Relay Compliance and Efficacy

Manifold Insights 12.20.22

Some highlights from some new papers that have been published recently bring to light MEV Relays and their impact on the Ethereum ecosystem. You can find the paper sources at the bottom of this post.

Flashbots releases code to their SGX solution

You may remember us mentioning this back in february when Intel stopped supporting certain SGX chipsets or when Secret Network got pwned last month - These sorts of guarantees w.r.t. SGX are hard to make without the proper focus, and from our perspective Intel does not care that much about the problems SGX tries to address. For a more competent approach of enclaves look no further than Bedrock Systems

SoK: MEV Countermeasures: Theory and Practice

Note, this graphic contains incorrect datasets due to how they validate which relay producing which block. Manifold does not always insert an artificial tx at the end of the block like flashbots does, which is how they attribute blocks.

Wait time of ‘Sanctioned Transactions’

…We compare the waiting time [28] of sanctioned transactions and regular ones. The waiting time of a given transaction is the time between when it first appears in the mempool and when it is minded in a block. We plot the distribution of waiting time for regular transactions and sanctioned transactions.

The median waiting times of regular transactions and sanctioned transactions are 8.87s and 14.93s respectively, thus sanctioned transactions have to wait for about 68% longer on average than regular transactions before they can be included in a block

Compliance with OFAC Uberalles Planetary Indulgence

Interestingly, (this) shows that only Flashbots is fully OFAC compliant before Nov 8th, 2022. While the other three relays, bloXroute (Regulated), Blocknative, and Eden, who claim to be OFAC compliant, still proposed blocks containing sanctioned transactions. On Nov 8th, 2022, OFAC updated the sanctions list, but all of the OFAC-compliant relays failed to promptly adapt to the new list

A Flash(bot) in the Pan

FlashMission Accomplished?

Despite Flashbots’ efforts towards its goals, we must question how much progress it has actually made in achieving those goals. And if has made enough progress to be considered a success.

Illuminating MEV

The first goal (Goal 1: Illuminating MEV) is to increase transparency of MEV in the mempool. If Flashbots has spurred more users to engage with other private pools then transparency cannot be said to be sufficiently addressed. It is not clear that, indeed, Flashbots was the driving factor in users moving towards other private pools, but private pools have certainly replaced public MEV to a large extent (Section 6.2). Despite our efforts to contact the Flashbots project with the hope of measuring, directly, their pool of pending transactions, we were unable to get through to them. This is a strike against Flashbots as in this regard, it is no different than any other “dark”, private pool. Conversely, their measurement dashboard [17] has many interesting plots and datapoints, though it contains no direct analysis of Flashbots’ real world use cases. Another concern is that the methodology for generating these plots is not sufficiently open, making comparison difficult—if not impossible. Overall, there is no clear answer to whether Goal 1 has been achieved. Flashbots is much more open than other private pools, but does not seem to be an optimal solution in illuminating MEV behavior.

Democratizing MEV)

The second goal of Flashbots (Goal 2: Democratizing MEV) is similarly complex. Through Flashbots, more people have access to MEV-permitting infrastructure than ever before. However, this comes at the risk of taking on losses (Section 5.2), especially for non-miners—who already have fewer resources than their miner counterparts. This is because some degree of knowledge in writing smart contracts and understanding MEV opportunities is required to use Flashbots. This is beneficial for whoever can afford the effort to write (and verify) such contracts, but leaves non-experts in the same position as before. In fact, they are in a worse position, because now their are many more users who will be trying to frontrun them. Again, there is no clear answer to whether MEV is democratized, so we stick to the same refrain: more than before, but less than it could be.

:Distributing Benefits

The third goal of Flashbots is to distribute the benefits of MEV extraction (Goal 3: Distributing Benefits). On this point, we note that after an initially steep climb in usage, Flashbots has seen a decline in the number of blocks published (see Figure 3). If rational actors are declining in their usage of Flashbots, then it can only be assumed that they are not finding it as profitable as other options. This is unsurprising given that non-miners (a sizeable population) are seeing much less profit than they were before Flashbots (see Figure 8b). And so for those users no longer using Flashbots, the MEV problem has not been addressed. On this count, it is clear that Flashbots is not achieving its goal. Benefits are even more skewed than before.


There GitHub can be found here: GitHub - a-flashbot-in-the-pan/a-flashbot-in-the-pan: Tools for measuring MEV extraction through Flashbots (IMC 2022).

[1]Yang, S., Zhang, F., Huang, K., Chen, X., Yang, Y., and Zhu, F., “SoK: MEV Countermeasures: Theory and Practice”, arXiv e-prints, 2022.

[2]Weintraub, B., Ferreira Torres, C., Nita-Rotaru, C., and State, R., “A Flash(bot) in the Pan: Measuring Maximal Extractable Value in Private Pools”, arXiv e-prints, 2022.

Editors Note

We will start publishing a biweekly Insights post going forward.

End of Year round up and AMA probably on the 27/28 depending on what we hear back. Thanks!


What’s the reason for the bimodal distribution of wait times for non-sanctioned transactions, but unimodal for sanctioned ones?

1 Like

Thoughts on AWS Nitro? I know it’s still basically the same trust assumption (trust your OEM with the setup) but I have not seen a comparison of side channel exploits and benchmarking of past known incidents b/w SGX vs. Nitro.

This is fascinating data. Do all the other censoring relay operators basically following the same blacklist as the FB relay? Specifically ofac_blacklist.json from the FB block validation repo. Interesting also there’s an open PR to update it dating back to…11/4/22 :face_with_monocle:

What are the other big private mempools out there aside from Flashbots? I thought Blocknative and Bloxroute were both relatively quite small? Are you aware of any of the big validator pools (CEX or decentralized) operating these “dark pools”? Perhaps maybe also a consortium of the big crypto MMs (the ones that are still left standing anyway)?

Is it clear though that Flashbots enabled incremental MEV extraction or simply organized activity that otherwise would have occurred anyway in a (trusted) somewhat orderly marketplace? I don’t think it’s actually very clear if there’s a causal relationship there to be fair to Flashbots…

They have been actively helping new, independent relays (iirc both the Gnosis DAO one and bootstrap activity by forwarding blocks to them. Or are you saying even when accounting for that, their share has been going down? Isn’t it reasonable that their market share would go down given they were effectively at 90-100% of the MEV market (though I suppose you could measure it a number of ways, internal builders vs. their relay)?

1 Like