Fee Estimation

Fee estimation is the process by which a wallet predicts the feerate a transaction needs to confirm within a desired number of blocks. Block space is a continuous auction: miners fill each block with the highest-paying transactions waiting in the mempool, so the "right" feerate is whatever outbids enough of the current queue. Getting it right matters in both directions — pay too little and a transaction can languish unconfirmed for hours or days; pay too much and you overspend on a permanent, irreversible payment.

How Bitcoin Core estimates

Bitcoin Core exposes fee estimation through the estimatesmartfee RPC, which returns an approximate fee per kilobyte for a confirmation target between 1 and 1008 blocks. The built-in estimator, CBlockPolicyEstimator, is history-based rather than mempool-snapshot-based. It sorts observed transactions into exponentially spaced feerate buckets (from 1,000 to 10,000,000 sat/kvB, spaced by a factor of 1.05) and tracks, for each bucket, how many blocks transactions at that feerate took to confirm. From this rolling record it derives the lowest feerate that historically achieved confirmation within the target. A CONSERVATIVE mode biases toward higher, more reliable estimates; ECONOMICAL reacts faster to falling fees. The design is deliberately manipulation-resistant: it only counts transactions the node saw enter the mempool and later confirm, so a miner cannot easily inflate estimates by stuffing their own blocks with fake high-fee transactions that never touched the public relay network.

Limitations and alternatives

Because it leans on past confirmations, the history-based estimator lags sudden regime changes — a burst of new demand or an abrupt collapse takes time to show up in the buckets. Many wallets therefore supplement it with live mempool inspection: read the current backlog, sort it by feerate, and compute where in the queue a candidate feerate would land. That approach reacts instantly but can be gamed by broadcast-only spam that never intends to confirm, which is why serious implementations blend both signals. What actually gets into the mempool in the first place is governed by mempool policy, so estimates can also differ between nodes running different policy settings.

When the estimate was wrong

Underpaying is recoverable. Replace-by-fee (RBF) lets the sender rebroadcast the same transaction with a higher fee, and child-pays-for-parent (CPFP) lets the recipient spend the stuck output with a generous child fee that miners evaluate as a package. Lightning channel transactions rely on a related mechanism — see anchor outputs — and adversarial interference with fee-bumping is its own topic under transaction pinning attacks.

A practical workflow for the node runner

Day to day, fee management is mostly patience arbitrage. Query estimatesmartfee for a realistic target rather than the next block — few payments genuinely need it — and sanity-check against your own mempool's depth. Fee pressure follows human rhythms: weekends and quiet hours routinely clear cheaper than weekday peaks. Save housekeeping for the lulls — consolidating fragmented UTXOs during a low-fee window is dramatically cheaper than being forced to sweep dust at high rates later, and deliberate coin control keeps future transactions small and cheap. When you must move now, set an RBF-enabled feerate you can live with and bump only if the queue moves against you.

For the sovereign operator, the practical takeaway is simple: run your own node and estimate from your own mempool. A third-party fee API is a trusted oracle that can be lazy, wrong, or malicious — quietly inflated estimates skim money on every transaction, and there is a mild conflict of interest in any fee oracle operated by entities that profit from higher fees. Your own node gives you the same view miners have, for free, with nobody in between. Fee pressure is also the long-term revenue story for miners: as the subsidy halves away, fees become the dominant share of the coinbase transaction, which makes honest, accurate fee estimation a piece of infrastructure the whole network quietly depends on.