1.4 Core risks in arbitrage

Arbitrage is not risk-free in practice. Capture OS deliberately tries to surface net opportunities, but you can still lose money if you don’t respect the following.

1.4.1 Slippage

Order books are not infinite.

• If you hit size X, your average fill price might be far worse than the “best bid/ask”.

• A +0.5% theoretical spread can vanish or turn negative once you move size through thin liquidity.

Good practice:

• Keep sizes realistic relative to pool depth.

• Assume worse execution than the first quote shows.

• Use slippage-tolerant calculations, not raw mid prices.

1.4.2 Fees

You pay:

• Trading fees (per leg, per venue)

• Potential routing/aggregator fees

• Priority fees during congestion

Every leg of a route eats into your edge. A “nice” 0.8% gross spread might be:

• –0.2% in fees

• –0.3% in realistic slippage

• → suddenly marginal or negative.

Capture OS focuses on net profitability after known fee models. You still have to factor in your own execution overhead.

1.4.3 Latency and race conditions

Between “opportunity detected” and “transaction confirmed”:

• Prices move

• Other bots might hit the same spread

• Pools get updated, routings change

This is why:

• Detection latency matters (Capture OS)

• Execution latency matters (your infra or Capture Bot)

• You don’t YOLO size on tiny edges that vanish in 1–2 slots

1.4.4 MEV, front-running, sandwiching

On any chain, if your transaction openly reveals:

• What you’ll buy

• Where you’ll sell

• In which size

Then MEV bots can:

• Jump the queue with higher priority fees

• Move prices against you

• Capture the spread themselves

Mitigations exist at the execution layer (private tx, bundled execution, sophisticated routing). Capture OS’s detection layer does not guarantee you’re safe from MEV — it just shows the mispricing.

1.4.5 Liquidity and fill risk

Even if the quote looks good:

• Pool depth might not be enough for your full desired size.

• LP position changes, concentrated liquidity moves, or protocol bugs can cause unexpected behavior.

Rule: never assume you get the printed price for arbitrary size. Start small, ramp only when you’ve observed real fills.