Cryptocurrency prices can vary across exchanges, creating opportunities for arbitrage. Traders can buy at a lower price on one exchange and sell at a higher price on another. However, these price discrepancies are often minimal—typically just a fraction of a percent—and last only for seconds. By the time a trader notices the spread and places their orders, the gap may have already closed.

The arbitrage platform Arbitron automates this process. It monitors order books and funding rates across 20 exchanges, executing trades automatically while keeping funds in the trader's exchange accounts.

ForkLog collaborated with the Arbitron team to explore how the platform differentiates between real and fake spreads, handles API keys, and why it charges fees only on profits.

Making Money from Price Differences

Each exchange has its own order book, buyers, sellers, and regional demand. Liquidity flows between platforms unevenly, and deposit and withdrawal fees can exacerbate price discrepancies.

Arbitrageurs don't predict market direction; they mitigate price risk by holding opposite positions. This delta-neutral strategy is detailed in Arbitron's guide.

Profits come from the spread and the funding rate (funding). While it appears straightforward on paper, the reality hinges on speed and coverage.

“A typical inter-exchange spread ranges from 0.01% to 0.5% and lasts for seconds. To consistently profit from these gaps, one must track thousands of pairs across multiple exchanges and submit two orders simultaneously. A human simply can't keep up; by the time they look at the numbers, the order book has already changed,” the Arbitron team explains.

Four Gaps Between Screen and Account

The numbers displayed by the scanner are raw; they merely represent the difference between two prices. The actual profit from this spread is significantly lower due to four factors: taker fees on both legs, the depth of the order book for the required volume, slippage during order execution, and the fading of the spread itself.

Calculating fees is straightforward: a complete cycle of two-legged arbitrage involves four market orders (opening both legs and closing them). Exchange fees vary significantly, so selecting the right pair is crucial.

Moreover, a spread lasts only until other arbitrageurs close it. The scanner updates every 10 to 15 seconds, meaning the most significant bottleneck is the trader.

23 Milliseconds to Binance

Arbitron connects to every trading pair on 20 exchanges, monitoring around 10,000 order books simultaneously. The platform retains the entire order book, not just the top price. Often, the best price has only a couple of dollars in volume, and the spread disappears with a real order. The service calculates the price at which an order will actually execute for $25, $100, $500, or $1,000.

Each account is assigned a dedicated server on AWS with a static IP, and traders can choose their region: Tokyo, Singapore, Frankfurt, or London. One relocation per month is free.

“Most major exchanges host engines in several cloud regions in Asia. Binance, Gate, KuCoin, Bitget, HTX, and CoinEx respond fastest from AWS Tokyo. MEXC and Hyperliquid also operate from there. Bybit and Phemex are based in Singapore, while OKX is closest to Hong Kong. Deribit in London and Poloniex in Europe are the fastest for derivatives,” Arbitron's knowledge base states.

A request to Binance from Tokyo takes about 23 ms compared to 206 ms from Singapore. Source: Arbitron.

Requests from a neighboring region take 10–35 ms, while those from another continent range from 200 to 500 ms. During this time, the liquid futures order book can update completely, causing orders to enter a book that no longer exists. In the team's measurements, such a trade resulted in nearly 0.8% slippage, several times higher than the typical spread.

Exchanges also limit the number of requests from a single address, and when using a shared pool, all clients deplete the limit simultaneously. According to Arbitron, external activity slows execution directly, whereas a personal IP eliminates this issue.

The trading core is developed in Rust, chosen for its lack of unexpected pauses on critical paths. The interface resembles a trading terminal, with prices highlighted on every change. If the data stream breaks, the number dims and is struck through.

Backtesting Instead of a Showcase

The platform features two search tools. Signals provide real-time data: traders receive notifications as soon as a spread crosses a specified threshold.

The scanner works differently from a typical screener: instead of the current spread, it shows the results of a backtest with a 5-minute delay. The platform runs the last 8 hours of recorded prices through the same two-threshold strategy used for trading cards. Each row reflects the outcome of such a simulation, with fees already deducted.

