Timeboost Transaction Ordering Policy

• Timeboost Transaction Ordering Policy is a mechanism that combines transaction arrival time and explicit bidding to prioritize blockchain transactions.
• It integrates an express lane auction with a fixed delay for standard transactions, ensuring low-latency finality and improved revenue capture.
• Empirical studies on Arbitrum reveal reduced redundant submissions, enhanced DAO revenues, and challenges in managing volatility risks.

The Timeboost transaction ordering policy is a mechanism that assigns priority to blockchain transactions by combining both latency (arrival time) and explicit bidding, replacing pure first-come-first-served (FCFS) queuing with a score-based or auction-mediated system. Originating in the rollup context, notably on Arbitrum, Timeboost is designed to mitigate inefficient latency races, internalize maximal extractable value (MEV), and provide economically efficient, low-latency finality guarantees. Its paradigmatic instantiations mix real-time scoring rules with sealed-bid, second-price auctions for exclusive "express lane" sequencing rights.

1. Formal Specification and Architectural Components

Timeboost operates by maintaining parallel submission lanes—a latency-advantaged "Express Lane" (EL) and a standard "Normal Lane" (NL). At designated intervals, exclusive access to the EL is auctioned off, while all non-EL transactions are deterministically delayed before sequencing.

The core functional workflow is as follows (Ko et al., 29 Dec 2025, Messias et al., 26 Sep 2025):

• Intervalization: Time is partitioned into contiguous rounds of fixed length $T$ (e.g., 60 seconds).
• Express Lane Auction (ELA): Prior to each round, a sealed-bid, second-price auction allocates exclusive EL rights for the upcoming interval.
• Transaction Handling: EL-winner transactions bypass any artificial queuing (0 ms delay). NL and non-winner EL transactions are delayed by a fixed period $\Delta$ (e.g., 200 ms).
• Sequencer Implementation: Transaction metadata is authenticated and ordered via score or explicit controller mapping, with final settlement of auction payments on-chain.

A precise pseudocode for the ELA is:

input: round r, deadline D_r < T_r at D_r: collect bids {b_{ir}} determine: winner i* = argmax_i b_{ir}, price p* = second-highest bid from t = T_r to T_{r+1}: if EL tx from i*: forward immediately else: delay by Δ, then enqueue FIFO at end: winner i* pays p* to protocol

(Ko et al., 29 Dec 2025, Messias et al., 26 Sep 2025)

2. Scoring Functions and Theoretical Properties

The canonical Timeboost scoring paradigm is (Mamageishvili et al., 2023, Schlegel, 2023): $$S(t, b) = \pi(b) - t, \qquad \pi(b) = \frac{g b}{b + c}$$ Here, $t$ is the transaction arrival time, $b$ is the bid, $g$ is the finalization horizon, and $c$ is a normalization constant. Transactions are globally sorted by descending score, enabling both latency and bid-based priority in a monotonic, data-independent manner.

Key properties:

• Finalization guarantee: Every transaction tx arrives at time $t$ and finalizes by $t + (g - \pi(b)) \le t + g$. Maximum latency matches FCFS policies, but bids can accelerate finalization.
• Bid-latency fungibility: Bidders can substitute between waiting and bidding, with equilibrium outlay spread between the external latency race and internal protocol revenue (Mamageishvili et al., 2023).
• Economic independence: The highest-value actor wins priority; auction or bid-based payments transfer economic surplus to the sequencer or protocol.

3. Auction Mechanism Design and Equilibrium Analysis

Timeboost instantiates its express-lane allocation via an ahead-of-time, sealed-bid, second-price auction mechanism:

• Bidding: All participants submit private bids for express-lane rights for the upcoming round (Ko et al., 29 Dec 2025, Mamageishvili et al., 23 Nov 2025).
• Allocation: The highest bidder gains exclusive EL access; the price paid equals the second-highest bid. The protocol imposes a reserve price (e.g., 0.001 ETH).
• Incentive compatibility: Under private values, truth-telling is a weakly dominant strategy.
• Certainty-equivalent valuation: Agents with risk aversion discount their bids by the variance of expected profit,

$$v_i = E[\Pi_i] - \frac{\rho_i}{2} \operatorname{Var}(\Pi_i)$$

where $\Pi_i$ is arbitrage profit and $\rho_i > 0$ is a risk aversion parameter (Ko et al., 29 Dec 2025).

• Structural equilibrium bid (time-priority auctions under volatility): $$b_i = \alpha_i + \beta_i m^{IV}_i \sqrt{P} - \gamma_i v^{IV}_i P$$ where $m^{IV}_i = E[IV]$, $v^{IV}_i = \operatorname{Var}(IV)$, $\gamma_i = (\rho_i/2) \beta_i^2$, and $P$ is the current asset price.

