Timeboost Auction Analytics Engine
Hi all,
I’ve built and deployed an analytics engine for Arbitrum Timeboost. It’s live, kept current within minutes of head, and Apache-2.0.
- Dashboard: https://timeboost.app
- REST API Reference: https://api.timeboost.app/openapi.json
- REST API Interactive Docs: https://api.timeboost.app/docs
- Questbook Link: Timeboost Auction SDK & Explorer
- Repo: GitHub
What it is
An analytics platform that ingests every relevant Timeboost data source, the full S3 bid archive, on-chain auction events and express-lane receipts, DEX swaps, and 1-second CEX prices, into one queryable store, and serves it through a dashboard, a documented REST API, and bulk dataset exports.
The purpose is to make the microstructure of the express-lane auction inspectable: not just how much revenue the DAO earned, but who bid what and when, which controllers actually execute vs revert, where the CEX↔DEX spread that funds the auction actually sits, and what a given bidding strategy would have earned over real history.
What’s indexed today (70M+ rows indexed)
- 4.43M bids — every bid the Auctioneer has published since launch, not just winners
- 551,822 auction rounds + 31.6M express-lane tx receipts on Arbitrum One
- 4.7M DEX swaps across 8 WETH-paired pools (Uniswap V3, Camelot V3, Sushi V3)
- 30.9M 1-second Binance klines for CEX↔DEX spread and markout analysis
- 14k reserve-pricer snapshots + all on-chain reserve price changes
Every source is continuously forward-filled. ClickHouse handles analytics, Postgres holds job state, and all cursors persist so the stack is restart-safe.
What it adds on top of the Entropy Advisors Dune dashboard
The Entropy Advisors Dune dashboard covers the top-line view — revenue, winners, daily aggregates. This tool is built to go a layer deeper:
- Every bid, not just winners — the full pre-close archive with bidder, amount, and second-level timestamp
- Auction-by-auction, block-by-block — every express-lane transaction joined to its round and controller, with revert status and in-block position
- CEX↔DEX spread at per-block resolution — 1-second Binance prices aligned to on-chain DEX state, rather than daily snapshots
- Backtesting — replay any bidding strategy over the entire round history and compare P&L against a naive baseline
- Bulk + programmatic access — a full REST API and Parquet exports, so the underlying data is queryable and forkable, not only viewable
In the dashboard
-
Auctions — round-by-round browser, per-round bidder list, controller stats
-
Health — revenue, HHI, competition, revert rates, per-source freshness
-
Pricing — live bid recommendations across 4 models (recent-performance, competitor-behavior, market-condition, floor-estimate) with a reasoning trace
-
Opportunity — CEX↔DEX spread heatmap, 6-horizon markout scatter, controller latency leaderboard
-
Simulator — backtest any strategy over the full round history, with a cumulative P&L curve
The REST API mirrors everything in the UI and adds Parquet exports (auction_bids, auction_results, swap_markouts, cex_dex_spread_per_block, execution_quality_daily) for bulk research.
What’s Next
When I started this, the goal was to make Timeboost participation less of a black box. The hardest part turned out to be assembling one coherent picture from sources that were never meant to sit together (the S3 bid archive, on-chain receipts, DEX swaps, and CEX prices), and that’s exactly what makes the result durable.
Right as I wrapped up, Offchain Labs posted the Constitutional AIP proposing a move from Timeboost to Priority Gas Auctions (PGA) (link). A change like this is exactly the kind of moment where this tooling earns its value: any transition in ordering policy is ultimately judged against the one it replaces, and that judgment needs hard data on both sides. This engine provides the Timeboost side of that comparison, giving the DAO an empirical foundation to assess whether the new policy delivers on what it promises.
How it can be used as the DAO weighs the transition:
- A lasting record. It preserves the full microstructure of the entire Timeboost era, every bid, round, receipt, and execution outcome since April 2025. The AIP that adopted Timeboost noted it “can only truly be evaluated with real-world data”; this is that dataset, queryable and Apache-2.0.
- A baseline for the comparison. Concentration (HHI), competition, revert rates, and DAO revenue are exactly the dimensions the PGA discussion turns on. Preserving this baseline gives delegates and researchers a concrete before/after reference if the policy changes.
- Mechanism-agnostic intelligence. Around 94% of express-lane value comes from CEX↔DEX arbitrage. That economics persists across ordering policies, so the spread reconstruction, markout, and latency analysis describe the opportunity surface itself.
- Immediate input to the discussion. The historical bid book, concentration trends, and revert decomposition are available today to any delegate, researcher, or aligned entity weighing the path forward.
The instance stays current and available, and the dataset stands as a durable archive of this era. Everything remains Apache-2.0, self-hostable, and configuration-driven by design, built so the methodology can evolve alongside whatever ordering policy the DAO adopts.
Happy to answer any questions in the thread.