# The Geography of (Concurrent) Block Proposing

Crypto.world

> A blockchain's resilience and censorship resistance depend on where its validators and proposers physically sit: operators in the same place share electricity, communications infrastructure, disaster…

**arXiv**: 2610.18812

**Published**: 2026-10-13

**Authors**: Ciamac C. Moallemi, Mallesh M. Pai, Dan Robinson, Christoph Schlegel

## AI summary

## Overview

**Research area:** Blockchain protocol design and mechanism design — specifically computer science and game theory (arXiv:2610.18812v1 [cs.GT], 13 October 2026), sitting at the intersection of market microstructure, auction theory, and the economics of validator/proposer geography.

**Technical level:** Advanced. The paper builds on first-price common-value auction theory (asymmetric signals, quantile constructions), Brownian information models, potential games, and weighted-median facility location. The intuition is accessible, but the formal machinery is heavy.

**Scope:** An equilibrium analysis of where block proposers choose to physically locate, first with a single proposer and then with multiple concurrent proposers competing for the same financially valuable opportunity, by the authors Ciamac C. Moallemi (Columbia University and Paradigm), Mallesh M. Pai (Rice University and Tempo Labs), Dan Robinson, and Christoph Schlegel (Flashbots).

## What This Paper Is About

A blockchain's resilience and censorship resistance depend on where its validators and proposers physically sit: operators in the same place share electricity, communications infrastructure, disaster exposure, and the same regulator. The paper asks whether the economic incentives of proposers push them to cluster around the most valuable sources of information, or to spread out — and whether adding more concurrent proposers changes that outcome.

To answer this, the authors make location an endogenous choice inside a two-stage game: proposers first pick where to operate, then compete in a first-price auction for an opportunity whose value depends on information that originates at geographically dispersed sources and arrives at different locations with different delays.

## Key Contributions

1. **Rents come from exclusive information.** The auction has a closed-form equilibrium (Theorem 2, with Corollary 1 for additive values) in which each proposer's expected profit is proportional to the standard deviation of the information it observes before its rival. Information both proposers see equally early earns no rent.
2. **A monopolist concentrates; competitors separate.** A single proposer locates at a variance-weighted median of the information sources (Theorem 1). With two proposers, the jointly unobserved variance is an ordinal potential, so equilibria exist and global solutions of the associated weighted two-median problem are equilibria (Theorem 3). Co-location requires one location that minimizes every source's delay (Corollary 2); with two sources, one proposer sits at each, however weak one source is (Proposition 3).
3. **More proposers produce distinct locations.** With at least as many sources as proposers, equilibria under "regular bidding" have pairwise distinct locations, though they may be close together (Theorem 5); the paper states that existence of such equilibria is open.
4. **Protocol design and price informativeness matter.** Location-dependent deadlines enter as effective delays (Proposition 6), and a single fixed auctioneer makes co-location an equilibrium (Proposition 7). Competition makes the auction outcome more informative about the value than an optimally located monopolist (Theorem 7), but equilibrium locations need not be the most informative ones (Proposition 9).

## Main Findings

- **The monopolist minimizes variance-weighted delay.** Writing the variance-weighted delay as D(x) = Σ_k w_k d_k(x) and the observed variance as I(x) = TW − D(x), the monopolist's profit E[(μ(x) − L)*+] is strictly increasing in I(x), so optimal locations are exactly the minimizers of D(x). With Euclidean delays these are variance-weighted geometric medians; if one source has w_j ≥ Σ*{i≠j} w_i, co-locating with that source is optimal. On a Euclidean segment with two sources of unequal variance, the proposer locates at the higher-variance source; with equal variances, every feasible point between them is optimal.
- **Competition monetizes exclusive information only.** With two proposers, the terminal value decomposes into a common component C, exclusive increments X_A and X_B, and a residual ε neither observes, with variances Γ, Ω_A, Ω_B, Ω_0 (Equation 12–13) summing to TW. The selected no-reserve equilibrium bid is b_i = C − (s_A + s_B)λ(X_i/s_i) with s_i = √Ω_i, and ex ante profit is Π_i = s_i/√π. A larger exclusive variance raises rent, not winning probability: at nondegenerate profiles each proposer still wins with probability 1/2.
- **The profit formula is distribution-free in shape but not in independence.** With additive values, profit equals one half the expected absolute difference of two independent draws of the bidder's own signal, π_i(c) = ½ E|X_i − X_i′|. The paper notes the formula does not require normality: if X_i is uniform on −a_i < x < a_i, the bid is c + (a_A + a_B)(q − 1) and profit is a_i/3. Independence is substantive — with correlated signals the construction need not be an equilibrium.
- **Location choice is a potential game.** Since Ω_A = D(x_B) − Ω_0 and Ω_B = D(x_A) − Ω_0, a proposer's exclusive variance rises exactly when the jointly unobserved variance Ω_0 falls. Hence −Ω_0 is an ordinal potential (an exact potential for the variance payoffs), pure location equilibria exist, and every global minimizer of min Σ_k w_k min{d_k(x_A), d_k(x_B)} is an equilibrium — but the equilibrium set can also contain coordinatewise minima that are not global.
- **Co-location is non-generic.** A co-located profile is an equilibrium if and only if that location minimizes the delay to every information source. With distance delays and at least two distinct source sites, no such single location exists, so equal locations cannot be a two-proposer equilibrium — regardless of how unimportant one source is.
- **Two sources on a line: one proposer at each end.** On 𝒳 = [0,1] with sources at 0 and 1, positive weights w_0 and w_1, and distance delays, the unique pure location equilibrium (up to swapping proposers) is (x_A, x_B) = (0, 1). For more sources on a line, a global solution is found by minimizing J(1, j) + J(j+1, K) over splits, where J(r, t) is the minimum weighted absolute-distance cost for a consecutive group.
- **Separation is paid for by the seller.** Revenue is V_0 − (s_A + s_B)/√π, so it is highest at V_0 when proposers see the same information — i.e., whoever collects auction revenue prefers the concentrated geography that proposer competition works against.
- **A finite reserve changes entry.** With a reserve L, one proposer can enter earlier and pool at the reserve while the other abstains; the paper defines entry thresholds q_R and p_R and gives the resulting bid β_L(q; c) = (q_R L + ∫_{q_R}^{q} h(s) ds)/q. Simply truncating the no-reserve bid at L is not in general an equilibrium. The baseline analysis uses the unrestricted no-reserve auction because a monopolist has no finite optimal bid without a lower bound.
- **Design levers.** A location-dependent deadline acts as an effective delay (Proposition 6), and a single fixed auctioneer restores co-location as an equilibrium (Proposition 7) — a route to the concentrated geography that a revenue collector would prefer.

