SimXa: Simultaneous Cross-Chain Arbitrage

This article is part of the series Building a Cross-Chain MEV Bot. It assumes familiarity with the SIA/SDA framework covered in How to Design a Cross-Chain Arbitrage Strategy.

1. The problem neither SIA nor SDA solves

Cross-chain arbitrage has a structural problem that neither SIA nor SDA solves cleanly.

SDA puts the bridge in the critical path. You execute the first leg, transfer the asset, execute the second. The price can move during the bridge window. The opportunity can disappear before the asset arrives. Capital is locked and unavailable for other trades while the transfer is in flight.

SIA avoids the bridge in the critical path by pre-positioning inventory on both chains before any opportunity appears. Both legs execute in parallel. But the inventory has to be the actual traded assets, ETH on chain A if you want to sell ETH, DOT on chain B if you want to sell DOT. Each route requires dedicated capital.

A strategy covering ten pairs ties up ten separate inventories.

Neither mechanism has access to cross-chain flash loans. On a single chain, flash loans are the standard tool for capital-efficient arbitrage. Cross-chain, that primitive does not exist.

SimXa is a SIA-based strategy designed around this constraint. It keeps SIA's parallel execution model: both legs execute in the same block, no bridge in the critical path. The difference is in what gets pre-positioned. Instead of route-specific inventory, SimXa uses generic collateral and on-chain lending to borrow the assets needed at execution time. One collateral pool covers any route.

2. How SimXa works

The strategy executes in two phases separated by the cross-chain settlement window.

The setup

Before any trade, collateral is distributed across chains, held in vaults, denominated in any stable or liquid asset. This collateral does not need to match the assets you intend to trade. It is the base from which lending positions are opened.

When a price discrepancy is detected, SimXa does not move capital from one chain to the other. It borrows the required assets on each chain independently, using the local collateral.

Block N, opening positions

Assume a price discrepancy in the ETH/DOT pair:

• Chain A: 1 ETH = 10 DOT (ETH overpriced)

• Chain B: 1 ETH = 8 DOT (ETH underpriced)

Two transactions execute in the same block, one per chain:

Chain A: Open a CDP using local collateral. Borrow 1 ETH. Swap 1 ETH for 10 DOT at the current rate. Running balance: −1 ETH (owed to CDP), +10 DOT.

Chain B: Open a CDP using local collateral. Borrow 10 DOT. Swap 10 DOT for 1.25 ETH at the current rate (1 DOT = 0.125 ETH). Running balance: +1.25 ETH, −10 DOT (owed to CDP).

In the same block, initiate two cross-chain transfers:

• 10 DOT from Chain A → Chain B

• 1 ETH from Chain B → Chain A

The arbitrage is complete at this point. Both swaps have executed. The profit is locked.

Block N+X, closing positions

When the transferred assets arrive:

Chain A receives 1 ETH, repays the CDP, recovers collateral.

Chain B receives 10 DOT, repays the CDP, recovers collateral. Remaining balance: 0.25 ETH (1.25 ETH received from the swap, 1 ETH used to repay the loan).

Both vaults are fully recovered and immediately available for the next opportunity.

Price risk is closed at Block N

Once the two swaps execute in Block N, the 0.25 ETH profit exists regardless of what happens to ETH or DOT prices between Block N and Block N+X. The open window does not create price exposure on the trade itself.

What it does create is loan exposure. If the collateral value drops significantly before the CDPs are closed, a liquidation becomes possible. That is the specific risk of the settlement window, covered in the risks section.

The shorter the window, the lower the exposure. Environments with native cross-chain messaging and fast finality, such as Polkadot with XCM, bring settlement latency to two blocks in an optimistic scenario.

This makes the strategy more viable in practice, even if the mechanics work anywhere the required infrastructure exists.

One structural constraint follows from using CDPs: you can only borrow a fraction of your collateral value, set by the loan-to-value ratio of the lending protocol. The collateral is route-agnostic. The same vault opens CDPs for any pair, on any block.

3. Capital structure: the Omni Vault

The collateral does not sit on a single chain waiting to be moved. It is distributed across all chains where the strategy operates, held in what the strategy treats as a single logical pool: the Omni Vault. Each chain holds a portion of it, and the full pool is monitored and rebalanced as one unit.

This design decision solves a timing problem. If capital had to be bridged before each trade, the bridge latency would become part of the critical path. By the time the capital arrived, the spread might be gone.

Pre-distributing collateral means that when an opportunity appears on any pair, the required vault balance is already on the right chain. The bot opens the CDP and executes immediately.

Profit management

Profits are added back to the local vault on the chain where they are realized, increasing the available collateral for future positions. This creates a compounding effect: larger vaults support larger CDPs, which allows capturing larger opportunities or holding more positions in parallel.

Users contributing to the vault can choose what asset to hold as collateral, as long as the asset is accepted by the lending protocol. The yield accumulates in the same asset or can be converted to a preferred denomination.

Rebalancing

Over time, vault values across chains will diverge. A chain with frequent arbitrage activity may see its vault depleted faster than others, reducing its capacity to open new CDPs.

Rebalancing is a maintenance operation, not a trading signal. When the imbalance exceeds a threshold, capital is transferred from heavier vaults to lighter ones to restore uniform coverage. This transfer happens outside the trading loop and does not affect open positions.

Yield stacking

The collateral does not sit idle between trades. If the deposited asset is a liquid staking token, it earns staking yield by design, independently of the arbitrage activity.

This creates a two-layer yield structure on the same capital. An investor holding a liquid staking token as collateral earns staking APY from the underlying position and arbitrage APY from the strategy at the same time, without choosing between the two. The collateral works twice.

4. Risks

SimXa extends SIA by making any route accessible with a single collateral pool. The tradeoff is a set of risks that classic SIA does not have, because SIA pre-positions inventory and has no settlement window to manage.

Settlement delay. Block N+X is not guaranteed. If the cross-chain transfer is delayed, the CDPs stay open longer and the collateral remains exposed.

Liquidation risk. If the collateral loses value significantly before the CDPs are closed, the lending protocol may liquidate the position. Severity depends on collateral volatility and settlement window length.

LTV constraint. Only a fraction of vault capital is deployable per trade, set by the loan-to-value ratio. The rest is locked as security. Yield stacking partially compensates for this.

5. Is SimXa better than SIA?

Not necessarily. Whether SimXa outperforms pure SIA depends on the specific context. The answer requires simulation against real data.

What is clear is the design space it opens. A practical approach worth exploring is a hybrid: run pure SIA on routes where inventory is already pre-positioned, and use SimXa for two specific cases:

• Capture opportunities on routes not covered by the existing inventory, without committing capital to a new dedicated position.

• Add exposure on an existing route using collateral held in a different token, layering additional capital on a route that SIA already serves.

In both cases, SimXa acts as a complement rather than a replacement. The SIA inventory handles the core routes. SimXa extends coverage where inventory is absent or constrained, at the cost of the settlement window and liquidation risks described above.

Whether that extension generates positive expected value in a given market is an empirical question. Simulating both strategies on the same historical data, with the same cost assumptions, is the right way to find out.

This article is part of the series Building a Cross-Chain MEV Bot. Previous: XChain Simulator · Decision Engine · How to Design a Cross-Chain Arbitrage Strategy Repo: xchain-mev-simulator