Inter-exchange arbitrage is not a bet on price direction, but an exploitation of market inefficiencies. The same asset at the same moment in time may trade slightly cheaper on one platform and slightly more expensive on another. The trader's task is to catch this gap, buy where the price is lower, and sell where it is higher. The problem is that spreads rarely exceed fractions of a percent and exist for only a few seconds. It is physically difficult for a human to react to such a window of opportunity while analyzing the order book and placing orders.
It is precisely this routine that the Arbitron platform automates. The service monitors order books and funding rates on 20 exchanges, then independently places both sides of the trade, which in professional slang are called "legs." Funds remain on the trader's exchange accounts — the platform is not a custodial vault.
Speed Decides Everything: Infrastructure and Latency
The key success factor in this type of trading is execution speed. Arbitron is connected to every trading pair on 20 exchanges, totaling about 10,000 order books simultaneously. The platform stores the full order book in memory, not just the top quote, which allows it to calculate the real execution price for volumes of $25, $100, $500, or $1000. After all, at the best price there is often only a couple of dollars, and with a real order the spread may simply disappear.
The infrastructure is built on dedicated AWS servers with a dedicated static IP for each account. The trader chooses the deployment region themselves — Tokyo, Singapore, Frankfurt, or London. This is critically important, as latency between continents is 200–500 ms compared to 10–35 ms from a neighboring region. During this time, a liquid futures order book manages to fully refresh, and the order arrives at a book that no longer exists. According to the team's measurements, such an error cost nearly 0.8% in slippage — several times more than a typical spread.
The trading core is written in Rust, a language the developers chose for its absence of sudden pauses on the critical execution path. This minimizes latency and ensures operational stability.
From Scanner to Real Profit: Backtest Instead of a Showcase
The platform offers two tools for finding opportunities. Signals display real-time data and notify the trader when a set spread threshold is crossed. The scanner, however, works fundamentally differently: it shows not an instantaneous gap, but the result of a backtest with a 5-minute delay. The platform runs the last 8 hours of recorded quotes through the same strategy used by trading cards, and then subtracts fees from this simulation.
This approach filters out the main illusion of arbitrage scanners — attractive percentages that are impossible to achieve in reality. The backtest also incorporates execution latency: the trader selects an interval of 0.5, 1, or 2 seconds, and each signal is calculated at the worst price during that period. A spread that flashes for half a second does not make it into the report.
Trading Cards: Automation with Risk Protection
Work in Arbitron revolves around trading cards, each of which links a pair of exchanges, a specific coin, and two spread thresholds — for opening and closing. Cards operate independently, so a failure of one does not bring down the others. The platform handles divergence in either direction: if the spread moves up, one configuration opens; if it moves down, a mirrored one. Many bots are rigidly tied to a single scheme and miss half of the movements.
Position protection includes three levels: a hard stop at a set loss threshold, a soft mode that stops opening new cycles and seeks a convenient moment to close, and an automatic stop 3% before the liquidation price. Separately worth noting is delisting protection — the platform monitors the official announcement APIs of major exchanges, checks the instrument list every 5 minutes, and reacts to a symbol disappearing from the feed. This is critically important, as when a contract is removed, one leg may vanish, leaving the second in the market without insurance.
Security and Economics: Commission Only on Profit
The non-custodial model relies on access rights: API keys have permission only for trading and do not allow withdrawing funds. Keys are encrypted using an "envelope" scheme with AWS KMS and the AES-256-GCM algorithm, with each user having their own encryption key.
The monetization model deserves special attention. The platform charges a commission only on realized profit, not on turnover or capital. Tariffs differ in feature sets: Scanner at $39 per month provides data only, Trader at $99 adds a dedicated server and trading, and Prime at $299 offers advanced capabilities. Importantly, losses are carried forward to future periods: income that merely covers a previous loss is not subject to commission. This is a fair approach that aligns the interests of the platform and the trader.
My analysis: Arbitron is an example of a mature tool for professional arbitrage, where attention to detail (from latency to spread validation) genuinely determines profitability. However, it is worth remembering that arbitrage is not "free money," but a complex discipline requiring an understanding of market microstructure and discipline in risk management. The platform removes the routine but does not absolve the trader of responsibility for configuring the strategy.