☄️

Research Programme: Sharing orderflow in PBS

A programme to study the causes, prevalence, downstream effects, and control of exclusive orderflow and orderflow sharing in the blockspace supply chain of Ethereum and based rollups.

Introduction

Currently, the Ethereum block builder market is oligopolistic, being overwhelmingly dominated by two entities: Titan Builder and beaverbuild. Their foothold in the market is thought to primarily arise from exclusive arrangements with order flow sources (OFSs) such as vertically integrated trading operations or third party Telegram bots.

It's generally assumed that significant concentration in the builder market poses risks to the objectives of the Ethereum blockchain, such as censorship-resistance or "neutrality."

In this programme, we will scrutinise these assumptions: How much concentration is too much? To what extent is concentration really driven by EOF? How prevalent is EOF, and why? And what can be done about it?

Track A: Drivers of EOF

The need for exclusive order flow

In short, we expect that builders must obtain exclusive transaction items because this the main component of their profit margins in the PBS auction.

The profit a winning builder derives from a single PBS auction is the difference between the fee income from the block and the amount they pay to the proposer (less "offchain costs," which we discount for now). Theoretically, in equilibrium this should be the difference between the fee income for the winning block and that of the next best block produced by another builder. That is, the builder's profit is exactly his edge -- the value he was able to aggregate that no other builder did. This is fundamental to the market structure of a single item auction, and cannot be alleviated by changing the details of the bidding mechanism (Milgrom, 2004).

There are other factors that could form part of a builder's edge, but these are thought to be minor:

  • Faster optimal packing of public tx material
  • Lower latency; effectively means late arriving transactions are exclusive, but without a special agreement
  • Fees for additional differentiated services. In practice such services are usually provided for free in order to attract EOF.

Questions.

  1. How does edge relate to profitability in other market structures such as multiple proposers or builder cartels?

What must EOF arrangements look like?

In its simplest form, an exclusive arrangement with a builder does not look very attractive to an OFS: since the partner builder cannot be sure to win the target slot, exclusivity incurs a risk of the transaction not hitting its target block that could otherwise be mitigated by multiplexing (muxing), that is, sending the transaction to multiple builders.

So why are OFS doing it? There must be some kind of sweetener involved. We will attempt to categorise the different approaches.

  • Rebates/subsidy. The vanilla bundle API of a builder handles fees via the same pay-as-bid mechanism as the public mempool. As the fee recipient, the builder could offer a rebate, effectively adjusting the bidding experience to one that's easier to manage. The mechanism offered by Flashbots Buildernet is a prototypical example.
  • Differentiated services, such as positioning, only available to exclusive partners.
  • Derisking by builder-side muxing. The muxing builder can still derive some edge from being the builder partner by requesting a rebate from the downstream builders by adjusting the refundPercent field of the eth_sendBundle API. In this case, the arrangement is not really exclusive, so we introduce the term priority order flow to describe it.

Example. Flashbots Protect provides a few extras to attract exclusive/priority OF:

Questions.

  1. How can we parametrise the set of deals so described?
  2. Is there some way to reduce the space of all possible deals to a manageable set of parameters? Do we need any more basic approaches than the ones listed here to capture it?
  3. What are the feasibility conditions for such deals? What exogenous signals drive these boundaries?
  4. Under what conditions are deals so rare that EOF or POF becomes a less significant driver of builder edge than other dimensions? Under what conditions are they so abundant that even builders with small market share may acquire them?

Measuring EOF

How can we measure the extent of exclusive order flow and the extent to which builder profits depend on it?

Two previous works purport to measure the extent of EOF. (Yang et al., 2024) defines a notion of pivotal provider for a builder, meaning one that is pivotal for winning blocks in the sense that the block would not win if the OF were removed. They identify major OF providers by common name or Ethereum address, and use the data to posit exclusive relationships. (The authors acknowledge this approach misses providers whose addresses are not known, e.g. BloXroute.)

Meanwhile, (Öz et al., 2024) defines exclusive signal as transactions not appearing in the public mempool or appearing as part of an "OFA bundle" (which mainly includes MEV-Blocker and MEV-Share bundles). Note that it isn't automatic that such signal is exclusive in the sense we mean here, i.e. that it is known to only one builder. Nonetheless, by the heuristics identified in (Yang et al., 2024), much of it probably is.

Questions:

  • How close are the actual builder profits from the value delta between the best and second best blocks?
  • Can we distinguish exclusive from non-exclusive transactions using publicly available data?
  • Among those, can we distinguish those that are exclusive only by dint of a latency advantage from those whose exclusivity derives from a business arrangement?
  • Assuming we can distinguish exclusive flow, can we quantify how dependent the builder is on that flow remaining exclusive?

Track B: Impacts of EOF

In the vein of (Bahrani et al., 2024), let us scrutinise the assumption that excessive builder concentration is bad.

