Timeboost Risk Analysis

The Timeboost proposal aims to change how transactions are ordered on Arbitrum One and Arbitrum Nova by introducing an express lane that immediately sequences (that is, without delay) transactions from the express lane auction winner for that round, while normal transactions face a proposed 200ms delay. Chaos Labs supports the proposal, viewing it as a safe upgrade, but we do highlight several minor economic risks.

Despite these minor concerns, Chaos Labs’ find no major blockers to the proposal’s implementation from an economic perspective.

Description

Timeboost is a proposed upgrade that would affect the way transactions are ordered in Arbitrum blocks. A summary of this proposal is included here for ease of reference.

Current Transaction Ordering

Transactions on Arbitrum are currently ordered on a first-come, first-serve basis (FCFS) . There are no priority fees, meaning that the value created by time-sensitive transactions accrues to searchers and does not flow back to the DAO.

First-come-first-serve transaction ordering has served Arbitrum chains well to date by allowing extremely fast block times. The default is 250ms, but this can go as low as 100ms, providing Orbit chains with the low latency needed for specialized gaming and DeFi applications.

Harmful MEV such as frontrunning and sandwich attacks are prevented as there is no way to priortize any transaction earlier in a block.

First-Come-First-Serve is also simple for teams to understand, implement and operate. This is a significant benefit to teams deploying Orbit chains using the Arbitrum tech stack.

One consequence of FCFS transaction ordering is that it results in actors spamming the network when opportunities for MEV present themselves. Searchers submit multiple orders to capture MEV opportunities such as arbitrage and backrunning, in the hope that one of them is routed to the sequencer first.

Timeboost Transaction Ordering

Timeboost improves on the current Arbitrum transaction ordering policy by creating an express lane, giving the auction winner’s transactions a time advantage for inclusion at the sequencer. This is implemented by introducing a latency for normal transactions defined by the NonExpressDelayMsec parameter, while the auction winner’s transactions will be sequenced and included immediately.

This creates a market for prioritizing transactions, while keeping the transaction costs for regular transactions comparable to the first-come, first-served (FCFS) model. By introducing market-based pricing for priority transactions, time-sensitive transactions cost approximately their long-term value, boosting sequencer revenue without affecting the cost structure for the majority of transactions.

Auction participants need to deposit funds (for making bids) into the auction contract ahead of time and bid for express lane control for the upcoming round, which lasts 60 seconds. The auction closes 15 seconds before the start of the next round.

It is important to note that Timeboost increases latency slightly for non-express lane transactions, while express lane transactions latency remains the same as under FCFS. This means that the express lane owner has their transactions included amongst regular transactions submitted to the sequencer NonExpressDelayMsec milliseconds prior. The timing of inclusion of express lane transactions remains uncertain for this reason, but where previously unknown profitable opportunities emerge, the express lane owner will have a healthy advantage to capitalize on them.

For more details on Timeboost, see the original post, and follow-up FAQs.

Economic Impacts and Risks

Timeboost is extremely likely to achieve its stated objectives of:

  • To allow the chain to reduce negative externalities from the racing behavior induced by MEV searchers. Additionally, it can socialize the benefits of the transaction sequencing market back to the ArbitrumDAO.
  • To offer the Arbitrum DAO an opportunity to capture additional revenue that does not come at the expense of users.

The increased latency for non-express lane transactions will significantly reduce the incentive to submit multiple orders to capture MEV in most cases. Searchers should rationally assume that the express lane controller will use its latency advantage to capture most of these and not bother trying.

At the same time, assuming at least two independent and competitive parties bid for the express lane, this latency advantage should be valued relative to the expected MEV opportunities in each round. This will increase revenues for the Arbitrum DAO without increasing the fees regular users pay, maximizing the potential for ecosystem growth in an economically sustainable way.

While Chaos Labs generally views the Timeboost proposal as a positive step forward for Arbitrum chains and a safe upgrade to first-come, first-serve transaction ordering, we recommend monitoring for a number of potential second-order effects of this upgrade. It is not apparent exactly how the marketplace for the express lane will develop ex-ante. In this section, we highlight some possible ways this marketplace could develop and the associated risks.

