Most arbitrage systems don’t fail because they can’t find opportunities. They fail because those opportunities were never really executable in the first place.
At first glance, arbitrage seems simple: find a price difference and execute trades. It’s easy to assume that the more signals a system sees, the better it performs. In reality, the opposite is true. The real challenge in arbitrage trading is not finding opportunities — it’s deciding which ones are actually worth executing.
Why Finding Arbitrage Opportunities Is Not Enough
Most systems focus on detection. They scan markets, compare prices, and generate signals. But detection alone doesn’t make a strategy profitable. Not every opportunity that looks profitable should be executed. Some appear too early, some are already outdated, and some break during execution. When a system tries to act on all of them, it doesn’t become better — it becomes noisier.
That’s why control matters more than volume. A well-designed arbitrage system doesn’t just react. It filters, validates, and rejects most of what it sees.
How Arbitrage Actually Works in Practice
In HETHA, arbitrage is not implemented as a standalone bot reacting to signals. It’s part of a broader system where the trading bot acts as an execution layer. It doesn’t try to “understand the market”. It decides whether a specific opportunity should be acted on — or ignored. This is where most systems break.
What Makes Arbitrage Hard in Real Markets
Before talking about implementation, it’s important to understand what actually goes wrong. The same opportunity can appear multiple times within seconds with small changes in profitability. Without control, a system executes the same idea repeatedly, locking capital and increasing risk without improving results.
Between detection and execution, the market moves. Prices shift, liquidity changes, and spreads disappear. A signal that looked profitable a moment ago may no longer exist by the time the trade is placed.
After a trade, balances are not instantly confirmed. If a system assumes they are, it may start using funds that are not actually available yet.
Over time, capital drifts across assets and exchanges. Even if opportunities still exist, the system may not be able to execute them simply because the required funds are in the wrong place.
How Control Is Implemented
This is where control stops being an idea and becomes system behavior. Instead of reacting to every signal, the system filters incoming data and temporarily locks repeated opportunities after execution. In practice, this means skipping redundant signals and waiting for stronger setups instead of chasing every fluctuation.
After execution, only the affected assets are temporarily locked until balances are confirmed. This ensures that decisions are always based on actual, not assumed, capital while allowing the rest of the system to continue operating.
Before execution, each opportunity is recalculated using the current order book. If conditions no longer match, the trade is rejected. This is how the system avoids acting on opportunities that no longer exist.
Not all opportunities are evaluated the same way. A faster mode based on top-of-book prices produces more signals, while a stricter mode that accounts for market depth produces fewer but more reliable ones. This is not a different strategy, but a different level of control.
Trade size is not fixed. The system adjusts volume based on available balance and market conditions, either scaling trades automatically or increasing size to capture stronger opportunities. In arbitrage, risk is often tied to how many trades you execute, not how large they are.
In some situations, execution certainty matters more than price precision. Using market orders can ensure that all parts of a trade are completed, but introduces uncertainty in the final price. These decisions are controlled, not applied blindly.
Over time, capital naturally drifts across the system. Without rebalancing, execution eventually breaks — not because opportunities disappear, but because funds are no longer where they are needed. Reverse operations redistribute capital with minimal cost, keeping the system functional.
In certain conditions, trades may be allowed to stay open longer, introducing elements of trend-based behavior. This can increase returns, but also increases exposure to market risk.
Finally, the system evaluates not just profitability, but whether a trade can realistically be completed. It does not try to predict price direction. Instead, it answers a more practical question: can this actually be executed under current conditions?
Why Most Arbitrage Systems Fail
Most systems are built around detection. But in real markets, not all signals are valid, not all trades can be completed, and not all profits are realizable. Without control, more signals don’t lead to better performance — they lead to more mistakes.
Arbitrage Fails at Execution, Not Detection
Arbitrage is not about finding price differences. It’s about controlling execution. The difference between a system that “finds arbitrage” and one that actually profits from it comes down to one thing: control.
Not a single parameter. Not a single feature. But how the entire system is designed.