Source: Arbitron.

The scanner calculates taker fees not based on the average market rate but according to the trader's personal rate: VIP level and discounts for native tokens are specified in the exchange key settings. This figure determines the breakeven point.

The backtest also accounts for latency: traders can choose intervals of 0.5, 1, or 2 seconds, with each signal calculated at the worst price over that period. A spread that flashed for half a second won't be included in the report, along with the main illusion of arbitrage screeners: attractive percentages that are unattainable.

Next to each opportunity, there’s a reliability rating from 0 to 100: the number of closed cycles in the backtest, profit margin relative to fees, and daily turnover of the less liquid leg. Market depth accounts for a third of the score, making it challenging to achieve a high rating with thin pairs. Any score below 50 is recommended as a yellow flag in the knowledge base.

Under each pair lies a history of spreads over the past 30 days. The monthly chart highlights moments when the spread exceeded the chosen threshold, with adjacent buttons allowing navigation to the next spike. This way, users can see whether the pairing was active throughout the month or just on a lucky day. The table also indicates when the next backtest result will be available.

Trading Cards

Operations on Arbitron revolve around trading cards. Each card links a pair of exchanges, a specific coin, and two spread thresholds: for opening and closing. They operate independently, meaning a failure on one does not affect the others.

The exit threshold is calculated along with the cost of reversing trades: the profit estimate already considers the costs of unwinding both positions. Without this adjustment, a spread may appear profitable on paper but actually result in losses.

The card works with spread movements in either direction: if the spread rises, the platform opens one configuration; if it falls, it opens a mirrored one. Developers claim many bots are rigidly tied to one scheme (for example, always short on Binance and long on OKX), causing half of the movements to bypass them.

Both orders are sent to the exchanges simultaneously, but sometimes one is only partially filled. In such cases, the algorithm either adds the missing volume or closes the already opened position. The trader receives a notification regarding the situation.

Before executing a trade, the card checks whether there is enough depth in the order book for the entire order: a spread that is real for $500 may be fictitious for $10,000. The “Depth Multiplier” setting scans the order book for the required volume and calculates the weighted average execution price.

The second filter, the validation window, defaults to requiring 1000 ms: the spread must remain above the threshold for one second without a single drop, or the countdown starts anew. Instant threshold crossings do not reach the trade.

Trading card settings: thresholds, order size, leverage, and protective limits. Source: Arbitron.

Traders can configure the scanner filter based on spreads, exchanges, and minimum liquidity or save a preset. The platform then creates and launches cards automatically. The number of cards is limited, both for each set of conditions and for the account overall: automation does not exceed the set risk parameters. For those who prefer to catch pairs manually, a manual mode with live order books from both exchanges is available.

In addition to inter-exchange spreads, the service also calculates funding arbitrage on the spot plus perpetual futures pair or between two derivatives. It even tracks inverse contracts, where collateral and settlements occur in the coin itself rather than in USDT, a feature that few competitors monitor.

Position Protection: From Stops to Delistings

Spreads do not always converge back, so the card includes three safeguards. The first is a hard stop: at a set loss threshold, the algorithm exits the market without waiting for a return. The second is softer: the platform stops opening new cycles and seeks an opportune moment to close. The third activates by default 3% before the liquidation price.

Delisting poses a greater risk than any stop: when an exchange removes a contract from trading, one leg disappears while the other remains exposed without coverage. Arbitron detects such events in three layers. The first queries the official API announcements from Binance, Bybit, Bitget, and OKX. The second checks the complete list of instruments every 5 minutes, while the third reacts to the disappearance of a symbol from the feed. Exchanges typically publish planned delisting warnings 3 to 14 days in advance.

Upon detecting a delisting, the platform halts position accumulation and sends a notification to the trader via Telegram. Existing legs can be closed through the standard interface, negating the need to go to the exchange. The service does not close them automatically; an exchange's API failure can easily be mistaken for a delisting, leading to a sale at an arbitrary price.