  • What are the potential negative impacts of builder market share concentration?
  • Of these, which are actually happening now? Can we forecast a timeline of anything happening?
  • How can concentration be measured?
  • To what extent is concentration driven by barriers to entry such as EOF? How much of this is EOF as opposed to other types of barrier?
  • Can we quantify the "amount" of EOF needed to achieve a builder market concentration below a given threshold?

Market power

By "market power" we mean the ability to take unilateral actions that non-negligibly affect the distribution of ultimate allocation or prices of blockspace.

Example. Within the scope of a given slot, the proposer has market power because they can choose to reject blocks, causing the block not to be allocated.

More interestingly, a builder with a foothold (reputation or exclusive deals) may have market power because they can achieve a high probability of winning by including just a subset of their available flows (including all exclusive flow), allowing them flexibility on the allocation of the remaining space.

What might the builder with market power choose to do with such advantage?

  • Adopt a monopoly pricing/allocation strategy, potentially resulting in lower overall welfare. This may also include basefee manipulation strategies.
  • Censorship by direct contracting, especially with existing EOF sources (e.g. censor opposing bids in an onchain auction).

Risk

A concentrated builder market leads to concentrated risk exposure, and exposure to correlated risk preferences (e.g. OFAC compliance), of the blockspace pipeline.

A diversified blockspace market ought to be more "resilient."

Fragmentation

Exclusivity means that builders can't share OF, hence it's difficult to get high confidence scheduling guarantees from the builder layer.

What about if there were no EOF at all? Would that be good?

Something about how no footholds => everything flows to the most efficient player => return to monopolisation and concentrated risks.

Track C: Controlling EOF

Suppose that EOF and builder concentration is a problem. We can monitor it, but how do we set targets? What levers do ecosystem participants have at their disposal to achieve those targets?

Questions:

  • What are the objectives of Ethereum with regard to builder market concentration?
  • Does Ethereum "prefer" a certain amount or certain types of EOF?
  • What tools do Ethereum protocol designers have to influence builder market concentration and the drivers of EOF?
  • What out-of-protocol systems could be built that influence the drivers of EOF? Which are currently buing built? Which are likely to be built?

Mitigating dependence of builders on EOF

Changing the market structure of PBS by the addition of a buyer cartel or introducing proposer competition.

  • Introducing proposer competition can only be done in-protocol. One expected change in this direction is FOCIL.

Removing sources of EOF

By adjusting EOF CAC.

  • Idea 1: Increase CAC by making exclusivity less attractive to OFS. Remove the tools that builders have to attract EOF by introducing competing features.
  • Idea 2: Decrease CAC by making exclusive deals feasible for builders with lower market share or reputation. This basically needs some kind of derisking via muxing, or priority instead of exclusivity.

Project management

Timeline

WIP

Task categories

  • Gathering and labelling data.
    • Qualitative OF data. Exploring and labelling transaction and bundle structure. Block explorers.
    • Numeric data (bids, profits). Use of blockchain data gathering tools like Eth execution API (cast, cryo, ethers, web3.py etc.). Dune queries including Flashbots tables.
  • Data visualisation. Python notebooks.
  • Economic analysis.
    • Mechanism design: creating and analysing incentive models. Key example: families of exclusive OF and OF sharing mechanisms.
    • Framing concerns in terms of economic objectives, e.g. allocative efficiency.
  • Mathematical risk models. Risk measures.
    • OFS inclusion/execution risk preferences.
    • Risk concentration in Ethereum liveness.
  • Agent models.
    • Market segmentation and OFS/builder partnerships
  • Statistics, i.e. creating and fitting statistical models.
    • Inferring agent types from observable data.
    • Time series forecasting

Call to action

If you're interested in this subject and feel you have something to contribute, especially in the above categories, then get in touch!

Funding

We will seek funding from the PBS Foundation, Flashbots Research, Ethereum Foundation, and others.

Resources

Bibliography

  • Milgrom, P. R. (2004). Putting Auction Theory to Work. Cambridge University Press.
  • Öz, B., Sui, D., Thiery, T., & Matthes, F. (2024). Who Wins Ethereum Block Building Auctions and Why? In R. Böhme & L. Kiffer (Eds.), 6th Conference on Advances in Financial Technologies (AFT 2024) (Vol. 316, p. 22:1-22:25). Schloss Dagstuhl – Leibniz-Zentrum für Informatik. https://doi.org/10.4230/LIPIcs.AFT.2024.22
  • Yang, S., Nayak, K., & Zhang, F. (2024). Decentralization of Ethereum’s Builder Market (No. arXiv:2405.01329). arXiv. https://doi.org/10.48550/arXiv.2405.01329
  • Bahrani, M., Garimidi, P., & Roughgarden, T. (2024). Centralization in Block Building and Proposer-Builder Separation (No. arXiv:2401.12120). arXiv. https://doi.org/10.48550/arXiv.2401.12120

Dashboards