What the series shows, and what it missed

The BlockPhi Global Liquidity Model is read here against five dated market periods, misses included, and one direction test on the current series. Each period from 2020 onward is read from the series as it stands today, with Bitcoin weekly closes as dated context only. The test compares six-week directions from January 2020 to 24 July 2026. This is a model estimate as of 24 July 2026, and past results do not indicate future results.

Updated weekly · Data as of

What the published study found

Three results from our published study on Global Liquidity and Bitcoin are shown here. These results come from our published research on 2022 to 2026 data, which used an earlier composite measure of Global Liquidity. The test on the current series is shown separately. The study used weekly data from 18 February 2022 to 24 April 2026, before the live record of the current series starts on 31 July 2026. All three results are model estimates. Past results do not indicate future results.

The lead. The study compared weekly changes in Global Liquidity with Bitcoin’s weekly returns at every lag from 1 to 20 weeks. The cross-correlation peaked at 11 weeks: Global Liquidity led Bitcoin by about 11 weeks. The study read the window as roughly 11 to 13 weeks and chose 11 weeks for its model. Model estimate on the earlier measure, on weekly data from 18 February 2022 to 24 April 2026. Past results do not indicate future results.

The size. A $1 trillion weekly rise in Global Liquidity was followed by a Bitcoin weekly return about 4.1% higher 11 weeks later: t 3.13, significant at the 99% level, net of Bitcoin’s own momentum. Model estimate on the earlier measure, on weekly data from 18 February 2022 to 24 April 2026. The estimate describes the study’s sample only. Past results do not indicate future results.

The reverse test. The study also asked whether Bitcoin’s returns lead Global Liquidity. A Granger test in that direction found no evidence of it: p = 0.99. Model estimate on the earlier measure, on weekly data from 18 February 2022 to 24 April 2026. Past results do not indicate future results.

Source: BlockPhi, “Global Liquidity and the case for Bitcoin” (published study, 2026).

The test on the current series

On the current series, the collateral base’s six-week change, set 11 weeks ahead, matched the direction of Bitcoin’s six-week return in 228 of 343 weeks from January 2020 (66.5%). Chance would be about 51%. The test runs on the collateral base, before the collateral multiplier, and counts direction only, not size. Model estimate. The test window ends 24 July 2026. Past results do not indicate future results.

Share of weeks in which the directions matched, %

How to read it. Consecutive six-week windows overlap, so the 343 weeks are not 343 independent observations. The 11-week offset comes from the published study, which chose it on weekly data from 18 February 2022 to 24 April 2026, inside this test window. It was not re-chosen on the current series. Every week in the test falls before 31 July 2026, so the test runs on reconstructed history. Recorded live since 31 July 2026. Weeks before 31 July 2026 are reconstructed with the current method, each input entering on a lag that follows its usual release schedule. They are not a record of what was known on each date. Model estimate, on weeks to 24 July 2026. Past results do not indicate future results.

Five periods

Five periods in date order; the fifth is read in two parts, 5a and 5b. Those from January 2020 are read from the series as it stands today. Each is marked Consistent, Partly consistent or Missed. Consistent means the series moved in the same direction as risk assets and as the funding data from other publishers over the period. Partly consistent means it did so for part of the period only. Missed means it did not. For a period marked Partly consistent, the text says which part did not.

Values are read from the series as of 9 October 2026 and are model estimates. Every period falls before 31 July 2026 and uses reconstructed history. Six-week changes are of Global Liquidity, the full series. Bitcoin weekly closes are dated context only and start in January 2022 on this page. Data from other publishers is listed with its source and date.

  1. Period 1 · September 2019 · Missed

    The September 2019 repo stress

    The series
    Not shown. The reconstruction starts in May 2019; charts on this site start in January 2020. This period is described in words.
    Other publishers, dated
    • Secured overnight financing rate 5.25% on 17 September 2019, with a 99th percentile of 9.00%. Federal Reserve Bank of New York, SOFR data, read 23 September 2026.
    • Bank reserves at a multi-year low of less than $1.4 trillion in mid-September 2019. Federal Reserve, FEDS Notes, 27 February 2020.
    • Federal Reserve total assets about $3.77 trillion on 11 September 2019 and $4.17 trillion on 25 December 2019. Federal Reserve H.4.1, via FRED (WALCL), read 23 September 2026.
    • US M2 up 2.0% from September to December 2019. Federal Reserve H.6, via FRED (M2SL), read 23 September 2026.

    The series did not call the September 2019 repo stress. In mid-September 2019 cash, not collateral, was short: corporate tax payments and Treasury issuance moved bank reserves into the Treasury’s account at the Federal Reserve. The response came from the Federal Reserve, whose balance sheet grew from mid-September to the end of December 2019 while US M2 grew far less.