Should either of these or other unknown risks materialize, further iteration on Timeboost could be necessary in the future. The risks highlighted below are meant to serve as a preemptive set of second-order behaviors to monitor rather than any blockers to the rollout of Timeboost from an economic risk perspective.

Secondary Markets for the Express Lane

A likely scenario is that secondary markets develop for access to the express lane where searchers compensate the express lane controller for access when they need it. Should this occur, the express lane becomes more valuable and therefore the auctions should clear at higher prices, increasing DAO revenue.

The express lane controller should rationally allow anyone willing to pay access to the express lane, creating a market for priority access on demand. Chaos Labs does not view secondary markets as a negative outcome, but acknowledges that much remains unknown around how they could unfold.

Competition in the Express Lane Auction

Independent competition is crucial for auctions to clear at efficient prices. The innovative and novel Timeboost ordering has no precedent currently and we therefore consider a range of potential market development outcomes for the express lane.

There is a risk that a small group of participants could dominate the Timeboost auctions, potentially due to their superior ability to capitalize on MEV or through other advantages. This dominance could lead to reduced competition or even collusion, undermining the fairness and efficiency of the system.

The dynamic reserve price mechanism acts as a defense against uncompetitive bidding scenarios. It is recommended to monitor the auction winners, their bids, and the MEV harvested over each round to ensure that the express lane does not become too profitable for auction winners.

Introducing MEV strategies not currently viable

Ordinary transactions will remain private until executed under Timeboost, which typically prevents the express lane owner from front-running user transactions. However, transactions like liquidations and pre-committed orders, such as limit orders, are at risk of being front-run by the express lane auction winner.

This occurs because the express lane owner has a latency advantage, allowing them to manage and execute these transactions with a higher likelihood of landing in their order of preference. While these same strategies are technically possible under FCFS, the probability of landing them is low enough to discourage them. It is unclear whether this will change under Timeboost to make these strategies more predictable.

In this case, the owner of the express lane has the opportunity to capture value that is not currently available at the expense of liquidated (or limit order-placing) users and/or protocols. It should be noted that if this is the case then presumably the express lane will become more valuable, increasing DAO revenue.

Example of potential Timeboost Ordering MEV on liquidation:

  1. The express lane owner identifies a liquidatable account on a perpetual futures protocol.
  2. The express lane operator maximizes their profit by executing a sandwich attack, taking advantage of their 200ms latency edge (or however Timeboost is configured). This predictability allows them to execute the liquidation unopposed and complete the sandwich strategy:
  • They first sell a large number of perpetual contracts on the relevant market, driving the price lower by either consuming order book bids or impacting the protocol’s pricing inputs, such as open interest imbalance.
  • They then liquidate the account, further pushing the price downward.
  • Finally, they close their position by buying back the contracts at the now-lower price.
  • As a result, users receive less collateral after the liquidation than they would have if the sandwich attack hadn’t occurred.

Comparison with Current First-Come, First-Serve Transaction Flow:

  1. An account becomes eligible for liquidation.
  2. Liquidators compete by spamming the network to execute the liquidation and claim the liquidation bonus.
  • Due to the competitive nature of liquidation rights, liquidators cannot easily predict or front-run transactions.
  • While this results in network spamming, the liquidations remain efficient.

Careful monitoring of express lane transactions should be sufficient at this stage as it is unclear whether Timeboost provides sufficient advantage to the express lane controller to execute harmful strategies. Keeping the NonExpressDelayMsec parameter shorter than block times could also be a useful mitigant in this regard.

Friction to the adoption of limit order books on Arbitrum Orbit Chains

Introducing a 200ms latency could render fully on-chain limit order books impractical in the future from a market maker UX perspective. Market makers rely on low latency to cancel orders promptly, and the delay would increase the potential for by the express lane controller to trade on their stale prices before they are able to cancel them, putting market makers at a significant disadvantage.

Considering the predominant DEX designs currently this is unlikely to be an issue on Arbitrum One or Nova, but application-specific Orbit Chains considering this might need to consider how this risk.

Market making on limit order books

On limit order book exchanges, there are typically an order of magnitude more cancelled orders than matched trades. The SEC reports that there are 14 cancelled orders for every matched trade in US stock markets. The reasons for this phenomenon are explored in more detail here and are found to be primarily driven by adverse selection risks.

