On-Chain Analysis for Traders — MVRV, SOPR & Exchange Flows (Basics)

Crypto has something no other market has: a public ledger of every transaction. In stocks you can only guess what other participants are doing — in Bitcoin you can see it on-chain: at what cost basis the coins that just moved were bought, whether they're heading to exchanges or leaving them, whether the market as a whole is sitting on profit or underwater. On-chain analysis is reading those footprints.

Plenty of hype has grown up around this field, though: dashboards sold as forecasting machines, indicators that "called every top" (after the fact), and confusing data with signal. This article walks through three pillars of the toolkit — MVRV, SOPR and exchange flows — with an honest note on what these metrics measure and what they do not tell you.

What On-Chain Analysis Is

Every Bitcoin transaction leaves a permanent record: from where, to where, how much, and — crucially — at what price the coins last moved. On this foundation, analytics firms (Glassnode, Santiment, CryptoQuant; free alternatives for some of the data: Dune, The Block) build metrics that aggregate holder behavior. The three most important families:

MVRV (Market Value to Realized Value). Market cap divided by "realized" cap — i.e., every coin valued at the price of its last on-chain move. In practice: the average unrealized profit or loss of all holders. MVRV of 2.0 = the average coin is worth twice what it was at its last move. Below 1 = the market is underwater on average. The related NUPL indicator shows the same thing as a percentage of market cap.

SOPR (Spent Output Profit Ratio). While MVRV looks at paper gains, SOPR looks at realized ones: for coins actually moved on a given day, it computes the ratio of the sale price to the acquisition price. SOPR > 1 — sellers are taking profit; SOPR < 1 — they're selling at a loss (capitulation). The most popular variant is STH-SOPR, calculated only for "short-term" holders (coins younger than ~155 days) — they're the ones trading, so their behavior reacts fastest to the market. A classic observation: in an uptrend, STH-SOPR bounces off the 1.0 level like support (dips get bought around the crowd's break-even point); in a bear market, 1.0 acts as resistance.

Exchange flows. The net balance of coins moving into and out of exchange addresses. Large inflows = coins being positioned "for sale" (potential supply); a preponderance of outflows = accumulation into self-custody wallets. A fair caveat from every decent handbook: flows say nothing about intent — a transfer to an exchange could be hedging, arbitrage, or preparation to sell, and labeling exchange addresses is a heuristic, not a certainty.

Rounding out the toolkit: dormancy (are old coins moving — historically, long-term holders waking up has accompanied market tops) and the percent of entities in profit.

Where do you get this data without a hedge-fund budget? Basic MVRV, SOPR and exchange-balance charts are available in the free tiers of commercial platforms and in open community tools (Dune dashboards, public breakdowns from The Block). Paid plans mainly buy resolution (hourly data instead of daily), deeper history and derived metrics. For the weekly regime-read workflow described below, the free tier is entirely sufficient — the edge isn't in access, it's in interpretive discipline.

What the Numbers Say

The most-cited argument "for" comes from Glassnode itself. The firm's data-science team ran its metrics through a machine-learning model (long-only on BTC) and studied which ones had the highest predictive power. Result: percent of entities in profit and STH-SOPR came out as the best single signals for taking long positions on Bitcoin — the model was hunting for a "Goldilocks zone" where the market is neither overheated nor capitulated, and the out-of-sample results were described as consistently positive.

A mandatory caveat before anyone treats this as a recipe: first, this is a study by a data vendor about its own data — Glassnode sells access to these metrics, so the conflict of interest is baked into the publication. Second, the firm explicitly states it discloses only the base metrics, without the transformations and parameters used in the model — so simply looking at raw STH-SOPR will not reproduce the reported results. Third, the detailed performance curves are only available to enterprise clients, so independent verification is limited. The honest takeaway: this is evidence that signal sometimes lives in this data — not proof that you'll be the one to extract it.

What's known more robustly, from a simple look at BTC's history: MVRV zones below 1 coincided with bear-market bottoms (2015, 2018–19, 2022), and extremely high readings coincided with top zones. That's useful as a regime map, with two caveats firmly in mind: the sample is barely a handful of cycles (N around 3–4, statistically very small), and the thresholds drift lower between cycles as the market matures.

