Bitcoin Market Regime Classification

01 What "Regime" Means in Market Context

Bitcoin market regime classification is the process of identifying which macro environment is currently shaping price behavior. Not what price did today, but whether the conditions that drive price over weeks and months are expansionary, contractionary, or ambiguous.

Most classification systems reduce this to three states. The simplification is intentional. A regime label is not meant to capture every nuance of the market. It is meant to give one clear input about which environment the market is in.

Risk-on conditions exist when liquidity is expanding, yields are falling or stable, hash rate is growing, and cross-asset correlation is positive. Capital moves into assets seeking financial return.

Risk-off conditions exist when liquidity is contracting, yields are rising, volatility is elevated, and defensive assets are outperforming. Capital rotates out of risk assets into cash, treasuries, or hedges.

Transition states are the most common and the hardest to classify. Indicators conflict. Some improve while others deteriorate.

The three-state model is a simplification that works well for Bitcoin because Bitcoin's price behavior is heavily influenced by macro liquidity conditions. But regimes do not have to be uniform across asset classes. Equities can be risk-on while crypto is transitional. The dollar can strengthen (risk-off for emerging markets) while US tech stocks rally (risk-on for Nasdaq). The label applies to the asset class being analyzed, not to "the market" as a whole.

For Bitcoin specifically, the dominant regime driver since 2020 has been the direction and magnitude of US dollar liquidity. When the Fed expanded its balance sheet and fiscal stimulus flooded the system, Bitcoin entered a sustained risk-on regime. When quantitative tightening began and rates rose aggressively, the regime flipped. The three-state model captures these shifts without overcomplicating the signal.

02 Why Single Indicators Fail

Every indicator has a failure mode. The yield curve inverted in mid-2022 and stayed inverted for over two years, a period in which Bitcoin rose from $16K to over $70K.

The hash ribbon recovery signal fired in January 2020, about two months before the COVID crash sent Bitcoin from $10K to $3.8K. Net liquidity was expanding through mid-2024 while Bitcoin consolidated sideways for months. The RRP drainage that ran through the late-2023 rally was a one-time structural event that stopped moving net liquidity once the facility emptied.

Single Indicator View

Yield curve: inverted.

Hash ribbon: recovering.

Net liquidity: rising.

VXX (volatility): elevated.

Result: two readings point one way and two the other. No single reading settles it.

Multi-Factor Classification

Macro: mixed. Liquidity improving but yields still restrictive.

On-chain: positive. Mining recovery confirmed, network growing.

Cross-asset: cautious. Volatility elevated, dollar weakening.

Result: Transition, leaning risk-on.

The failure is not that any single indicator is wrong. Each one captures real information about one dimension of the market. The failure is treating any one dimension as sufficient for a decision that depends on all of them.

03 Historical Regime Transitions

The regime framework is easiest to see in the period since 2020 because the macro drivers were unusually pronounced. The Fed's balance sheet expanded by over $4 trillion, then contracted. Fiscal stimulus flooded and then drained. These moves were large enough that the regime transitions were relatively clear in hindsight, which makes them useful for illustrating the concept. Earlier periods (2017-2019, 2013-2015) followed similar patterns at smaller scale but with less clean macro data available.

Mar 2020 - Nov 2021

QE, fiscal stimulus, near-zero rates. Every macro component pointed the same direction. Net liquidity surged. BTC ~$4K to $69K.

Nov 2021 - Jun 2022

QE taper, aggressive rate hikes, RRP above $2T. Net liquidity contracted. The Luna/UST collapse accelerated the drawdown. BTC $69K to $18K.

Jul 2022 - Sep 2023

QT underway but RRP above $2T locked liquidity out of the system. Hash rate recovering. Yield curve deeply inverted. Conflicting signals across every category. BTC rangebound $16K-$30K.

Oct 2023 - Mar 2024

RRP draining rapidly, releasing liquidity. ETF anticipation provided a catalyst. BTC $27K to $73K.

Apr 2024 - Present

QT slowed, then ended in December 2025. RRP near zero, so no more drainage. TGA volatile around debt ceiling dynamics. Equities strong while crypto choppy. BTC $54K-$126K with high volatility, a record in October 2025 and a fall back below $60K in 2026.

04 Regimes Do Not Switch Cleanly

The timeline above makes regime transitions look discrete. In practice, the shift from risk-on to risk-off in late 2021 was a three-month process. The Fed signaled tapering in September. Markets ignored it through October. By November the signal had compounded with rising RRP and declining breadth, and price was falling with the macro data. There was no single day the regime "switched."

This is a fundamental modeling problem. Most financial literature assumes Gaussian distributions for returns: normal curves, standard deviations, geometric Brownian motion. Reality in crypto is messier. Returns are fat-tailed, especially at higher frequencies. Kurtosis runs far above what a normal distribution predicts, which means the extreme moves that define regime transitions happen more often and more violently than the math suggests. Hamilton's 1989 regime-switching model formalized the idea that the same data point means different things depending on which state generated it. A 5% drawdown in an expansionary regime is a different event from a 5% drawdown in a contractionary one.