Under Timeboost, when an arbitrageur controls the express lane, as we expect it to be much of the time, market makers are disadvantaged in the latency they experience relative to the arbitrageur. As a result, their stale prices would be vulnerable to arbitrage before they have the opportunity to cancel or update their orders.

This can be tangibly explained with the following example. Imagine market maker XYZ in the ETH market submits a bid at 2499 and an ask at 2501, in line with the best bid and ask on CEXs. Then imagine the external price on these CEXs quickly changes to 2504 bid, 2506 ask. The arbitrageur will, with certainty, buy from XYZ on Arbitrum and sell on a CEX, putting XYZ into a short position off-market.

While this risk exists with the current first-come-first-serve transaction ordering, arbitrageurs do not receive favorable treatment. Instead, the arbitrageurs will likely spam the network to compete against each other and the price-maker for the arbitrage opportunity. Under Timeboost, arbitrage is more likely, and market makers will have no option other than to widen their spreads as a result.

Conclusions

While the ultimate impact of the Timeboost upgrade on the microstructure of time-sensitive transactions on Arbitrum is impossible to completely predict ahead of time, Chaos Labs finds the economic risks acceptable.

We recommend closely monitoring the time sensitivities in and out of the express lane to ensure sufficient competition remains to mitigate abuse of the privilege in hard-to-predict ways. We also recommend monitoring to ensure that it does not lead to the enablement of cross-domain MEV, which could begin to harm users.

Chaos Labs has performed an economic risk audit of the Timeboost upgrade and finds no material blockers to implementation. Our analysis uses public materials and does not cover technical or security risks.

Disclaimer

The reports, and other information and materials (“Materials”) made available by Chaos Labs, Inc. (“Chaos”) are made available “AS IS” and “WITH ALL FAULTS.” YOU USE AND RELY ON THE MATERIALS AT YOUR OWN RISK. For the avoidance of doubt, unless otherwise set forth in a separate written agreement between Chaos and you, Chaos makes no representation or warranty with respect to the Materials or any other services provided or actions or omissions by or on behalf of Chaos or any other party, including without limitation the Arbitrum Foundation, Arbitrum Research and Development Collective (the “ARDC”), the Arbitrum Decentralized Autonomous Organization (the “DAO”), or any individual member of any of the foregoing entities (collectively, the “Arbitrum Parties”). Chaos disclaims any and all representations and warranties with respect to the Materials and any services performed by Chaos in connection with the ARDC, whether express, implied, statutory or otherwise, including without limitation warranties of merchantability, noninfringement, fitness for a particular purpose, and any warranties or conditions arising out of the course of dealing or usage of trade. Chaos specifically does not represent that the Materials will be error-free. The Materials may include content, software, information, and data provided or made available by third parties, including Arbitrum Parties (“Third-Party Materials”). Chaos has no obligation to investigate the source of such Third-Party Materials and does not represent that such Third-Party Materials are accurate or secure, nor that you have a right to use any such Third-Party Materials.

For the avoidance of doubt, Chaos will not be liable to you or any third party under any theory of liability for any claims or causes of action arising from or related to the actions or omissions of any Arbitrum Parties, and you agree that you will not, and will not cause or permit any third party to, bring any claim against Chaos in connection with any of the Arbitrum Parties’ actions or omissions. Chaos is not, and shall not be deemed, a partner, fiduciary, or other legal representative or agent of any of the foregoing.

4 Likes

quite interesting report, especially the consideration for order book protocols.
Thank you for the analysis.

great analysis and would love to see more discussion on this

This seems a very significant risk that should be further discussed before implementation.

I saw your comments on twitter already introduced some elements of discussion, especially about the risks involved, is there any reason not to host the discussion here?

1 Like

Yes, I think here is as good a place as any.

I would love to see a direct response from either Arbtirum or Espresso on these specific concerns

1 Like

Supporting ,great: the Timeboost proposal introduces a priority channel for transaction ordering, which will effectively reduce the MEV (Maximum Extractable Value) on the network’s

it is astounding you have managed to find no issues with timeboost, when mechanism designers have been pointing out issues left and right on twitter.

this is a very low quality analysis that seems to be lightly edited chatgpt output.

are you just the crypto mckinsey, paid to promote and push through whatever arbitrum wants to push through?