How It Works Step by Step (a Numeric Example)

An illustrative walk-through of the process — on-chain context as a filter, not a standalone system:

1. Set the regime (once a week). You check MVRV: 1.8 — the market has moderate paper profit, neither euphoria nor capitulation. NUPL is in the "optimism" zone. Conclusion: mid-cycle, neither a generational buy signal nor a top alarm.
2. Check short-term-holder behavior. BTC's price drops 12% during the week. STH-SOPR falls to 0.97 and climbs back above 1.0 within two days — short-term holders sold in a mild panic just below break-even, and the market absorbed it. Historically, this kind of "test and defend" of the crowd's break-even point looked like a correction within a trend, not the start of a bear market.
3. Cross-check against exchange supply. Exchange inflows during the drop: no anomalies, exchange balances keep drifting down. No sign of coins being mass-positioned for sale.
4. Only now, a trading decision. On-chain didn't give an entry signal — it gave context: "this drop looks like a correction, not distribution." The entry itself and its level still come from your own toolkit (market structure, momentum, risk management), and position size from position-sizing rules.
5. The reverse scenario, for symmetry. MVRV 3.2, dormancy rising (coins from the bear-market bottom are waking up), exchange inflows spiking, STH-SOPR treating 1.0 as resistance for the first time in months — the context says "late cycle, distribution." That's not a short signal; it's an argument for trimming exposure and tightening stops.

One workflow tip that ties this all together: log your readings and your own interpretation before the market resolves, ideally on a fixed weekly cadence. On-chain data is unusually easy to read as "prophetic" after the fact — every correction has some signal in the data that could have been spotted. A reading journal shows ruthlessly how much of that signal you actually saw ahead of time, versus how much your memory added afterward. It's the cheapest test of whether this kind of analysis adds anything to your process at all.

Risks — When On-Chain Analysis Does NOT Work

• Confusing data with signal. A metric describes a state, not the future. "MVRV is high" doesn't mean "it'll drop tomorrow" — it means "there's a lot of fuel for profit-taking." Markets can sit at extreme readings for quarters; anyone who shorts the reading itself pays for it with their position.
• Small cycle sample. Most "reliable" thresholds (MVRV 3.7, NUPL 0.75, etc.) are based on three or four BTC cycles. That's too little for statistical confidence, and the thresholds visibly drift between cycles. An indicator calibrated on 2017 and 2021 could have already called a top too early, or not at all, in 2025.
• Heuristics and revisions. Clustering addresses into "entities," labeling exchanges, the 155-day cutoff for STH — these are all analyst assumptions, not on-chain facts. Providers periodically revise methodology, and historical metric values can shift after the fact.
• ETFs and custody muddy the picture. With a large share of BTC supply now sitting with ETF custodians, classic flow and "percent in profit" metrics increasingly cover coins whose economic owners never touch the chain at all. Old interpretations need adjusting.
• No information about intent. An exchange inflow, an old coin moving, a whale transfer — each of these has several non-exclusive explanations (sale? hedge? wallet change? lending?). A single on-chain event means nothing; only sustained changes in the aggregates carry meaning.
• A horizon mismatched to day trading. On-chain data breathes on a rhythm of weeks and months. Trying to trade it intraday is like using a barometer to measure temperature.

The no-hype verdict: on-chain analysis is a genuine informational edge crypto has over other markets — nowhere else can you see the acquisition cost and behavior of the entire holder population. MVRV, SOPR and exchange flows honestly answer the question "what regime is the market in and what is the crowd doing." But there's a gulf between "I can see the state of the market" and "I have a trading signal" — a gulf data vendors happily paper over with marketing. Treat on-chain data like a weather forecast for a sailor: it won't tell you when to sail out for profit — it'll tell you when not to sail out at all.

FAQ

Rafał — Krypto Bez Ściemy

Trader and founder of Strefa Tradingu. He’s been breaking crypto down on YouTube for years — no hype, no signals, with a focus on market structure and risk management.

kryptobezsciemy.plInstagramYouTubeKlub Zero