Can Repair Carry On?

History

From Capitulation to Repair

As discussed in Note 4, Bitcoin’s rapid recovery carried price back above the Short-Term Holder Cost Basis, now around —, and weekly closes have held above it since. At the same time, the cost basis of the most recent buyers began to stabilise around the same region. Over the past several days, that improvement has held, allowing our Value Regime Detection Framework to confirm a shift from Capitulation into Repair. We are now — days into the current Repair regime.

The significance of this phase lies in whether that recovery can persist. Holding above the Short-Term Holder Cost Basis reduces the pressure on recent buyers and gradually improves their ability to tolerate further volatility. As that pressure eases, market confidence can recover and fresh demand is more likely to engage without being immediately tested by underwater positions.

For the Repair regime to keep developing, the key condition is that the cost-basis ordering defining the regime remains intact.

Mystery

A Familiar Repair

With the regime shift confirmed, price provides the first test of whether the market is beginning to behave like a Repair as well.

The current transition followed a sharp advance toward —, closely resembling the run-ins to the 2019 and 2023 Repair regimes. Across those episodes, price gained roughly — over the 14 days leading into confirmation. Momentum then remained positive over the following month, with both completed Repairs extending the move by another — before their paths began to diverge.

So far, the current episode has followed a familiar opening sequence. But two historical observations are far too few to treat the price path itself as a reliable template. The more pertinent question is whether the investor behaviour that sustained those prior Repairs is also returning, and whether the underlying conditions can persist, particularly on the spot-demand side of the market.

The Framework

To assess whether Repair can persist, we examine both sides of the market: the return of demand through ETF and network-wide capital flows, and the supply response from recent buyers moving back into profit.

Since spot Bitcoin ETFs began trading in the U.S. in January 2024, a simple rule of thumb has become unusually persistent: “If weekly inflows reach $1–2 billion, institutions are buying aggressively, so price should rise next.” Embedded in that view is the assumption that ETF flow is a leading signal, because creations and redemptions are treated as directional buying or selling.

Figure 3 offers the first challenge to that reading. It plots weekly net ETF flow in BTC against spot and highlights five rallies, each defined as an advance of at least 25%, while allowing for drawdowns of no more than 15% along the way. In every case, ETF inflows strengthened during the rally. Before and after those advances, flows were materially weaker and, in many windows, outright negative.

The fifth of those rallies is still running, so its flows can be read as they arrive. The current rally is showing the same pattern. ETF inflows have averaged roughly — per week, with the latest week alone bringing in about —, or —.

Across the completed episodes, average weekly inflows during the rally were substantially higher than in the surrounding windows, while the current rally is following the same broad pattern so far. That does not, by itself, prove that price leads ETF flow. But it does weaken the simple idea that large inflows consistently arrive first and then pull price higher.

The useful part is that this relationship can be tested rather than assumed. The framework that follows examines the timing at shorter horizons, allowing us to identify whether this behaviour persists or begins to change.

The earlier analysis of ETF flows through rallies showed that inflows generally strengthen once a rally is already underway. The next question, then, is whether those flows actually lead price or simply respond to it.

To test this more directly, Figure 4 examines the timing across the full ETF history, comparing correlations from ten trading days before to ten days after each observation. Positive lags capture cases where ETF flows move first; negative lags capture cases where price leads and flows follow.

The strongest relationship appears on the same day, at roughly 0.45, followed closely by the one-day price-leading lag at around 0.42. From there, the correlation fades quickly, especially on the side where ETF flows would need to lead.

This does not prove causality, but it weakens the case for treating daily ETF flows as a reliable leading signal. So far, the data fits better with flows moving alongside price or responding shortly after it.

The daily analysis gives us the clearest view of short-term timing, but it may also carry noise that disappears when flows and returns are measured across broader windows. To test whether the same relationship holds beyond day-to-day movements, we extend the analysis to weekly, bi-weekly and monthly sampling blocks.

As Figure 5 shows, price and ETF flows move increasingly in sync as the sampling window expands. Same-period correlation rises from roughly 0.45 on daily data to 0.69 weekly, 0.78 bi-weekly and 0.88 monthly. What does not strengthen, however, is the tendency for ETF flows to move ahead of price.

Over longer horizons, the two increasingly look like parts of the same market move. In other words, ETF flows have behaved more like a response to changing market conditions than a reliable signal of what price will do next.

