2 / 4 · 6 min
Before you compare: units, currency, ticker
Half of all 'arbitrage opportunities' exist only because something incomparable was compared. Before marvelling at a difference, bring both sides to the same form.
Comparing two prices looks simple: subtract one from the other. In reality a contract multiplier, a settlement currency, a different asset under the same name and a delivery delay may all stand between them. Each of those creates a 'divergence' that does not exist. Below are four traps, every one of which we caught in our own data, and all four cost us corrupted numbers until they were closed.
Contract multiplier
- A contract can be for one coin, for ten, for a hundred or for a thousand.
- Small coins are often traded in bundles: the contract is called 1000PEPE and costs exactly a thousand times more than one coin.
- Comparing such a contract with a plain one gives you a 'spread' of a thousand percent — and it is entirely your own mistake.
- A correct comparison always goes through dollar value: quantity × price × multiplier.
Settlement currency: how we corrupted our own history
A real case from our work. The Gate exchange served candles for stock indices in dollars, while the index itself lives in points — the numbers differed by roughly 1.16 times, exactly the exchange rate. Our corruption guard compares foreign candles with ours and rejects divergences larger than two times. A difference of 1.16 did not exceed the threshold, and the wrong candles settled quietly into history. The lesson: a threshold tuned for gross errors is blind to neat ones. Now prices from such a source are multiplied by the rate before the comparison, not after.
One ticker, different assets
Our second case: the ticker ON. On one venue it is a token, on another the stock of Ontrak. For eleven days the coin's history was filled with the stock's prices, and it could not be noticed by name — the names matched. It became noticeable by price: the numbers made no sense for that coin. Hence a practical rule worth adopting: an instrument is identified by price and venue, not by ticker. A ticker is a convention, a price is a fact.
Timestamps: two of them, and both are needed
Every event has the time the exchange stamped on it and the time the data reached you. They differ, and the gap differs per venue. Comparing by receipt time makes it easy to conclude that 'Binance moves first' when in fact its feed simply arrives faster. Any conclusion about who leads whom requires exchange timestamps; your own stamps are good only for checking whether the feed itself has fallen behind.
Comparing spot with a perpetual and calling it arbitrage
Spot and a perpetual contract are different instruments with different economics. Their difference is called basis and is usually explained by funding: if holding a long position is expensive, the future trades above spot, and that is a normal state rather than a market error. You can earn on that difference, but it is not 'buy here, sell there' arbitrage — it is a two-legged trade with a carrying cost and its own set of risks. The last lesson of this course is devoted to it.
The only quantity in which positions from different venues can be added and compared. Everything else — number of contracts, volume in coins, notional — depends on the rules of a specific exchange and does not carry across venues.
In our data bundle contracts are reduced to a single coin, index prices to their native currency, and instruments sharing a ticker are separated by venue. How that works — in the sources breakdown.
How we reconcile dataFind a bundle contract
Find a coin priced in thousandths of a cent and look at how its future is named on different exchanges. You will meet versions with and without the 1000 prefix. Work out how many times the 'prices' would differ if compared directly. It is the cheapest vaccination against false spreads.
On one exchange a coin costs 0.00004, on another 0.04. What do you do?
Do not celebrate a thousandfold spread — check the multiplier: almost certainly the second instrument is a bundle of a thousand coins. If the multipliers match, the next check is whether it is the same asset at all: tickers coincide often, the projects behind them differ. Only if both checks pass does the difference deserve attention.