a healthy dip from a regime change. LTH-SOPR moves the long-range backdrop: sustained readings above 3 mean old coins are entering the market (late-cycle distribution), while sustained sub-1 readings through a prolonged drawdown mean old-hand capitulation. Because both lines can swing several percent on a quiet day, treat a multi-week sustained crossing as signal and a single-day print as noise.

Reading the canonical cycle anchors against both SOPR series surfaces the page’s central asymmetry: STH peaks barely move, LTH peaks collapse. STH-SOPR tops at 1.29, 1.40, 1.26, 1.19, 1.11, 1.21 across the six cycle peaks. LTH-SOPR tops at 34.9, 298, 56.2, 13.5, 5.1, 5.6 over the same windows. Bottoms have held a much tighter band: STH lows of 0.85 / 0.89 / 0.84 / 0.94 and LTH lows of 0.35 / 0.27 / 0.57 / 0.35. Prices below are our own daily closes; SOPR is read from the live cohort series.

The cleanest view of the regime shift is the maximum reading inside each cycle’s topping window, in order. Pulling the peak of each line from the live series:

STH-SOPR has been remarkably stable. Six cycle peaks span 1.11 to 1.40 — a roughly 30% range. Even the lowest STH cycle peak (Nov 2021 at 1.11) sat clearly above breakeven, and the most recent print (Mar 2024 at 1.21) is squarely in the long-run norm. The reactive cohort behaves the way it always has: when prices rise, short-term holders book profit.

LTH-SOPR has compressed by an order of magnitude, and the reason is mechanical, not behavioural. The 2013 peaks printed 35 and 298 because the 155-day-old cohort in 2013 still held Patoshi-era coins last moved at fractions of a dollar — every spend at $1,000+ generated a four- or five-figure ratio. By 2017 that cohort had aged forward but still carried substantial sub-$100 cost-basis coins, producing the 56 peak. By 2021–2024 the long-term cohort had matured to a point where most LTH spends have cost bases in the thousands, so the peak ratios now sit in the single digits.

Lost coins compound the bias. Every Patoshi-era coin that has not moved since 2010 sits permanently in the cost-basis baseline at near-zero, but those coins almost never spend. Chainalysis research has estimated 2.78 to 3.79 million BTC permanently lost; Sergio Demian Lerner’s Patoshi research identifies roughly 1.1 million Satoshi-era coins untouched since 2010 (recap on Yahoo Finance). They add realized-cap mass without ever generating SOPR-affecting spends, so as their share of the dormant supply grew, the LTH cohort’s effective average cost basis drifted forward toward more recent prices.

The practical corollary: read LTH-SOPR as a cohort-relative threshold, not an absolute one. “LTH-SOPR > 3 = top” has held up better than most realized-cap-family ceilings precisely because it scales with the cohort. Both 2021 peaks and the 2024 high cleared 3 — only the absolute heights have shrunk.

1. Custodial reshuffles register as “spends.” SOPR counts every output spend equally, whether or not beneficial ownership changed. ETF creation and redemption, exchange cold-to-hot wallet transfers, and OTC settlement all count as spends and feed the daily average. This custody noise floor has grown meaningfully since US spot Bitcoin ETFs launched in January 2024 — a structural, ongoing distortion, not a one-off.

2. Large custodial wallet migrations are known artefacts. When an exchange or custodian reshuffles its reserves in bulk, long-dormant coins last moved at fractions of a dollar can suddenly appear as “spends” at current prices, producing a multi-day spike that is custody mechanics, not market behaviour. Watch for unexplained single- or multi-day jumps that coincide with reported large internal transfers, and treat those prints as distorted.

3. The early-chain LTH line is barely market information. The pre-2013 LTH readings in the hundreds — including the 18 July 2010 print near 121 — come from a tiny supply where the 155-day cohort’s cost basis was effectively zero. A “298” in 2013 and a “5.6” in 2024 are not comparable observations of the same phenomenon; the cohort’s cost-basis floor moved underneath the ratio. Any cross-cycle comparison that includes the first three years is reading an accounting artefact as a signal.

4. The 155-day boundary is empirical, not mechanical, and the daily line is noisy. Coins either side of 155 days behave on a continuum, not a step; a 120- or 180-day cut would tell a similar story with different numerics. On top of that SOPR is one of the noisier on-chain series — daily values swing several percent on no real news, especially in low-volume regimes. Most analysts apply a 7-day moving average for reading; the raw daily series is preserved here for transparency. Sustained multi-week crossings are the reliable unit, not single prints.

If you accumulate on a schedule, the Mutual-loss regime (STH < 1 and LTH < 1) is one of the highest-conviction tactical accumulation signals on the site. It has fired at every cycle bottom on record, and the bottom-side compression has been mild — even the shallowest, the 2022 post-FTX trough at STH 0.94 / LTH 0.35, fired cleanly. STH-SOPR crossing 1.0 from below after extended weakness has historically marked the restart.

If you are timing the cycle, lean on the spread rather than the level. LTH-SOPR > 3 has stayed a reliable distribution flag even as the absolute peaks compressed, but do not expect the next top to reach 2013’s or 2017’s heights — the cohort that produced them has aged out. Pair SOPR with STH/LTH realized price for the cost-basis lens and MVRV for the unrealised-profit picture; SOPR is one of the faster signals here, moving on the order of days rather than weeks.