Most accurate gas oracle, live gap vs realized priority fee

The absolute difference, in gwei, between each gas oracle's predicted p50 priority fee and the realized p50 priority fee in the next mined block, ranked on the p99 of that gap over 24h (the worst 1% of blocks, where gas spikes actually cost money). Owlracle currently leads the active chain tab at 0.109 gwei. The covered column adds the inclusion-side view: the share of time the prediction sat at or above the realized p50. What it does NOT measure. USD over-pay cost (depends on gas_used × ETH price) or fundamental oracle reliability beyond priority-fee accuracy.

Because at current fee levels the p50 gaps are fractions of a micro-gwei apart across oracles, on a 100k-gas transaction a 0.001 gwei error is about a thousandth of a cent, so a p50 ranking orders economically indistinguishable noise. The p99 gap captures the volatile minutes (mempool spikes, NFT mints, MEV bursts) where predictions genuinely diverge and a wrong number either overpays or misses the block. The typical p50 and p90 gaps remain visible as secondary columns.

The share of time over 24h that the oracle's posted p50-tier prediction was at or above the realized p50 priority fee of the latest mined block. Absolute gap treats over-prediction and under-prediction the same, but they are not symmetric for users: over-predicting means slightly overpaying, under-predicting means a transaction paying exactly the predicted fee would have ranked below the block's median and risks waiting. Inclusion-confidence oracles (Etherscan) should score high here by design; percentile trackers sit near 50% by construction. Read gap and covered together: low p99 gap + high covered is the sweet spot.

Both are EIP-1559 with a proper dynamic base fee, which makes 'priority-fee prediction error' an apples-to-apples comparable metric across them. Avalanche C-Chain was dropped because its auto-tuning fee market collapses the priority fee to ~0 by design, which makes the prediction-error metric uniformly ~0 across all oracles and non-informative. BNB Chain is excluded because it isn't EIP-1559 (effective fee = gasPrice only, priority is structural noise). Optimism / Base / Arbitrum and other L2 OP Stack rollups are excluded because their priority fee is ~0 (centralised sequencer, no MEV) and the relevant cost is L1 data fee, a different metric that belongs in a separate bench. Solana uses lamports per CU with Jito MEV, different fee model entirely.

Read the p99 gap (does it blow out during spikes?) together with the covered rate (does it err on the side of inclusion or under-bid?) for the chain your product runs on. Owlracle currently leads the active tab at 0.109 gwei (p99) but the right oracle depends on whether you can tolerate over-pay (then prefer high covered), whether you optimise for tail behaviour, and whether the oracle's free quota fits your call volume. The bench gives you the live numbers. it cannot tell you which trade-off your product wants.

L2 OP Stack chains have priority fee ≈ 0 because the sequencer is centralised (no MEV competition, no public mempool). The actually meaningful cost on an L2 is the L1 data fee, what the sequencer pays Ethereum mainnet to post the batch, which is a different prediction problem. We're considering a separate `l2-data-fee-prediction` bench for that. Including L2s in this bench would silently produce flat ~0 numbers that pollute the comparison.

Per chain, every oracle is polled at its tier-tolerant cadence and the predicted priority fee per tier is buffered with the predicted block height. When that block is mined, the harness pulls the full block via `eth_getBlockByNumber(.., true)` on the chain's PublicNode RPC and computes the realized percentile from the actual `maxPriorityFeePerGas` values across every included transaction. Absolute error (|predicted, realized|, in gwei) is recorded per (oracle, tier, chain) as both a gauge and a histogram. p50 / p90 / p99 are computed via Prometheus `quantile_over_time` over the last 24 h. Empty or low-tx blocks are flagged separately because the realized percentile is noisy when transaction count is near zero.

EIP-1559 makes the base fee deterministic (it adjusts by ±12.5% per block based on the previous block's gas used), so every oracle agrees on base fee within a fraction of a gwei. The hard part is predicting the priority-fee distribution in the *next* block. Priority fees are set by users in response to mempool congestion, which shifts on swap activity, MEV bot deployments, NFT mints and DEX volume in a way no historical-lookback model can fully anticipate. The leaderboard surfaces which oracle's lookback / inference scheme tracks reality best per chain, sustained over 24 h.