  2. Period 2 · March to June 2020 · Partly consistent

    March to June 2020

    The series
    • Collateral base $73.73 trillion on 21 February 2020, $75.27 trillion on 20 March 2020 and $80.21 trillion on 26 June 2020.
    • Central bank money $21.33 trillion on 20 March 2020 and $25.50 trillion on 26 June 2020.
    • Global Liquidity +2.83% from 21 February to 13 March 2020, and +6.62% from 20 March to 26 June 2020.
    • Six-week change +3.35% on 13 March 2020, +4.92% on 8 May 2020 and +8.48% on 18 December 2020, the highest in the history shown.
    Other publishers, dated
    • Treasury collateral multiplier of the largest dealers: 7.5 on average in 2019, 5.5 in the second week of March 2020. Infante and Saravay, Federal Reserve, FEDS Notes, 4 December 2020.
    • Initial margin at central counterparties up about $300 billion over March 2020; peak variation margin call $140 billion on 9 March 2020. BCBS, CPMI and IOSCO, Review of margining practices, 29 September 2022.
    • US M2 growth at its peak, 26.8% year on year, in February 2021. Federal Reserve H.6, via FRED (M2SL), read 23 September 2026.

    The collateral base rose between 21 February and 20 March 2020, partly from the Federal Reserve’s balance sheet and more from sovereign bond prices as yields fell. So the series rose into the March 2020 sell-off and gave no warning of it: a collateral measure at market value reads a bond rally as expansion. Funding markets were under strain in the same weeks: the largest dealers reused less collateral and margin calls rose. The response that followed came mostly from central banks, and the six-week change peaked on 18 December 2020. Deposits told the same story that cycle, later and on a monthly clock: US M2 growth peaked with the February 2021 data.

  3. Period 3 · 2022 · Consistent

    2022

    The series
    • Global Liquidity $169.15 trillion on 25 February 2022, its 2022 high, and $155.25 trillion on 21 October 2022, its 2022 low: a change of −8.22%.
    • Six-week change −1.47% on 1 April 2022 and −5.01% on 30 September 2022, the lowest in the history shown.
    Bitcoin, dated
    • $39,232 on 25 February 2022.
    • $19,165 on 21 October 2022.
    • $16,508 on 25 November 2022, its lowest weekly close of 2022.
    Other publishers, dated
    • US bank reserves down $869 billion, 22.7%, from 30 March to 28 September 2022; overnight reverse repo take-up up $581 billion over the same dates. FRED (WRESBAL, RRPONTSYD), read 23 September 2026.
    • US M2 down 1.1% from March to September 2022. Federal Reserve H.6, via FRED (M2SL), read 23 September 2026.
    • Nominal broad dollar index at about 128 on 26 September 2022, its 2022 high. Federal Reserve, via FRED (DTWEXBGS), read 23 September 2026.
    • The 30-year gilt yield rose 160 basis points in a few days. Bank of England, letter to the Treasury Committee, 5 October 2022.
    • The Bank of England bought gilts from 28 September 2022 and ended the purchases on 14 October 2022. Bank of England, statement on the end of gilt market operations, 17 October 2022.

    Global Liquidity fell from February 2022 and made its 2022 low on 21 October 2022, a week after the Bank of England ended its emergency gilt purchases. Most of the fall in the collateral base came from the stronger dollar. Lower bond prices took off a smaller part, new issuance added a little, and the Federal Reserve’s balance sheet in the series changed little. The six-week change reached its lowest in the history shown in the week ending 30 September 2022, the week those purchases began. Bitcoin’s lowest weekly close of 2022 came in the week ending 25 November 2022.

    In the United States, M2 fell over those months too, and bank reserves fell far more, while cash moved into the Federal Reserve’s reverse repo facility. The difference was the bond price leg and the split by source, not the direction.

  4. Period 4 · September 2023 to March 2024 · Partly consistent

    September 2023 to March 2024

    The series
    • Collateral base $89.09 trillion on 29 September 2023 and $92.50 trillion on 29 March 2024.
    • Global Liquidity +4.03% from 29 September 2023 to 29 March 2024.
    • Six-week change +0.53% on 10 November 2023, its first positive week of the period, and +3.40% on 8 December 2023. Negative again from 26 January to 16 February 2024, reaching −2.04% on 16 February 2024.
    Bitcoin, dated
    • $26,911 on 29 September 2023.
    • $37,319 on 10 November 2023.
    • $44,193 on 8 December 2023.
    • $41,818 on 26 January 2024.
    • $52,161 on 16 February 2024.
    • $69,903 on 29 March 2024.
    Other publishers, dated
    • Federal Reserve total assets down about $500 billion, reserve balances up $303 billion and overnight reverse repo take-up down $925 billion, 27 September 2023 to 27 March 2024. Federal Reserve, balance sheet developments report, May 2024.
    • US ten-year Treasury yield 4.98% on 19 October 2023 and 3.88% on 29 December 2023. Federal Reserve H.15, via FRED (DGS10), read 23 September 2026.
    • US M2 below its level of a year earlier in each month from September 2023 to February 2024, and down 3.4% year on year in October 2023. Federal Reserve H.6, via FRED (M2SL), read 23 September 2026.

