I’ve been spending some time exploring this dataset. I made an initial tweet about it the other day. The dataset contains historical Flashbots transactions which found their way into blocks.
If you don’t know what Flashbots is, or MEV is, you will already be confused. Other people have done a much better job than I will describing these things. Google now…
This post is me feeling around the dataset, trying to understand. A disclaimer. I’m still very much trying to make sense and may have made some wrong assumptions/ conclusions. Please message me if you find something errant. I’d not be surprised if there was!
Data Source and Assumptions:
- I used the blocks.flashbots.net API to generate this dataset
- I limited the analysis to the last ~50 days, recent blocks since block # = 13602434 (Nov-12-2021 04:44:10 PM +UTC). There are 978,354 transactions in this subset.
- There are a few types of “bundles” {miner_payout, flashbots, rogue, unknown}. I’ve focussed on the “flashbots” bundles.
- Addresses for routers like Uniswap, Sushiswap, etc are excluded.
- This analysis focusses in on what I’ve decided is the #1 Flashbots bot. What is this “#1” based on?
#1 Flashbots Contract
As % of total transactions in the analysis period, the #1 contract is ~11% of the dataset (72,480 transactions for this one contract).
At any time on etherscan you can see there are many pending transactions for this contract and many historical transactions (> 2.6 million presently).
It looks like Tuesday 22 June was their highest transaction count in a single day
How about the EOAs calling this contract?
There appear to be 210 EOAs that call this contract. Many of these EOAs generate a large number of transactions. They must have a detailed access control scheme.
What else can we find that these EOAs touch?
We can see what else those EOAs are sending to? First, we find that there is another top Flashbots contract in this “EOA web”. I had labelled this other contract “#25” by my ranking. Also note that Etherscan has this contract tagged as "Fund" and labelled “Fund: 0x979...383”. (Perhaps there is some other reason for this, and these are unaffiliated? I don’t know why that would be…)
What if you search for transactions to accounts that are not between this web of affiliated EOAs, and not to the 2 bot contracts, what do you find? There are only 13 addresses that remain and a few are interesting.
- Another EOA that triggers that #25 bot above. This EOA is also interesting because it has inbound transfers from another account I’ve tagged based on data in this dataset. I’d labelled this other account in the From column as “fbots top 20 EOA feeder account”. This account seems to fund the EOAs that invoke these bot contracts.
- Completing the circle, you can see this account transfer into another account, which feeds … our #1 MEV bot.
- Another EOA triggers yet-another top 50 MEV bot (again, ranking based on this dataset). That makes 3 bots. This one ranks #15 in the recent period FWIW.
- This related account appears to be the transfer account to Avalanche Bridge (again my labels visible).
I tried to get a sane graph for the interconnected web and ran out of patience here. The green arrows are egress from the feeder account. The blue arrows are invocations from the EOAs to the main bot account. (This filters out some accounts with low transaction counts.). There exists a better graph, but not today graphviz…
Finally
- Humility - maybe all this is nonsense and I’m finding noise/ misunderstanding something.
- I find the openness of the data fun and inspiring. Quite a contrast to tradfi!
- There will be more data explore to come …