The Macro Consensus
Ask a thoughtful macro investor what drives bitcoin and the answer arrives within a sentence: global liquidity. Michael Howell at CrossBorder Capital has built a decade of research around the claim that bitcoin tracks a broad measure of central-bank balance sheets and shadow-banking credit with a lead of roughly thirteen weeks. Lyn Alden's framework points in the same direction through different machinery: bitcoin is a long-duration, cash-flow-less asset, exquisitely sensitive to the discount rate, and therefore exquisitely sensitive to the supply of dollars. The chart that gets shared most often overlays bitcoin against global M2 and shows the two lines drifting together for years.
The case is intuitive. Bitcoin has no earnings to anchor its valuation, so its price has to come from somewhere. The somewhere is flows. Flows come from liquidity. The chain is short, the logic clean, and the visual evidence persuasive.
The SHAP Verdict
The dashboard's machine-learning pipeline rated US M2 the tenth most important feature out of thirteen for predicting bitcoin's seven- and thirty-day direction. Walk-forward validation across four cycles produced the same low rank in every round. The dollar index, the other macro variable on the dashboard, did not earn a top SHAP rank either. It sits in the indicator set because it cleared the independence test that filters for features uncorrelated with the rest, not because the model rated it highly.
This is an awkward result to publish given how often the rest of this site cites the same liquidity literature. It is more useful to ask why the rank is so low than to pretend it is not.
What Counts as Liquidity
Liquidity is not one number. The US M2 feed on the dashboard is the narrowest defensible version: deposits, money-market balances, and small time deposits at US institutions. Global M2 adds the eurozone, Japan, China, and the United Kingdom, translated into dollars. Net Fed liquidity (Fed balance sheet minus the Treasury General Account minus reverse repo balances) measures the dollars circulating in the financial system rather than parked at the central bank. Howell's CrossBorder index goes wider still, layering in shadow-banking credit and central-bank reserve assets to produce a measure of roughly $170 trillion.
The dashboard's M2 reading is therefore a proxy for a proxy: it captures the US slice of one of four reasonable definitions. The honest reading of the SHAP rank has to accommodate that. A wider liquidity proxy might rank higher. The narrow one ranked low.
Why the Tide Is Already in the Price
But the broader story does not depend on which proxy is chosen. Even if global liquidity ranked five or six rather than ten, it would still rank below the price-derived features at the top of the list. The reason is structural: liquidity's effects on bitcoin are already absorbed by the features that describe what bitcoin itself is doing.
Power Law Position measures bitcoin's deviation from its long-run trajectory. That trajectory was carved by fifteen years of liquidity cycles. When liquidity expanded during 2020-2021, the deviation widened; when liquidity contracted in 2022, it narrowed. The Power Law reading is, in effect, a low-pass filter on liquidity history rendered in bitcoin's own units.
MVRV captures the gap between bitcoin's market value and the aggregate cost basis of coins held on-chain. The gap widens when buyers chase price above their predecessors' entry. That chase is funded by liquidity. By the time MVRV reads hot, the liquidity that produced it has already flowed.
BTC Fees measure demand for blockspace, which spikes when capital is moving onto and around the network. Network congestion is the on-chain receipt for liquidity-driven flows. The model saw the receipt before it saw the macro aggregate.
Three of the dashboard's top features therefore encode liquidity's consequences with higher fidelity than US M2 encodes its cause. The model preferred the receipts.
The Dollar Index's Place
DXY is on the dashboard on different grounds. It did not earn a top SHAP rank but cleared the independence test, and it captures a dimension the other twelve features do not: the relative price of dollars against the world's other reserve currencies. When dollars are scarce globally (DXY rising), risk assets including bitcoin tend to bleed regardless of whether US M2 is expanding. The two macro features are conceptually complementary. Together they earn 9.3% of the composite: modest but deliberate.
What This Means for the Composite
The macro cluster is light because the price-derived cluster has already done liquidity's accounting. A heavier macro weight would not add information; it would double-count signal already pooled at the top. The pattern repeats one layer above the tech-beta story: the upstream driver matters, but the model reads it through downstream proxies that contain more pricing detail.
This is why naive macro overlays ("M2 is up, accumulate") tend to underperform when tested against bitcoin's own history. The smart capital that responds to liquidity has already moved by the time the macro aggregate prints, and the result of that movement is sitting in the Power Law band and the MVRV reading before the M2 release goes out.
What Would Change This
Two scenarios would raise liquidity's rank. The first is a broader liquidity proxy. Replacing US M2 with a true global aggregate (the G4 sum, or net Fed liquidity, or Howell's index when the data becomes available) might surface signal the narrow version misses. Whether it would do so meaningfully is an empirical question, not a rhetorical one. A future iteration of the model will test it.
The second is a regime where Power Law stops absorbing the tide. If bitcoin escapes its long-run trajectory in either direction, a permanent break above the upper band or a structural collapse below the lower one, the price-derived features lose their anchor and the macro inputs would have to do more work. That regime has not arrived. When it does, the SHAP ranks will move.
The Receipt and the Cause
Liquidity is the cause. The dashboard's top features are the receipt. The composite reads the receipt because that is what shows up first, with the most detail, in the data the model was trained on. The macro cluster sits at 9.3% not as a denial of the liquidity thesis but as a quiet vote of confidence in it: the cause has already been counted, only under a different name.