How to Design a Cross-Chain Arbitrage Strategy

1. The wrong starting point

The first question most teams ask when designing a cross-chain arbitrage strategy is: SIA or SDA?

It is the wrong question. SIA and SDA are execution mechanisms. They describe how a trade moves across chains. Choosing between them before defining your assets, capital structure, and information model is like choosing an engine before deciding what to build.

The result is a strategy assembled backwards. When it breaks under real conditions, the redesign is expensive.

There is a correct order. Four steps, each one constraining the next.

2. Step 1: Assets and unit of account

Before capital structure, before routing, before bridge selection: decide what you want to hold and what profit means.

If your target is USD, every trade that ends in a non-USD asset is incomplete until converted. The conversion cost is real and must be modeled. If your target is BTC, a trade that generates ETH is not profitable until you account for the BTC conversion step.

If you accept any asset, you have no unit of account and EV is undefined.

This choice also determines whether inventory drift is acceptable. A strategy that tolerates holding mixed assets across chains has very different capital requirements than one that enforces strict denomination.

Everything downstream depends on this decision. Make it first.

3. Step 2: Capital policy

Once you know your assets and unit of account, you can define how capital is structured across chains.

The key decisions: how much capital on each chain, how to split between core inventory and opportunistic buffer, and when to rebalance. Each depends on the asset choices made in step one.

One principle applies regardless: rebalancing belongs outside the critical path. It is a maintenance operation, not a trading signal. A strategy that rebalances in response to price movements is no longer arbitraging: it is speculating.

If capital sits idle most of the time, there are two likely causes: the allocation is skewed toward a direction the market does not support, or the total capital deployed exceeds the available opportunity in this market. Either way, the answer is not to wait: it is to resize.

4. Step 3: Information structure

With capital defined, you can now ask the structural question: is there an informational leader between your venues?

One chain tends to discover price first. The other follows. The gap between them is the spread you are trying to capture, and how long that gap persists determines its value.

Whether or not a clear leader exists, the allocation across directions must be validated through simulation. A structural leader tells you where to look: the data tells you whether the opportunity is actually there and how to size it.

This step also tells you when not to trade. If spreads are too narrow, too short-lived, or too competitive to generate positive EV after costs, no execution optimization will fix it.

[leader-follower article — covers the leader-follower dynamic and capital asymmetry in depth]

5. Step 4: Execution mechanism (SIA vs SDA)

Only at this point does the SIA vs SDA question make sense.

SIA means both legs are sequence-independent: neither depends on the other completing first. The most common implementation is inventory-based, with assets pre-positioned on both chains before any opportunity appears.

SDA puts the bridge in the critical path. Capital starts on one chain, the first leg executes, the asset transfers, the second leg executes on the destination. Lower inventory requirements, but price can move during the bridge window and partial execution becomes a real failure mode.

The choice follows from the previous steps. If inventory is sufficient on both chains and the bridge is slow or expensive, SIA is structurally cleaner. If capital is limited or the bridge is fast and reliable, SDA becomes viable.

The leader-follower structure from step three shapes the allocation within whichever mechanism you choose, not the mechanism itself.

6. Why the order matters

Each step in this framework depends on the previous one. That dependency is the reason the order matters.

You cannot define capital policy without knowing your assets. If your unit of account is USD and you discover mid-design that your capital policy requires holding native tokens on both chains, you are redesigning step one, not step two.

You cannot evaluate the information structure without a defined capital policy. The question "does a leader exist and how should I weight each direction" only has a meaningful answer if you know how much capital is available per chain and what the rebalancing constraints are.

You cannot choose between SIA and SDA without both. The execution mechanism is the last variable, not the first. It is determined by everything you built before it.

Building backwards is possible. Teams do it constantly. The cost is that every stress condition forces a redesign of earlier decisions, and those redesigns compound.

7. Conclusion

A cross-chain arbitrage strategy is not a choice of execution mechanism. It is a set of decisions that build on each other in a specific order: what you hold, how capital is structured, what the market's information structure looks like, and only then, how trades execute.

Most strategies fail under stress not because the execution mechanism was wrong, but because the earlier decisions were never made explicitly. When the market changes, there is nothing to fall back on.

The four steps do not guarantee profitability. They guarantee that the strategy is built on explicit, traceable decisions. That is the minimum foundation for a system you can debug, adapt, and improve.

Step three, the information structure, is covered in depth in the companion article: [leader-follower article].