Every Ethereum slot is a lap. We time how fast each block reaches the world, rank the regions, and race them against a what-if car: an estimate of the same block with mump2p.
Starting from this exact moment 24 hours ago, real blocks travel around the world again one after another, at their real speed. Each dot is a Xatu sentry lighting up the moment it actually saw the block; the quiet seconds between blocks are skipped.
What Optimum measures on mainnet with its own gateways. Self-reported.
Each lap is one hour of the day. Cars finish when the block reaches that region (hourly median), slowed down so you can watch. The what-if car is the last-placed region with mump2p (estimate).
The full day, slot by slot.
Not a measurement. Dark grey is the time until the first sentry sees the block, which depends mostly on when the proposer releases it. Silver is the extra time the network needs to carry it further. The estimate divides only that second part by —, the network spread factor Optimum reports for mump2p versus libp2p (Optimum Hub, self-reported). The first-seen time is kept as it is, which is our modelling choice. The one independent mainnet study so far (StereumLabs, Sep 2026) measured blocks arriving up to about 120 ms earlier on most consensus clients (gateway v1.1.1, 36 nodes in one datacenter); the authors note the design could not reach statistical significance and call for a rerun on a newer gateway version. The estimate here is conservative: Optimum's own data also shows earlier first arrivals, which it leaves out.
Ready to post. The card and text update with each new day.
For each slot we take the canonical block and the moment every Xatu sentry first saw it on gossip, in milliseconds after the slot began. A slot's time is the median across sentries. Regions are the sentries' continents.
Purple marks the fastest 5% of the day's slots, green is faster than the day's median, yellow is slower. Orange means the median arrived after the attestation deadline (a third of the slot). DNF is a slot with no canonical block.
Xatu sentries are a sample of the network, not every node. Data is published daily, so this page replays the most recent full day. The what-if car applies Optimum's self-reported spread factor and is an estimate. Xatu sentries are not a clean no-mump2p baseline, because accelerated blocks are re-gossiped into the normal network.