If ETF flows mostly move with price rather than ahead of it, the next question is what sits inside those inflows. One possible explanation for part of this co-movement is the cash-and-carry trade, a delta-neutral strategy that can generate meaningful ETF inflows without necessarily reflecting a bullish view on Bitcoin.

Cash-and-carry desks buy a US spot Bitcoin ETF, most often IBIT for its liquidity and tight creation and redemption, while shorting an equivalent notional amount of CME futures. The position is directionally hedged, so it creates ETF inflows without expressing a view on Bitcoin, and the return comes from the front month futures basis converging towards spot.

A high basis is not on its own attractive. It has to beat the return available on a lower-risk alternative. Figure 6 therefore measures Net Carry as the annualised front-month CME basis less the 3-month Treasury bill yield.

That spread has varied considerably, from roughly +17% in early 2024 to nearly -18% later that year, before recovering to around — in the latest observation. The axis below is trimmed to the middle of that record, so the deepest negative readings sit beyond it. Positive Net Carry means the futures premium exceeds the T-bill return, making the trade more attractive. Negative readings imply that even before ETF fees, financing and roll costs, the lower-risk alternative offers a better return.

More recently, Net Carry briefly fell below zero in late July before returning to positive territory throughout August, ending the latest week at roughly —. The basis trade therefore remains above its risk-free hurdle before additional costs, although the spread is considerably less attractive than the +6.7% reached in mid-July.

With the economics of cash-and-carry established, the next step is to see whether that incentive leaves a visible footprint in ETF flows.

Figure 7 compares weekly net ETF flow with the annualised Net Carry available during the same week. Across all 137 observations, the relationship is moderately positive, with a Pearson correlation of roughly 0.55 (p < 0.001) and an R² of around 0.30. In other words, stronger carry has generally coincided with larger ETF inflows, while weaker or negative carry has tended to appear alongside softer flows.

The green points highlight the 80 weeks that occurred during the five identified price rallies. The correlation remains similar at roughly 0.54, but the concentration of large inflows becomes more revealing: 14 of the 15 largest rally-week inflows occurred while Net Carry was positive.

This does not prove that cash-and-carry activity is the sole driver of those ETF inflows, but the pattern is consistent with carry economics contributing to ETF demand when the basis trade is most attractive.

Recent observations remain broadly consistent with that pattern, though not uniformly. Net Carry remained positive across all four August observations, while ETF flows were positive in three of those four weeks, including the two latest observations. The intervening outflow is a useful reminder that carry alone does not determine ETF demand.

The relationship between price, futures basis and ETF flows can be understood as a sequence in which one source of demand gradually creates the conditions for another.

The process begins with rising demand pushing Bitcoin higher. As price momentum builds, demand for futures can lift the futures premium and widen the basis. If the annualised basis rises sufficiently above the Treasury yield, pushing Net Carry into positive territory, the trade becomes more attractive to cash-and-carry desks. New positions can then be opened by buying the ETF spot leg while simultaneously shorting CME futures, adding another source of ETF inflows. Figure 8 summarises this sequence schematically.

This mechanism is naturally self-limiting. As short futures positions accumulate, they put pressure on the premium and gradually compress Net Carry. Once the spread becomes too narrow, new entries lose their appeal. If Net Carry turns negative, existing cash-and-carry positions may also begin to unwind, requiring desks to close the short futures leg and sell the corresponding ETF exposure. As a result, some ETF outflows may reflect the economics of closing a hedged carry trade rather than a purely bearish view on Bitcoin.

After challenging the one-dimensional assumption that ETF inflows simply represent directional spot buying, we can now read ETF flows alongside changes in CME open interest and Net Carry to infer the likely motive behind them.

ETF flows show whether exposure is entering or leaving the funds, while changes in CME open interest indicate whether futures positioning is expanding or contracting. Together, those two signals provide the first-pass distinction between directional activity and the opening or closing of a cash-and-carry trade. Net Carry then tests whether the economics of the basis trade support that interpretation.

The last two weeks show why this distinction matters. In the week of 17 August, ETF inflows reached 28.6K BTC while CME open interest rose by about 12.3K BTC. With Net Carry positive at around 1.75%, the combination aligned with Cash-and-Carry Entry. A week later, inflows remained strong at roughly 11.8K BTC, but open interest fell by around 6.8K BTC. Net Carry stayed positive at about 2.61%, yet the Primary Regime shifted to Directional Buying.

