13 October 2026. Corrected 15 October 2026: fee figures recomputed after a base-fee decoding fix in TEMBRI; see the correction reply below. The edit history shows the original.
Motivation
I work on pricing position within the block on Ethereum L1, mainly out of protocol through MEV-Boost, and also in protocol under EIP-1559. Any such design needs to know what the front of the block pays today, what types of transactions are there, and how they pay.
Summary
The data covers sixteen sampled days, about 115,000 blocks: 1 July 2025, 1 July 2026, 1 to 7 August 2026 and 3 to 9 October 2026. It is execution data only, from TEMBRI, an open-source Rust tool I wrote, run on public Xatu data.
- Swaps in the first 10% of block gas pay on average 8 to 17 times the priority fee of swaps between 30% and 60%, on every day. Most of the fall happens before 20%. Fee levels fell by about three quarters from 2025 to 2026; the ratio did not fall.
- About half the swaps in the first 5% of gas paid no priority fee in early August, and about one in eight in early October.
- Titan accounts for much of the fall in zero-tip transactions at the front, at a nearly unchanged block share.
- Swaps that pay a priority fee are about twice as frequent at the front as overall, in both months.
- Builder labels show persistent differences in the correlation between priority fee and position.
Measurement tool: TEMBRI
TEMBRI (Transaction, Ethereum, Mempool, Block and Relay Intelligence) follows transactions from the mempool to the block. The current release reads Xatu execution data and produces the daily tables below. Mempool, relay and direct-payment data are in progress.
Code: GitHub - TLM-Research/tembri: Rust toolbox for reproducible Ethereum execution and mempool measurement. · GitHub (Apache-2.0), commit c30cdb9. The October run is configured in config/weekly_execution_reports.toml.
Data and method
- Source: Xatu canonical execution Parquet (ethpandaops): blocks, transactions and logs. Days are complete UTC dates. Blocks per cohort: 7,118 (1 July 2025), 7,167 (1 July 2026), 50,239 (August week), 50,219 (October week).
- Position: gas used up to and including the transaction, divided by block gas used, so a transaction is placed by where it ends.
- Position bands (disjoint): first 10% of gas, 10% to 20%, 20% to 30% and 30% to 60%. A band’s fee is the simple average over its transactions.
- Priority fee:
min(max priority fee, max fee - base fee)for type-2 and later transactions;gas price - base feefor legacy ones, floored at zero. Zero tip means 0 wei per gas. - Types: a swap has a Uniswap V2 or V3 swap event from one pool only. A multi-pool swap has swap events from two or more pools, so it includes both routed swaps and arbitrage.
- Builder: the block’s extraData string, which is self-reported. “Locally built” means it shows an execution-client version.
- Week columns: median of the seven daily values (ratios computed per day), with the range in brackets.
1. The front of the block pays more in priority fees
Average swap priority fee (gwei) by position band:
| 2025-07-01 | 2026-07-01 | Aug 2026 week | Oct 2026 week | |
|---|---|---|---|---|
| First 10% of gas | 15.63 | 4.42 | 3.83 (2.47 to 4.64) | 4.57 (2.13 to 6.59) |
| 10% to 20% | 5.26 | 1.41 | 1.00 (0.78 to 1.11) | 1.43 (0.80 to 1.91) |
| 20% to 30% | 3.05 | 0.70 | 0.53 (0.43 to 0.71) | 0.84 (0.44 to 1.06) |
| 30% to 60% | 1.64 | 0.36 | 0.28 (0.27 to 0.49) | 0.48 (0.26 to 0.56) |
| Ratio, first 10% / 30% to 60% | 9.5 | 12.3 | 9.4 (8.3 to 17.2) | 9.8 (8.2 to 13.2) |
The same ratio is 4.8 to 14.8 for multi-pool swaps and 2.5 to 4.2 for token transfers. Zero-tip transactions also reach the front (section 3), so priority fees alone do not capture all payment for front-of-block placement.
2. Transaction types at the front
A type’s share of the transactions in the first 5% of gas, divided by its share of all transactions (both by count):
| 2025-07-01 | 2026-07-01 | Aug 2026 week | Oct 2026 week | |
|---|---|---|---|---|
| Swap | 3.1 | 3.5 | 3.6 (1.8 to 4.2) | 2.0 (1.4 to 2.2) |
| Swap, positive fee | 3.1 | 2.4 | 2.2 (1.2 to 2.5) | 2.0 (1.3 to 2.1) |
| Swap, zero fee | 3.0 | 8.6 | 9.1 (4.7 to 12.2) | 3.0 (1.8 to 3.6) |
| Multi-pool swap | 3.0 | 3.2 | 2.6 (1.8 to 3.4) | 2.5 (1.9 to 2.8) |
| Token transfer | 0.6 | 1.2 | 1.7 (1.2 to 2.2) | 1.1 (1.1 to 1.2) |
| Native transfer | 0.6 | 0.5 | 0.5 (0.4 to 0.8) | 0.8 (0.7 to 0.9) |
Swaps that pay a priority fee are over-represented about twice in both months. The fall in the swap ratio comes from zero-tip swaps. Token transfers are half of the transactions completing in the first 5% on every 2026 day, against a fifth in 2025.