    The Federal Reserve’s balance sheet shrank over these months, and the series’ Federal Reserve leg fell with it. The collateral base still rose, mostly because the government bonds the system borrows against recovered from the October 2023 yield peak: collateral value that no M2 series and no central bank headline records. The headline said tightening. The series showed easing. Its six-week change turned positive on 10 November 2023.

    Bitcoin rose from 29 September to 10 November 2023 while the six-week change was still negative. And the six-week change was negative again from 26 January to 16 February 2024, while Bitcoin rose again. The series matched the regime, not every move within it.

  5. Period 5a · April and 10 October 2025 · Missed

    April 2025 and the 10 October 2025 sell-off

    The series
    • Six-week change +1.01% in the week ending 4 April 2025 and +1.93% in the week ending 11 April 2025.
    • Six-week change +0.41% in the week ending 10 October 2025.
    Bitcoin, dated
    • $112,996 on 10 October 2025.
    Other publishers, dated
    • Longer-term US Treasury yields about 30 basis points higher on 11 April 2025 than on 2 April 2025; repo rates within recent ranges; no evidence of an unwind of the basis trade. Roberto Perli, Federal Reserve Bank of New York, speech, 9 May 2025.

    The series gave no warning of the April 2025 Treasury sell-off. That was a collateral price event, not a funding one: Treasury yields fell at first, then longer-term yields rose sharply in the days after 2 April 2025, while repo rates, though volatile, stayed within recent ranges. A weekly series describes a two-week event. It does not predict it.

    The series gave no warning of the 10 October 2025 sell-off either. It reads the regime over weeks, not event crashes.

  6. Period 5b · Late 2025 · Partly consistent

    Late 2025

    The series
    • Six-week change below zero in every week from 17 October to 5 December 2025, reaching −1.12% on 7 November 2025, its lowest reading of the period.
    • Six-week change positive in every week from 12 December 2025 to 27 March 2026, at +1.19% on 19 December 2025 and +1.73% on 27 March 2026.
    • Global Liquidity +10.47% over the 12 months to 6 February 2026.
    Bitcoin, dated
    • $106,476 on 17 October 2025.
    • $90,275 on 12 December 2025.
    • $95,521 on 16 January 2026.
    • $70,531 on 6 February 2026.
    • $66,359 on 27 March 2026.

    The level of the series rose over the year to 6 February 2026. What the series added was the split by source and the six-week change. About half of the rise in the collateral base was new sovereign issuance, net of smaller central bank balance sheets, and about half was the weaker dollar. Bond prices took a little off. This is a model estimate, derived from the components, which are offered under licence; on this site they are shown frozen at the week ending 18 September 2026. The series shows that split; a single total does not.

    The six-week change was below zero from 17 October to 5 December 2025, and Bitcoin’s weekly close was lower on 12 December than on 17 October 2025. Bitcoin fell again from 16 January to 6 February 2026, 10 to 13 weeks after the six-week change reached its lowest reading on 7 November 2025. One period does not make a rule. Past results do not indicate future results.

    From 12 December 2025 the six-week change was positive in every week to 27 March 2026, while Bitcoin’s weekly close on 27 March 2026 was lower than on 12 December 2025. That part did not match, so the period is marked Partly consistent.

One lesson

In 2020, 2022 and 2025 to 2026 the level of a money stock and the level of our series told a similar story, late. In 2023 to 2024 they disagreed: the series rose with risk assets while US M2 was below its level of a year earlier through February 2024. The six-week change and the split by source spoke earlier in 2020, 2022 and late 2025, and showed which of currency, central banks and bond prices was moving. They said nothing useful in September 2019, in the first weeks of March 2020, in April 2025 or on 10 October 2025. For weeks at a time in late 2023, early 2024 and early 2026, the six-week change and Bitcoin went opposite ways. That is the record. Read the statistics above with it. Past results do not indicate future results.

Global Liquidity growth

Global Liquidity’s change over the last 3 months, annualised, and over the last 12 months, each Friday to 9 October 2026. Each line starts once it has a full window of weekly values from 3 January 2020. Model estimate. Past results do not indicate future results.

Per cent, weekly to 9 October 2026.

Weeks before 31 July 2026 are a reconstruction with the current method.Source: BlockPhi Global Liquidity Model
Global Liquidity growth on the last Friday of each year from 2020 to 2025, and the latest week. Per cent. A dash means the window reaches before January 2020. Model estimate.
Week ending3-month change, annualised12-month change
25 December 2020+44.22%–
31 December 2021+6.28%+8.50%
30 December 2022+16.98%−3.29%
29 December 2023+17.60%+3.92%
27 December 2024−11.20%+1.29%
26 December 2025−0.19%+9.65%
9 October 2026+8.23%+5.61%
Page history
24 September 2026: First published. 3 October 2026: Links to the methodology’s point-in-time section.

The series is offered to institutions under licence.

Terms, delivery and notice periods are agreed in each licence.

For funds, market makers, family offices, treasury companies and data platforms.

Request access

Request access: contact@blockphi.com or www.blockphi.com/institutions#request