Quantum Drive Live

Our prediction, next to what depositors actually earned.

Every width and delay this site recommends is re-fitted each night to real positions, not backtests. Below, each pool shows the rate the model predicted for a deposit at its Best Fit and, beside it, the rate real depositors at that width earned from fees collected on-chain. The gap between the two is published as the model's accuracy.

Fitted to
real positions
Last fit
nightly
Model accuracy · 30d
realized ÷ predicted
Pools reported
model version
$
Headline % = daily fee rate × 365 (the same scale as the Earnings Rate on your position cards).
Per pool

Predicted vs realized by depositors

Predicted = the exact-share concentrated rate at the pool's current Best Fit width and delay. Realized = gross fees real depositors at that width collected, from the book. Tap a row for what people at each width earned.
Lift curve

Fitted from real positions vs the old formula

Lift is how many times the whole-pool rate a concentrated position earned. The dashed line is the formula the app used to fall back on, 1.25 · w−1.19. The dots are what real positions implied, bucketed by half-band w, sized by how many positions sit in the bucket. Where the dots sit under the line, the old formula overstated.
Your contribution

Positions of yours that fed the model

Read-only. This page mirrors the wallet connected on Quantum Scout; it never asks you to sign anything and shows nothing beyond the public wallet address.
Data dividend: rebate coming. The plan is a lower per-wallet performance fee for wallets whose positions improved the model, applied in governance batches through the 48-hour timelock. It is a specification today, not a live rule; nothing on this page changes your fee.
Plain English

What this page proves, and what it does not

What it proves

  • For every pool listed, we publish the rate our model predicted for a deposit at the recommended width and the rate real depositors at that width actually earned. Both are gross trading fees collected on-chain, on the same scale.
  • The accuracy figure is the gap between those two over the last 30 days: how right the model was last month, in public, not a backtest score.
  • The fit is redone every night from real positions. A parameter only moves when at least five positions agree and the fit gets tighter; otherwise the feed says "insufficient data" and this page says so too.
  • Every served number is uncapped. The old formula's curve is drawn alongside the fitted points so you can see exactly where it overstated tight bands.

What it does not

  • These are gross fees. They do not subtract impermanent loss, the performance fee on collected fees, gas, or the tokens' own price move. A position can collect every dollar shown and still be worth less than the deposit.
  • A realized rate is history, not a promise. Volume can vanish overnight; a pool that paid 1,000% last week can pay nothing this week.
  • Small n is weak evidence. One depositor at one width is an anecdote, five is a hint, fifty starts to mean something.
  • Nothing here is financial advice. Read the risks and the fees before depositing.

QuantumPools is experimental software. Providing liquidity can lose money. Feed: /drive-live.4663.json · How Quantum Drive works · Risks