3. Zero-tip transactions at the front
Share of transactions with zero priority fee:
| 2025-07-01 | 2026-07-01 | Aug 2026 week | Oct 2026 week | |
|---|---|---|---|---|
| Swap, all | 4.5% | 17.8% | 18.5% (16.6% to 20.0%) | 8.8% (8.1% to 10.0%) |
| Swap, first 5% | 4.3% | 43.4% | 49.3% (42.4% to 53.8%) | 12.5% (10.3% to 16.5%) |
| Multi-pool swap, all | 5.1% | 16.7% | 14.1% (12.3% to 18.0%) | 9.9% (7.9% to 14.3%) |
| Multi-pool swap, first 5% | 10.4% | 21.0% | 17.8% (16.6% to 20.8%) | 6.4% (4.7% to 11.5%) |
| Token transfer, first 5% | 2.6% | 15.5% | 8.0% (4.6% to 8.9%) | 11.3% (8.6% to 13.9%) |
A zero-tip transaction at the front may be paid for by a direct transfer to the builder or by other transactions in its bundle, or be the builder’s own or exclusive flow. Execution data cannot tell these apart.
From August to October, zero-tip swaps in the first 5% fell from about 35,300 to 7,800, and zero-tip transactions of all types from 1.44 to 0.93 per block. By builder, all types (about 52,000 blocks per period):
| Builder | Block share | Front zero-tip txs per block | Share of all front zero-tip txs | Share of own zero-tip txs placed at the front |
|---|---|---|---|---|
| Titan | 47.0% / 47.9% | 2.15 / 1.08 | 70% / 55% | 24.0% / 18.0% |
| Quasar | 20.0% / 19.5% | 1.55 / 1.29 | 22% / 27% | 14.6% / 9.5% |
| BuilderNet | 11.8% / 14.8% | 0.39 / 0.66 | 3% / 10% | 6.8% / 9.7% |
| Eureka | 8.6% / 4.4% | 0.51 / 0.49 | 3% / 2% | 5.9% / 3.9% |
| bombora | 1.3% / 3.0% | 1.26 / 0.98 | 1% / 3% | 10.3% / 2.3% |
| Locally built | about 7% / 6% | 0 / 0 | 0 / 0 | 0 / 0 |
Each cell is August / October; “front” is the first 5% of gas.
Titan halved its zero-tip front transactions per block, at a nearly unchanged block share. The split by type within each builder is not yet measured. In October, 16.8% of bombora’s transactions pay no priority fee, but only 2% of those are at the front.
4. Builder ordering
Block share and mean within-block Pearson correlation between priority fee and position:
| Builder | 2025-07-01 | 2026-07-01 | Aug 2026 week | Oct 2026 week |
|---|---|---|---|---|
| Titan | 40.7%, -0.35 | 51.8%, -0.45 | 47.3%, -0.31 (-0.46 to -0.29) | 47.9%, -0.38 (-0.46 to -0.31) |
| Quasar | 4.7%, -0.41 | 16.4%, -0.55 | 20.0%, -0.54 (-0.55 to -0.53) | 19.6%, -0.50 (-0.57 to -0.48) |
| BuilderNet | 32.9%, -0.47 | 11.6%, -0.47 | 12.4%, -0.46 (-0.47 to -0.45) | 14.9%, -0.44 (-0.49 to -0.40) |
| Eureka | 8.8%, -0.19 | 7.4%, -0.24 (-0.27 to -0.22) | 3.9%, -0.14 (-0.16 to -0.13) | |
| bombora | 1.5%, -0.55 | 1.2%, -0.58 (-0.60 to -0.56) | 2.9%, -0.53 (-0.56 to -0.52) | |
| Locally built | 3.0%, -0.63 | 6.0%, -0.67 | 7.0%, -0.65 (-0.67 to -0.60) | 5.8%, -0.76 (-0.76 to -0.72) |
Locally built blocks have the most negative correlation, which fits fee ordering but does not show their source or rule. The correlation also reflects transaction types, gas, bundles and nonce order. Each builder’s value persists across the two weeks; Titan moves most, from -0.31 to -0.38.
5. Low-tip native transfers fell
Native transfers fell from about 1.4 million a day in the August week (average priority fee 0.20 gwei) to about 0.5 million in the October week (0.68 gwei), the 2025 level. Total transactions fell by only about 0.7 million a day (345 to 247 per block); other types rose by about 0.2 million. The cause cannot be seen from execution data.
Limits
- Swaps are detected only on Uniswap V2 and V3.
- Public and private flow are not yet separated.
- Direct payments to the fee recipient are not yet measured; Xatu trace partitions are empty for these days.
- Sixteen sampled days; no continuous run yet.
- About 2% of October blocks have no builder label.
Next
- Split public and private flow with Xatu mempool data and Mempool Dumpster, calibrated on locally built blocks.
- Direct payments per block, from the fee recipient’s balance change.
- Zero-tip front transactions by builder and type: which of Titan’s flows fell.
- A 30-day run including high-volatility days.
- Uniswap V4, Curve, Balancer and the main aggregators; token flow to separate arbitrage from routed swaps.
Comments, corrections and pointers to related measurements are welcome: danguo01@gmail.com.