Macro
What the candles cannot see. This page is descriptive: it shows how closely crypto has been moving with the wider market, not what that market is about to do.
Bitcoin and the Nasdaq
Correlation of daily returns — not of price levels. 30 sessions 90 sessions
A single figure over five years would read about +0.4 and would hide everything that matters. The curve does not: bitcoin traded like a growth stock through 2022, came fully unstuck from the Nasdaq during 2023 — briefly negative — and has been moving with it again since late 2025.
Read a high reading as « the market is currently being driven by something this site cannot see ». Both crashes of 2026 were macro events: a Fed nomination, tariffs, margin calls on metals, ETF outflows. None of that is in a candle, an order book or a funding rate.
What this has been measured to be worth
Printed next to the drawing on purpose. Two lines moving together invite the reader to assume one drives the other; the measurements say otherwise. Tested over five years on five assets:
| Series | Measured effect on crypto | Verdict |
|---|---|---|
| Nasdaq, S&P 500, dollar | A Nasdaq fall of more than 1.5% is followed by +0.38% on BTC at 3 days (baseline +0.33%), p = 0.47 | no effect |
| Gold | An effect appears on BTC (p = 0.018) but replicates on no other asset — out of sample, p from 0.084 to 0.356 | not established |
| US 10-year yield | A mild decline, same direction on all five assets tested, about one point at 3 days (ETH p = 0.023, BNB p = 0.044; SOL and XRP not significant) | best candidate, unproven |
This is why the macro block reaches the model as context but has no dedicated field in its output schema — a field forces a reading, and no measurement here supports forcing one.
The five series, last 180 days
Each series reduced to its own drift in per cent, the only basis on which an index near 20,000 and a yield near 4 can share an axis.
Correlation to the Nasdaq, by asset
Same calculation, run on every asset on the board. A blank cell means the asset has not been tracked long enough for that window — the figure is withheld rather than computed on too few points.
| Asset | 30 sessions | 90 sessions |
|---|---|---|
| Cardano | +0.36 | +0.45 |
| Polkadot | +0.34 | +0.45 |
| Avalanche | +0.31 | +0.40 |
| Solana | +0.16 | +0.40 |
| Chainlink | +0.17 | +0.38 |
| Ethereum | +0.16 | +0.38 |
| Bitcoin | +0.18 | +0.36 |
| BNB | +0.21 | +0.35 |
| Dogecoin | +0.18 | +0.34 |
| XRP | +0.06 | +0.33 |
| Litecoin | +0.12 | +0.32 |
| TRON | +0.07 | +0.15 |
How this is computed
- On returns, never on levels. Two series that both rise produce a high correlation whether or not they are related — they share a trend, not their movements.
- Dates are intersected, never filled. The Nasdaq closes at weekends; crypto trades every day. Filling the gaps would compare a Saturday crypto move against a frozen index.
- Returns that span a gap are dropped. A missing stretch of data would otherwise enter the calculation as one enormous « daily » move. This is not hypothetical: a 364-day hole in the cached BTC closes inflated one window to +0.82 where it belongs at +0.35.
- The points of the curve are not independent. Two neighbours share every observation but one, so the line looks smooth by construction. It is there to be read, not to be counted.
- Correlation is neither causation nor prediction. Part of what shows up during crashes is mechanical: when liquidity leaves, everything falls together.