Table 2 formalises this logic. ETF flow and the weekly change in CME open interest determine the Primary Regime. The Net Carry Layer then classifies the economic cross-check as Supports, Neutral, or Weakens. Positive Net Carry Supports Cash-and-Carry Entry and Weakens Cash-and-Carry Exit; negative Net Carry does the reverse. Directional Buying and Directional Selling remain Neutral because carry is not diagnostic when ETF flow and futures positioning move in opposite directions. These labels estimate the dominant positioning regime rather than investor intent.

Figure 10 applies these rules across the ETF history, showing rotations between Directional Buying, Directional Selling, Cash-and-Carry Entry, and Cash-and-Carry Exit. The Net Carry Layer adds a second read on whether carry economics reinforce or weaken the Primary Regime. The week of 20 July, for example, showed Cash-and-Carry Entry but slightly negative Net Carry, placing the second layer in Weakens.

In the latest classified week, the Primary Regime is Directional Buying, with the Net Carry Layer remaining Neutral. ETF demand stayed positive, but the decline in CME open interest suggests that the inflow was not accompanied by the expansion in futures positioning we would expect from fresh Cash-and-Carry Entry. This marks a notable change from the previous week, when rising ETF flows and expanding open interest were reinforced by positive Net Carry. For now, the evidence points towards a greater directional component in ETF demand rather than a continued build-up of hedged carry positions.

Having examined spot demand through the ETF channel, we can now zoom out and look for its broader footprint across the Bitcoin network. One of the clearest aggregate measures for this purpose is Realised Cap, which values each coin at the price at which it last moved on-chain, effectively approximating the network’s aggregate cost basis.

As Figure 11 shows, its 30-day rate of change provides a useful proxy for the direction and intensity of net capital flows. Repair regimes have historically marked the point where this measure moves out of persistent contraction and back into positive territory. Across the two completed Repairs in this window, the average 30-day change settled within roughly —.

The current episode is already averaging —, although that reading is elevated by the Repair’s short duration and the sharp repricing accompanying the recent rally. As the sample extends, the average may moderate towards a smaller positive rate if subsequent changes are less extreme. Even so, the early shift is consistent with capital flow turning from net withdrawal towards renewed net inflows.

Having followed demand from ETF flows into network-wide capital formation, we can now turn to supply. Since the latest rally leg accelerated on 19 August, spot has moved from — to —, while Short-Term Holder Supply in Profit shifted from — to —, a — percentage-point change.

Figure 12 shows how that repricing is being realised. Daily Short-Term Holder Realised Profit moved from — to — over the same window, a — change. Higher profit-taking is not bearish by itself. In Repair, it is a natural consequence of recent buyers moving back into profit. The question is whether that supply begins rising faster than demand can absorb it.

This is why the two panels matter together. When a very large share of Short-Term Holder supply is in profit while realised profit is also accelerating, the market is simultaneously exposing more profitable supply and seeing more of it spent. That combination can warn of rising short-term exhaustion risk within Repair, without implying a broader cycle top.

The Road Ahead

Having examined how demand and supply responded to the recent move from Capitulation into Repair, the next question is what keeps the market in this regime. Demand has improved as capital flows turned positive, while the rally has also moved a larger share of Short-Term Holders into profit and increased the potential for profit-taking.

Under the QFI Labs Value Regime Detection Framework, Repair remains in force until a competing cost-basis ordering clears the relevant classifier deadbands. On the downside, the 1d–1w Cost Basis, now —, would need to move sufficiently below the Short-Term Holder Cost Basis at — to trigger a return to Capitulation. On the upside, moving above the True Market Mean at — is not enough by itself to trigger Breakout; that relationship must also clear its relevant deadband, and it currently sits — points short of doing so. This is why the raw ordering and the label can disagree for a time.

Figure 13 translates that structure into the price levels we can monitor. As long as neither boundary clears the threshold required to change the regime, the behavioural dynamics discussed throughout this report remain relevant. This is therefore the key structure to watch as Repair develops.

A natural question after Repair confirmation is how much of the broader cycle advance has historically already occurred. The recent rebound makes that question especially relevant, with Bitcoin already — above its June low near $58K.

Rather than trying to identify the exact bear-market low, Figure 14 compares the historical outcome of accumulating through Capitulation with waiting until Repair had been confirmed.

Across the two completed cycles, waiting for Repair confirmation still left substantial subsequent upside. Daily accumulation during Capitulation produced returns of — and — to the next all-time high. Waiting until Repair still produced —, capturing an average of — of the earlier buyer’s gain.

History does not guarantee repetition. It shows that, within this framework, waiting for regime confirmation still preserved a substantial share of the subsequent cycle advance.