Empirical estimations confirm strong positive sensitivity to expected volatility ($\theta_1 >0$), but large discounts for volatility forecast uncertainty ($\theta_2 < 0$); see subsequent empirical results (Ko et al., 29 Dec 2025, Mamageishvili et al., 23 Nov 2025).

4. Comparative Analysis with FCFS and Batch Auctions

The performance of Timeboost is benchmarked against two canonical alternatives: FCFS (pure latency race) and block (batch) bidding (Schlegel, 2023, Mamageishvili et al., 2023).

Policy	Score System	Finalization Bound	Expected User Outlay	Latency Waste
FCFS	S = –t	≤g (if bounded)	E[cost] = 1/2	Yes (external)
Pure Bidding (Block)	S = b in slot	=g (block duration)	E[bid] = 1/6	Low
Timeboost	S = π(b) – t	≤g – π(b) ≤ g	E[bid] = 1/6	None

Timeboost can achieve or even exceed the efficiency and revenue of block auctions for low marginal protocol cost ($c/g$), nearly eliminating the externality from latency investment (Mamageishvili et al., 2023, Schlegel, 2023).

5. Empirical Findings: Centralization, Revenue, MEV Capture, and Predictability

Observational data and empirical studies from Arbitrum highlight several outcomes and limitations (Messias et al., 26 Sep 2025, Mamageishvili et al., 23 Nov 2025, Ko et al., 29 Dec 2025):

• Highly centralized control: Two entities (Selini Capital and Wintermute) win ≈91% of express-lane rounds. A Herfindahl index ≈0.45 and Gini coefficient >0.7 indicate extreme concentration.
• Auction revenue and participation: DAO revenue ≈1090 ETH over 151,423 auctions (Apr–Jul 2025), but clearing prices and participation declined over time, e.g., mean clearing price fell from ≈0.015 ETH to <0.005 ETH.
• MEV opportunity timing: While express-lane access allows earlier inclusion (median position ≈0.46 in block), most profitable arbitrages cluster at block ends, limiting the value of early access.
• Spam and revert behavior: Timeboosted txs have a 21.75% overall revert rate (27.6% for Wintermute, 25.8% for Selini). Secondary-market reselling of rights proved unprofitable and led to excess revert rates and block stuffing.
• Bid–profit predictability: Per-minute Pearson correlation between winning bid and actual markout is weak (≈0.32 Wintermute, ≈0.26 Selini), but aggregate correlation rises over longer windows (to >0.8 at 30–60 min). Bidders appear to extrapolate from recent markouts rather than forecast precise minute-ahead returns (Mamageishvili et al., 23 Nov 2025).

6. Game-Theoretic and Welfare Effects: Link to Spam and DAO Revenue

A formal model shows that moving from pure FCFS to Timeboost reduces wasteful MEV-related spam by rerouting user incentives from redundant copy submission (with high revert rates) to explicit auction payments (Zhu, 10 Dec 2025):

• Equilibrium submission intensity decreases for both winners and losers; total redundant submissions drop strictly under Timeboost compared to FCFS.
• Sequencer/DAO revenue increases since revert fees (previously wasted to network) convert into transparent auction payments.
• Empirical difference-in-difference (DiD) estimates: A 0.7 SD decline in redundant copies and 0.7 SD increase in sequencer/DAO revenue post-Timeboost implementation, both relative to other layer-2s, significant at p < .01 (Zhu, 10 Dec 2025).

A caveat: while overall spam is reduced, express-lane winners may still overestimate their private advantage or suffer from sequencing race conditions, resulting in higher-than-expected revert rates for timeboosted transactions (Zhu, 10 Dec 2025).

7. Design Challenges, Limitations, and Future Directions

Several lessons and caveats emerge from the formal and empirical literature:

• Variance risk premium: All ahead-of-time auction formats impose a valuation discount for volatility forecasting uncertainty, especially in high-frequency environments (Ko et al., 29 Dec 2025).
• Allocative limitations: Sealed-bid second-price auctions deliver exclusivity but are revenue-inferior in low-participation regimes and reinforce capital-based centralization (Messias et al., 26 Sep 2025).
• Predictive frictions: The inherent difficulty of minute-ahead volatility forecasting undermines allocative efficiency at high frequency; bidders rely more on past realized markouts than on predictive analytics (Mamageishvili et al., 23 Nov 2025).
• Hybrid mechanisms: Policy recommendations include extending auction horizons (longer rounds), introducing dynamic reserves or subscription models, and integrating cryptographic order-fairness guarantees (Messias et al., 26 Sep 2025, Ko et al., 29 Dec 2025).
• Cross-domain extension: Express-lane frameworks could generalize to other rollups or permissionless sequencer architectures, especially when paired with verifiable or randomized ordering engines (Zhu, 10 Dec 2025).

In sum, Timeboost and its associated auction frameworks represent an overview of latency-based scoring, auction-theoretic mechanism design, and MEV-internalization, with significant trade-offs among decentralization, efficiency, and protocol revenue that must be addressed in next-generation transaction ordering protocols.