Problem Description
Ethereum block intervals are generally described as 12 seconds long, but the peer-to-peer nature of the network and the prevalence of timing games distorts block intervals in practice. For a subset of transactions this causes them to be included in blocks with timestamps from before the transactions are sent.
We want to analyze false-negative error rates in traditional measurements around censored transactions or order flow auction (OFA) inclusion times, where it may at first appear that transactions were intentionally excluded, but in reality the block was constructed before the transaction was seen by the builder. A more advanced set of insights could include trying to decode and geographically footprint the internal pipelines of block builders, i.e., is there a public mempool ingression that is closer to one set of instances from a block builder or do all instances have their own ingressions?
The distribution of transaction inclusion times adjusted for the winning bid arrival time of the blocks. We see the disappearance of the head of the profile, with comparatively few transactions being seen by the Mempool Dumpster nodes after the winning bid time (node coverage is imperfect).
Related Notes
‣
‣