## Methodology in Plain English

The authors build a stylized two-stage model. In stage one, each of M proposers picks a location in a compact metric space; each of K information sources sits at a fixed site, and information from a source reaches a given location with a known, continuous delay bounded by the settlement time T. In stage two, proposers see the information that has arrived by the deadline and bid in a first-price auction for a right with a common terminal value.

The value is modeled as a public starting value plus a sum of independent Brownian innovations, one per source, weighted by the source's variance w_k. Because Brownian increments over disjoint time intervals are independent, the authors can decompose the terminal value exactly into four independent pieces: what both proposers know (common), what only A knows ahead of B (exclusive to A), what only B knows ahead of A (exclusive to B), and what neither has yet seen. The variance of a proposer's exclusive piece becomes the single number that drives its rent.

For the auction stage, they use a quantile-based bidding construction from the asymmetric first-price common-value auction literature (credited to Cheng and Tan), and verify deviations rather than just first-order conditions. Specializing to Gaussian signals gives closed-form bids and rents. Because each proposer's profit is a monotone (square-root) transform of its exclusive variance, and because exclusive variances move one-for-one against the jointly unobserved variance, the location stage becomes a potential game. That lets the authors characterize equilibria using weighted-median and two-median facility-location problems, then check specific geometries (points on a line, distance metrics) and extensions (finite reserves, location-dependent deadlines, a fixed auctioneer).

## Why This Matters

**Impact on research.** The paper supplies a micro-founded link between auction rents and geography, complementing reduced-form and simulation-based work on validator positioning (Yang et al.) and market-share-based analyses of concurrent proposers (Babel, Milionis, and Xue). It also contributes a new use of potential games: because rents depend on the square root of exclusive variance, the residual (jointly unobserved) variance becomes an ordinal rather than exact potential, turning proposer location into a two-median coverage problem. The result that competition pushes proposers apart is a counterweight to the usual concern that competition only limits any single proposer's control over inclusion.

**Real-world applications:**

- Siting of proposer, builder, and searcher infrastructure relative to trading venues and other price sources, and the resulting latency advantages in on-chain order flow auctions.
- Design of multiple concurrent proposer protocols (motivated by Fox, Pai, and Resnick; Garimidi, Neu, and Resnick; and Cadence), where the number of proposers is a protocol-design choice with geographic side effects.
- Placement of attesters and other intermediaries, which enters the model as a location-dependent bidding deadline.
- Jurisdictional and infrastructure-risk policy for staking operators, including proposals to explicitly reward geographic diversity (Roeschlin et al.) and to verify location claims (Milionis et al.).

**Industry relevance.** The analysis implies that nominal openness of participation can coexist with strong pressure to operate in a narrow set of locations, and that simply adding participants need not spread them out — a direct concern for teams designing proposer competition, for staking providers choosing data-center regions, and for anyone who collects auction revenue and therefore benefits from concentrated geography. The observation that a single fixed auctioneer restores co-location is a concrete design warning.

## Future Directions

- **Existence for more than two proposers.** The paper proves that equilibria with regular bidding have distinct locations when there are at least as many sources as proposers (Theorem 5), but explicitly states that existence of such equilibria is open.
- **Which equilibria are selected.** The location game can have coordinatewise minima that are not global minimizers of the two-median problem; characterizing the full equilibrium set, and refining among equilibria, is left unresolved.
- **Richer continuation and reserve structures.** Payoffs depend on the selected auction equilibrium (the monotone zero-lowest-quantile-rent equilibrium); other unrestricted equilibria can support co-location, and the finite-reserve location game with entry thresholds q_R and p_R is only partly analyzed.
- **Beyond the model's abstractions.** The authors abstract from consensus, from network topology beyond pairwise delays, from collusion among proposers, and from other block-building tasks; endogenizing attester placement, delay technology, and geography-based rewards are natural extensions.

## Target Audience

Protocol designers and researchers working on proposer-builder separation, multiple concurrent proposers, and validator decentralization; mechanism-design and market-microstructure economists interested in how information geography shapes auction rents; and quantitatively sophisticated members of staking and block-building teams who need to reason about where infrastructure should sit. Readers without a game-theory background can follow the framing and results but will need the formal sections to engage with the proofs.

## Authors’ abstract

The resilience and censorship resistance of a blockchain depend on where the validators and proposers of that blockchain physically operate. We study the economic incentives behind this location choice, both with a single proposer and with multiple concurrent proposers. Formally, we study a game where proposers first choose locations and then compete in a first-price auction for an opportunity whose common value is driven by information from geographically dispersed sources.…

[The full abstract on arXiv](https://arxiv.org/abs/2610.18812)

[Read the original paper](https://arxiv.org/abs/2610.18812)

---

Canonical: https://crypto.world/paper/the-geography-of-concurrent-block-proposing-2610.18812
