Bitcoin fee-sniping reorg EV chart

This page models a rational miner considering a private fork to reorganize the last block and claim an unusually large fee. The chart scans hashrate q and plots the number of additional net public-chain blocks the miner would tolerate before the continuation value becomes non-positive.

Model inputs

Economic payoff Extraordinary fee, F Fee the miner is trying to capture, in BTC.

Ordinary private-block reward, R Subsidy plus normal fees per private block. Existing private blocks, B Usually 0 at the start of a last-block reorg attempt.

Penalty / risk cost, Φ Expected reputational, detection, legal, or market-impact cost if the attack is attempted or succeeds.

Opportunity cost Cost model

Honest reward per public block Used only by the dynamic cost model. Extra cost per block event Financing, operational, or pool-defection cost.

Fixed c Used only by the fixed cost model.

Fork state and search bounds

Starting public lead, d₀ Use 1 for an attempt to reorg the most recent block. Max public lead to scan Higher values allow longer-shot continuation thresholds.

q min q max q step Default is the requested 0.1 increments.

Continuation threshold by miner hashrate

Results

q	p = 1 − q	c used	Attack?	Start EV	Best start give-up lead G*	P(win) at G*	E[events] at G*	Give up when public lead reaches	Chart y: additional net public blocks tolerated

Formula notes

The model uses the conservative one-block-ahead success condition: the private fork succeeds when its lead reaches d = -1. It gives up at a public lead boundary G. For each current public lead d, the page scans all G > d and keeps the best positive expected value.

p = 1 − q

ρ = q / p

Pwin(d,G) = (ρ^(d+1) − ρ^(G+1)) / (1 − ρ^(G+1)), when q ≠ p

Pwin(d,G) = (G − d) / (G + 1), when q = p

T(d,G) = { (G+1)[(1 − ρ^(d+1))/(1 − ρ^(G+1))] − (d+1) } / (p − q), when q ≠ p

T(d,G) = (d+1)(G−d), when q = p

EV(d,G) = Pwin(d,G)(F + B·R − Φ) + R·M(d,G) − c·T(d,G)

M(d,G) is the expected number of additional private blocks mined on paths that end in success, weighted by success probability. It is solved in JavaScript from the recurrence:

M(d) = q · [Pwin(d−1,G) + M(d−1)] + p · M(d+1)

M(−1) = 0, M(G) = 0

The plotted y-value is:

y = max(0, giveUpLead − d₀)

So with d₀ = 1, a y-value of 1 means: "try the reorg, but give up if the public chain gets to a 2-block lead." If a point hits the configured max public lead, increase the scan bound to see whether the positive-EV region extends farther.