I've published a new Mempool Observation today.

It's about the daily withdrawal broadcast of BitMEX. They broadcast multiple megabytes in around two minutes every day.

This has an effect on the whole network: The feerate estimator's estimates, and, as a reaction, the observed average feerate spikes. The minimum feerate for block inclusion is higher for a few hours and the time-to-confirmation sees an increase.

All because their users deposit directly into 3-of-4 P2SH multisig (redeem scripts with four uncompressed public keys, it's 2020...). That's more than 500 bytes per input. I discuss a few ways of reducing the transaction count and size. Most promising seems to use Schnoor + Taproot combined with output batching. With that it could even be possible for them to keep their '3BMex' vanity addresses by iterating a leaf of the merkle tree that's used as tweak until they get a Bech32 vanity they like. (Not saying that's a good idea: the privacy and fungibility benefits from Taproot would be gone.)

Full article is here:

The daily BitMEX broadcast at 13:08 UTC - Using a fingerprint to reason about a footprint - https://b10c.me/mempool-observations/2-bitmex-broadcast-13-utc/

Planning to write up my observations on seasonality next but haven't started yet. Seasonality seems to overlap with what gmaxwell mentioned in one of this issues (https://github.com/0xB10C/memo/issues/66).