Smooth Regime Probability (illustrative)

Instead of a binary switch, the system tracks probability across all three states. The directional lean (here: toward risk-on) and its rate of change are what the classification reports.

Modern implementations treat transitions as smooth rather than discrete. The system tracks a probability distribution across all three regimes simultaneously, and the directional lean evolves tick by tick. In clearly risk-on or risk-off environments, the classification is easy. During transitions, the ability to quantify how that lean is shifting is what separates useful classification from noise.

05 What Automated Detection Looks At

Regime classification combines three categories of inputs. Each operates on a different timescale and captures a different dimension of the market. The categories are not weighted equally. Macro inputs move slowly but carry the most weight because they set the backdrop that everything else trades against. On-chain inputs are faster and specific to Bitcoin. Cross-asset inputs are the fastest and capture the real-time flow of capital between risk and safety.

Macro

Slowest. Heaviest weight.

Net Liquidity direction

Yield Curve state

Fed policy posture

RRP trend

On-Chain

Faster. Bitcoin-specific.

Hash Rate trend

Miner capitulation

Mining economics

Accumulation patterns

Cross-Asset

Fastest. Risk appetite.

VXX (equity volatility)

UUP (dollar strength)

SPY / QQQ breadth

Correlation direction

The interaction between categories matters more than any single input. Net liquidity can be rising while the yield curve is inverted and the dollar is strengthening. That combination tells a different story than net liquidity rising with a normal yield curve and a weakening dollar. The regime is determined by the pattern across categories, not by any one reading.

The output is not a trading signal. It does not say "buy" or "sell." It says which environment the market is in, and other systems decide what to do with that context. This separation between context and action is important both for system design and for regulatory clarity.

06 Where PreReason Fits

PreReason's Cross-Asset Regime briefing delivers a regime classification as part of a structured context payload. Agents receive the regime label, directional lean, confidence scoring, and the individual metrics that produced the classification. The evidence page documents the measured results.

The difference is what the LLM spends its reasoning on. With raw JSON, the agent first computes trends, then interprets significance, then determines the regime, then decides what to do. With a pre-analyzed briefing, the computation and interpretation are already done. The agent goes straight to the decision.

Raw JSON (agent must interpret)

{
"net_liquidity": 5983000,
"change_7d": 0.012,
"yield_10y": 4.31,
"yield_2y": 4.28,
"vxx": 18.4,
"uup": 27.1,
"hash_rate": 842.3,
"hash_rate_30d_sma": 831.7,
"hash_rate_60d_sma": 819.2
}
// Agent must: compute spread,
// determine trend directions,
// classify regime from scratch,
// THEN make a trading decision

The raw rows behind cross.regime measured 23,250 tokens (September 2026). No interpretation. All reasoning on the LLM.

PreReason Briefing (decision-ready)

An excerpt of get_context cross.regime (MCP), as served on October 12, 2026
...
## Market Regime
**Current Regime:** EXPANSION
Liquidity rising (+1.3% 7d, outside the regime vote's ±0.5% band). Dollar (UUP) flat (+0.5% 7d, inside the regime vote's ±0.8% band).
| Indicator | Direction | Strength |
|---|---|---|
| Liquidity | rising | moderate |
| Dollar (UUP) | flat | strong |
| BTC/SPY Correlation | 0.31 | low_positive |
| Hash Ribbon | 1.02 | above_1 |
| 200D MA Distance | 16.2% | more than 15% above |
| Price-to-Production | 1.00x | marginal |
*Direction here is the regime vote's reading, on its own bands; a metric row above reads direction on that metric's own band.*
**Regime Score:** 5 | **Confidence:** 65/100
...
*Sources: Coinbase | Federal Reserve Board | U.S. Treasury | Finnhub | Federal Reserve Bank of New York, Markets API. The Effective Federal Funds Rate (EFFR) is subject to the Terms of Use posted at newyorkfed.org. The New York Fed is not responsible for publication of the EFFR by PreReason, does not sanction or endorse any particular republication, and has no liability for your use. PreReason is not affiliated with the New York Fed. | ECB | BIS | FRED, Bank of Japan data | Powered by CoinGecko | PreReason (calculated)*
*cross.regime | pro | Terms: /api/context/glossary*

The full cross.regime briefing measured 4,618 tokens (September 2026). Regime resolved. Agent focuses on the decision.

This is still evolving. Smooth transition modeling, per-asset-class regime divergence, and higher-frequency detection are areas of active development. The current implementation uses a fixed set of indicators and weighting logic. The concept described in this article is the framework. The implementation improves as more data and more backtesting refine what works.

For builders working on their own classification, the raw data to context guide covers how to structure regime data for LLM consumption. The API documentation covers access through MCP or REST.

Related Insights