Crypto Arbitrage Bot

The idea behind an arbitrage bot

The same coin can trade at slightly different prices on two exchanges at the same moment. If it is $100.00 on exchange A and $100.60 on exchange B, a trader could buy on A and sell on B and keep the $0.60 difference per coin. An arbitrage bot watches the prices on both and sends the two orders when the gap looks wide enough. There are other forms too, such as a loop of three pairs on one exchange, which is covered in triangular arbitrage.

In theory this has no view on where the price is going, because the buy and the sell happen together. In practice the two orders are separate, they take time, and each has a cost. That is where the gap goes.

Why fees, latency and transfer time eat the gap

• Fees. The buy pays a fee and the sell pays a fee. Both are charged on the order value. If the fee is 0.1 percent per side, 0.2 percent of the gap is gone before anything else. Fee tiers vary by exchange and volume.
• Spread and slippage. The quoted price may be for a small size. A larger order fills at a worse average price. See spread and slippage.
• Latency. Latency is the delay between seeing a price and having an order filled. The gap lasts a short time. A bot that reacts a moment late faces a smaller gap or none.
• Transfer time. If the bot holds money only on exchange A, it must move coins to B to sell them. A transfer can take minutes and has a fee. During that time price can move by more than the gap.
• Limits and outages. Withdrawals can be paused or capped, an exchange can be slow in busy markets, and an order can be rejected on one side only.

Some bots avoid transfers by keeping money on both exchanges and trading each side from its own balance. That removes the transfer, but it ties up more capital and leaves the bot exposed to price moves on both balances. Balances also drift out of line and need to be moved at some point, so the cost is delayed rather than removed.

A worked example with made-up numbers

A made-up bot buys 50 coins at $100.00 on exchange A, a $5,000 trade, and sells them at $100.60 on exchange B. The gross gap is 0.6 percent, which is $30.00. The table takes off each cost in turn.

Of the $30.00 gross, $22.06 is gone, which is 73.5 percent of the gap. All the numbers are made up, but the pattern is the point: a gap that looks wide shrinks to a thin margin. If the fee were 0.2 percent per side the same trade would lose money.

How an arbitrage bot fails

• Leg risk. One order fills and the other does not. The bot now holds a position with price risk it never meant to take.
• The gap closes in transit. The figure above shows a gap shrinking from 0.6 to 0.2 percent. With the same costs, the result goes from $7.94 to -$12.02.
• Stale data. A price feed that lags shows a gap that is already gone.
• Fee changes. A drop in a volume tier or a change in the schedule moves the break-even gap.
• Counterparty risk. Money sits on more than one exchange, and each one is a place where withdrawals can be frozen.

How to check one honestly

List every cost on both sides, including the order type the bot would really use. Work out the break-even gap and compare it with the gaps the bot actually sees over weeks, not with the best one. Count how often both orders fill. A backtest that assumes both fill at quoted prices is the same trap as perfect fills in any backtest. Use a demo account or paper trading where the venue offers it, and read the crypto trading bot guide for key safety.

What this page does not tell you

It does not say that arbitrage works or fails for anyone. It shows the arithmetic and the ways the gap disappears. Whether a gap exists, how wide it is and how long it lasts depends on the venue and the moment.