Is
2.4%a typo? because it sounds like pretty small improvement.
This is a lower bound based on preliminary analysis, we expect actual numbers to be significantly higher but only wanted to present numbers with high confidence.
2 Likes
Yes, this is likely to be the case. People are likely to be willing to pay more for transaction ordering than for transaction inclusion.
4 Likes
Is there an actual paper about this? With actual details of the transaction bundle specifications, how the sealed-bid auction is actually held, how searchers and miners find each other, etc.
There is no paper about this yet. We plan on publishing papers in the next few months as part of the research roadmap as we collect feedback from the community.
You can consider this post + the reference implementation of MEV-Geth to be the canonical specification. Happy to clarify any components if unclear.
There is no discovery process for miners and searchers at the moment, we expect this capability to develop over time.
4 Likes
Love the MEV inspect dashboard. Do you plan to host a publicly available version for anyone to access?
Regarding MEV-Geth, could miners “steal” the MEV from searchers? It seems like searchers do the hard work of identifying the MEV and miners could (simplified) just insert their own address to reap the rewards.
1 Like
Yes - we are currently working on infrastructure so that we can host the dashboard publicly. In the meantime, the data can be replicated by running the mev-inspect service.
This question has come up a couple times on our discord too, I will consider updating the post with additional clarification on the game theory if this answer proves to be adequate.
You are right that in the mev-geth proof of concept, miners have full visibility of the bundles as they come in. Miners could, if they wanted to, steal the most profitable bundle for themselves. This is why Complete privacy has a big 
next to it.
Here is the game theory argument why they will not:
- it is possible for a searcher to identify if a miner is stealing his bundles by monitoring the chain
- it is possible for the searcher to withhold bundles from miners who he finds to be miss-behaving
- mev-geth bundle auctions are a repeated game between miners and searchers
- miners are incentivized to keep the game going in order to maximize long run profits
In addition, stealing bundles is a non-trivial engineering effort which I find unlikely miners will be interested in doing. It would require modifying mev-geth, implementing logic for generalized frontrunning, and taking on additional exchange risk from receiving payments in tokens rather than ether.
In summary, miners would only be incentivized to steal bundles if the revenue from that bundle outweighs the total expected future revenue from continuing the relationship with the searcher AND the risk adjusted cost of execution. I don’t expect this to be a sufficiently enticing proposition.
I expect miners who participate in the Flashbots proof of concept to use their reputation as a signal of trustworthiness, especially as some of them prepare to become staking pools for ETH2.
In the longer run, we hope to transition to a fully trustless solution as described in this section:
3 Likes
Fascinating work, arguably one of most important topics in the space.
Question around mechanisms on the miner side: rather then reputation, which is retrospective (“miner X is known to front-run bc they already did so Y times”), would it be possible for miners to signal they are running a “good” auction mechanism for ordering transactions? Is that where you’re going with this?
I’ll ping @fiiiu who is coordinating the research around this question https://github.com/flashbots/mev-research/blob/aa07a7f40a2b20d8a545132303553c437364b5b5/FRPs/FRP-1.md
@lauren great question!
Just to clarify, by “good” here we were aiming at something more “mechanistic” if that makes sense (think “efficient”, “fast to compute”, etc.–defining it is a valid research question!)… “Good” in the ethical sense is deferred to paper 2.
The plan is to incorporate the auction mechanism in mev-geth, as we now have the sealed-bid, first price one in the PoC. Miners opting in to mev-geth would run this by default, as miners today order txs by price.
Miners of course could still provide an endpoint for searchers to send transaction bundles, but prioritize them in different ways. I don’t think there would be a way to prove future behavior, since they could change the code they run at any point. There could perhaps exist “harder” mechanisms than purely reputational ones (somehow publishing/staking for a hash of the code they are running? I don’t know), but I don’t think they’d bypass the problem above…
If you’re interested on these topics please comment on the mev-research repository, or join our weekly research calls! 
4 Likes
Another interesting question:
does this paper answer the question on how dishonest the miners are (how many miners are frontrunners themselves)?
As an example, if you see a front-running transaction with SMALLER gas price included in front of the original transaction with LARGER gas price, this would for almost sure prove dishonesty of the miner, would not it ?
1 Like
Within the objectives of illuminating the dark forest, we aim to provide transparency to what is happening in the mempool and on chain. This includes how transactions are ordered within blocks and how they are propagated through the p2p network. I would say that effort is independent from ethical judgement of honesty or dishonesty. The ethical questions are wide ranging and deserve to be discussed on their own once the data is available.
Well :)))
I agree - I should not have used the “dishonest” word - the guys are just making money.
Let me rephrase :))
Is it possible to measure what percentage of mined blocks order transactions specifically to maximize frontrunning profits (from which you would deduce that miners cooperate with front runners)?
Here are a couple metrics that I think are interesting and reasonably easy to obtain:
- % of blocks that have transactions out of order (excluding miner payouts)
- % of transactions that are included on chain but never observed in public transaction pool (ie: % of transactions that go through dark pools)
Do you have any other in mind ?
cc @ao98 who is looking at metrics right now
2 Likes
I think these two are really good!
BTW, the problem may(?) become worse for ETH2, since for ETH2 the proposer knows way in advance it will propose the block, while a miner does not know for sure whether the block will win or not.
3 Likes
It’s not clear yet exactly how the dynamics change with ETH2. I expect the change in WHO proposes blocks will matter more than HOW. The lower barrier to entry for new entities like exchanges or hedgefunds to become significant ETH2 block producers may make it much easier for highly specialized actors to extract MEV in a selfish manner.
3 Likes
The “sealed bid” isn’t sealed for miners, who are improperly treated as distinct from seekers.
I agree miners remain a trusted party in the proof of concept design of mev-geth. I think searchers need to expect miners will try to use their preferential information in a selfish way and prepare accordingly.
What kind of actions are you worried miners will take?
How do people feel about the existence of private mempools like taichi.network? If you had to take a position, would you support or oppose them?
On one hand, they feel a bit shady to me, perhaps because they move toward a two-tier network where some people have an unfair advantage over others. One might counter that the service is available to anyone who is willing to pay for it, but it still doesn’t feel very good.
On the other hand, they provide a valuable service by solving a problem (frontrunning) for users.
1 Like
Anyone can send private tx via taichi.network, it’s free to use, bro. You do not need to pay for it.
1 Like
Ah, that is good. Still, the vast majority of users are not technically sophisticated enough to use it.