Market makers exit doomed instruments ahead of time, causing spreads to widen five to ten times in the final hours, and funding can spike to ±1–2% over an eight-hour interval. Arbitron recommends closing positions 24 to 48 hours before shutdown while there are still buyers in the order book.

This protection feature was highlighted by the host of the YouTube channel “Diana, Arbitrage, and Love,” who called it one of the platform's most useful functions in her video review.

“I have traded coins that were delisted. Spot trading stops, and you are left with coins you cannot sell, while futures are simply delisted. […] So this protection is very timely here,” Diana noted.

Keys That Cannot Withdraw Funds

The non-custodial model is based not on trust but on access rights. Arbitron obtains API keys that allow trading only and does not request withdrawal rights, ensuring that it cannot withdraw assets.

Arbitron stores keys using the “converter” scheme with AWS KMS, Amazon's hardware storage. Each entry is encrypted using the AES-256-GCM algorithm. Each user has their own encryption key, and there is no single master password.

A separate key should be issued for each third-party service, disabling withdrawal rights and linking it to a static IP through an exchange whitelist.

Source: Arbitron.

Every trade is logged in full: price, volume, fees, order number, time. The execution speed across each exchange is visible, along with how much one leg lagged behind the other. Balances with open positions are compiled in the portfolio section, while the terminal opens in Telegram, with automatic login and a full set of settings.

Paying Only When You Profit

There are three pricing plans, differentiated not by trading volume but by what the platform is allowed to do. The Scanner plan, priced at $39 per month, provides access to data without execution: all scanners with full depth and no delay, signal feeds, live spread streams, depth history, and executable volume. Traders place orders themselves, and there are no profit-based fees on this plan.

The Trader plan, costing $99 per month, adds the features that typically attract users: a dedicated server, a static IP with one free monthly change, trading using personal keys, and up to 10 strategies simultaneously. The fee here is 35% of realized profits.

For ForkLog readers, Arbitron has opened a special registration link: in the first month, the profit fee will be 20%, and there is no joining fee.

The third plan, Prime at $299, is marked as preparing for launch on the pricing page. It features a lower fee of 30%, an increased strategy limit of 50, and additional capabilities such as directional trading in one leg, multiple accounts on one exchange, and a private Telegram channel. Annual payment for any plan grants two months free: $390, $990, and $2990, respectively.

The platform charges nothing else: no transaction fees, no data fees, and no fees on account balances. Traders pay exchange trading fees directly to the exchanges, which are deducted before calculating profit fees. Three additional services can be paid separately: VPN, priority IP change, and a second relocation within the same month.

The platform calculates fees weekly and only on realized results: a week without profits incurs no costs. Losses do not vanish with the reporting period but carry forward. Income that merely offsets a loss is effectively zero, and calculations resume once the prior peak is exceeded. A loss of $100 in the first week and a gain of $100 in the second results in no fees owed.

Cashback on trading fees applies to seven of the 20 exchanges: 60% on Toobit, 40% on Bitget, BloFin, Gate, and MEXC, and 30% on Bybit and KuCoin. However, to qualify, accounts must be registered through Arbitron's link and have the UID linked.

Who Should Use Arbitron

Arbitron is designed for those who understand the basic principles of futures trading and are willing to dive into trading strategy settings. Stablecoins for both legs and accounts on at least two exchanges are necessary. Platform representatives recommend starting with $3000, splitting the amount roughly evenly. Initial setup can take from 30 minutes to several hours, depending on the number of platforms connected. Most of the time is spent creating keys and transferring funds.

In exchange, the service takes care of the routine: monitoring thousands of pairs, simultaneous execution, cost accounting, and tracking delistings. Users discuss strategies in the Arbitron community on Discord and Telegram.

Connected exchange accounts can only be used with the platform: external bots and manual trades outside of it can disrupt position accounting and hedging.

Learning to use the platform takes from several days to a week, depending on experience, with the outcome determined by how carefully thresholds and limits are set. Understanding the mechanics in advance is aided by the open knowledge base and public spread scanner.