Why timeboost hurts the arbitrum MEV space.

This is a highly opinionated piece. If you disagree with some of my points or notice inconsistencies feel free to hit me up.

What is arbitrum and the sequencer?

Arbitrum is a Layer 2 blockchain which settles to Ethereum. Transactions on arbitrum are primarily submitted to a centralized sequencer, which then posts these transactions to Ethereum. I will first explain the pre-timeboost situation and then expand to what timeboost changed and why I think that it is hurting the MEV ecosystem on arbitrum. Arbitrum achieves a very fast blocktime of 250ms through ordering transactions purely sequentially, all transactions submitted to the sequencer are ordered after the FCFS (first-come-first-serve) principal. This makes it impossible for users to be frontrun/sandwich attacked. Therefore we are only gonna focus on MEV strategies which dont actively extract value from users. On arbitrage and liquidation opportunities the MEV game became a pure latency race. Whoever got their transaction included first won the opportunity for themselves.

How to win the latency race

Winning this latency race requires multiple optimizations and strategies, I won’t give away too much here but co-locating/having extremely good peering to the sequencer is a must in this space, as well as listening to the sequencer feed where applicable. One of arbitrums arguments for timeboost was that getting the lowest latency was reserverd for only a handful of high profile players because of the cost and complexity to colocate (keep that in mind for later).

The reasoning behind timeboost

Arbitrums claims that FCFS architecture resulted in a lot of bots spamming the chain to basically “poll” on chain state and instantly execute when an opportunity becomes visible. A lot of bots were also keeping up dozens if not hundreds of parallel connections to the sequencer feed to eliminate the randomness in what order the sequencer sends websocket messages to the subscribing clients. This is of course puts unreasonable amounts of load on the sequencer infrastructure. There also was another issue with the FCFS model, arbitrum had no way to extract value from MEV operators. I understand and support reclaiming MEV on either the chain or protocol level when it comes to MEV which actually extracts value from users. However that kind of MEV was already borderline impossible on arbitrum anyways since the transaction ordering is sequential and there is no mem pool to inspect transactions before they are getting executed. Arbitrage and Liquidations are both necessary for protocol health and I personally don’t see any reason why protocols or chains should get a revenue cut of MEV bots who invested thousands in infrastructure and R&D. However this is heavily opinion based and I might be biased in this regard.

Timeboost and how it works

Timeboost is a novel ordering mechanism which instantiates two lanes for transaction submission. A standard lane and an express lane. The standard lane is the default lane for all transactions and adds a 200ms delay to transaction inclusion. The express lane on the other hand can skip those 200ms. Every 60 seconds theres an auction on who will be the expresslane controller for the following 60 seconds. A transaction controller can then sign transactions which will be timeboosted, skipping the 200 ms. To fight big entities hogging timeboost control, arbitrum supported projects like kairos who would actively try to win auctions to then sub auction timeboost access by introducing a similar transaction ordering strategy to ethereum where MEV bots bid for earlier inclusion.

Why timeboost isn’t fixing what it is supposed to fix

One of the main arguments for timeboost was to reduce transaction spam, however if we think about how it’s trying to achieve that and why spammers are even spamming in the first place, one will quickly realize that timeboost has zero effect on liquidations and atomic arbitrage.

AAVE - An example

I will explain this based on AAVE liquidations. A user on aave is considered liquidatable when the healthfactor of a user drops lower than 1. The healthfactor can be calculated via this formula:
Health Factor = (Total Collateral Value * Weighted Average Liquidation Threshold) / Total Borrow Value
Both total borrow value and total collateral value depend on the prices of the reservers borrowed/supplied by the user. AAVE uses chainlink oracles as their primary price source which regularly push a price on-chain. The optimal liquidation would happen at i + 1, where i is the tx index in which the price got updated.
This is of course not possible in a deterministic way, neither pre or post timeboost. At the time that the price update transaction is visible, that block is already produced by the sequencer and there is no way to get into the optimal i + 1 slot anymore. Since liquidations are highly profitable on arbitrum and can yield vast sums of liquidation bonus some liquidators found a way around this. By monitoring off chain prices it is possible to anticipate chainlink updates, these bots then spam thousands of transactions in the upleading seconds to get into the earlier mentioned optimal i + 1 slot. This isn’t solved by timeboost, in fact, timeboosted transactions gain absolutely zero advantage here.
If anything it made spammers stronger since deterministic bots face an extra 200ms delay which has no effect on the strategy used by spammers. The top 3 spamming liquidation bots consumed around 8% of the total gas spent on arbitrum in the past month at the time of writing. (2026-02-08)

Arbitrage - Another example

It’s by no means a secret that L2s facilitate spamming bots, it is common practice for bots to not actively react to arbitrage opportunities that they poll off chain but rather have a few routes that are likely to generate profit and poll these constantly by sending transactions, reverting if the the path isn’t profitable.
Again timeboost is not helping here by any means since the predictive branch of arbitrage is essentially nerfed by timeboost.

Who Timeboost Actually Serves.

Timeboost is primarily used by big trading firms and market makers running CEX-DEX arbitrage. Notably Wintermute and Selini Capital, timeboost auctions got incredible highly priced since its introduction last year. At the time of writing Wintermute and Selini Capital paid around 220,000$ in ETH in the past 7 days. This shows that timeboost favors big parties and any efforts to bring timeboost to the people failed horrendously. Kairos, which subauctions timeboost express lane access won 1.0% of the auctions in the past 7 days. Underlining my point that timeboost is nothing more than arbitrum selling itself out to big trading firms.
Another thing which had a bitter aftertaste to me is that wintermute was the third biggest party voting for timeboost in the arbitrum proposal.