Inter-exchange arbitrage is a hunt for milliseconds. The same coin is priced differently on different platforms, and the price difference is ready-made profit. But the problem is that such discrepancies rarely exceed fractions of a percent and disappear faster than a trader can blink. Manual trading loses to machines here, and this is exactly the niche that the Arbitron platform fills.
The service automates the entire cycle: it scans order books and funding rates on 20 exchanges, simultaneously places both sides of the trade—the so-called "legs"—and does not take funds into its own account. Money remains on the user's exchange accounts, making the model non-custodial and reducing counterparty risks.
Why a "bare" spread is a trap
Scanners only show the difference in quotes, but the path to real profit is thorny. Four factors eat away at the margin: taker fees on both legs, the actual order book depth for the required volume, slippage during execution, and the natural decay of the spread. A full cycle of a two-legged trade is four market orders, and platform fees play a decisive role here.
The main problem is speed. The spread lives for seconds, and if the scanner updates every 10–15 seconds, it is already showing history. A human physically cannot react in time, so Arbitron takes over execution, leaving the trader only to configure the parameters.
Infrastructure: 23 milliseconds to Binance
The platform is connected to every order book on 20 exchanges—that's about 10,000 order books simultaneously. The full book is stored in memory, not just the best price, which is critical: at the top, there is often only a couple of dollars of volume, and a real order of $500 simply won't fill at that price. The service calculates the executable price for volumes of $25, $100, $500, and $1000.
Each account gets a dedicated AWS server with its own static IP. The trader chooses the region themselves: Tokyo, Singapore, Frankfurt, or London. This is not a whim but a necessity. A request to Binance from Tokyo takes about 23 ms, while from Singapore it's already 206 ms. From another continent—200–500 ms, and during that time, the order book of a liquid futures contract updates completely. In the team's measurements, such a delay cost nearly 0.8% in slippage—several times more than a typical spread.
The trading core is written in Rust—a language that tolerates no sudden pauses on the critical path. The interface resembles an exchange terminal: the price is highlighted with every change, and when data is interrupted, the number dims and is crossed out.
Backtesting instead of illusions
Instead of the usual screener with momentary figures, Arbitron offers a scanner that shows the result of a backtest with a 5-minute delay. The platform runs the last 8 hours of quotes through the same strategy that trading cards use and subtracts fees. The taker fee is calculated according to the user's personal rate, including VIP level and discounts for the native token.
Execution delay is also factored into the calculation: the trader chooses an interval of 0.5, 1, or 2 seconds, and each signal is calculated at the worst price over that period. A spread that flashed for half a second does not make it into the report. This filters out the main illusion of arbitrage screeners—beautiful percentages that are impossible to reach.
Trading cards: automation with protection
The work revolves around trading cards, each of which links a pair of exchanges, a coin, and two spread thresholds—for opening and closing. Cards live independently, so a failure of one does not bring down the others. The exit threshold is calculated together with the cost of reverse trades; otherwise, the spread looks profitable on paper but actually brings a loss in practice.
The discrepancy is worked in both directions: if the spread goes up—one configuration, if down—a mirrored one. Many bots are rigidly tied to a single scheme and miss half the movements. Both orders are sent simultaneously, but in the event of partial execution, the algorithm tops up the volume or closes the position, notifying the trader.
Before a trade, the card checks the order book depth for the entire order, and the validation window requires the spread to stay beyond the threshold for at least 1000 ms without a single dip. Instant crossings are filtered out. Three safeguards protect against an adverse scenario: a hard stop, a soft exit, and automatic closure 3% before liquidation.
Delisting—the main risk
Delisting is more dangerous than any stop: when an exchange removes a contract, one leg disappears, and the other is left without insurance. Arbitron catches such events in three layers: polling official announcement APIs, reconciling the full list of instruments every 5 minutes, and reacting to a symbol disappearing from the feed. The platform halts position building and notifies the trader via Telegram, but does not close trades automatically—an API failure can easily be mistaken for a delisting and lead to selling at a random price.
Market makers leave a doomed instrument in advance, and in the final hours, the spread widens 5–10 times, while funding accelerates to ±1–2% per eight-hour interval. Experienced traders close out 24–48 hours before the delisting, while there is still liquidity in the order book.
Security and fees
API keys are stored using an "envelope" scheme with AWS KMS and encrypted with AES-256-GCM. Each user has their own encryption key; there is no single master password. Keys do not have withdrawal rights, so the platform physically cannot take assets.
The payment model is fair: a fee is charged only on realized profit. The Scanner plan at $39 per month provides data without execution, Trader at $99 adds a dedicated server and trading with a 35% fee on profit. Prime at $299 is for professionals with a 30% fee and expanded limits. Losses are carried forward to future periods, and income that merely covers the hole is not taxed.
Who this suits
Arbitron is designed for those who understand the basics of futures trading and are willing to figure out the settings. You will need stablecoins for both legs and accounts on at least two exchanges. The recommended starting capital is from $3000. Mastering it takes from a few days to a week, and the outcome depends on how carefully the thresholds and limits are set.
My take: arbitrage is not passive income but full-fledged work with infrastructure requirements. The platform solves the main problem—speed and cost accounting—but success still depends on discipline and proper configuration. For those ready for this, Arbitron is one of the most well-thought-out tools on the market.