,
8 tweets,
2 min read
Read on Twitter
1/ I appreciate your efforts to compare exchanges, especially in terms of real/fake volume it is important work. But API response time as a measure of API quality is fairly meaningless, also for high frequency trading (which is what I do). I'll try to explain.
2/ Yes, API response time is important in a way. If I send a call to and endpoint, I would rather receive a reply in 100 ms than 200 ms, let alone 1000 ms. But API response time is not just a result of server and API quality. Server location is an important factor.
3/ If I colocate my server with an exchange, I'll get lower ping than if my call goes across the globe first. So if you measure from your London office, you're biased towards exchanges that are located near your office. Any trader that values ping will colocate with the exchange.
4/ Let's say you colocate with all exchanges, then measure response time to eliminate connection latency. Is that an objective measure of API quality? Not really. Some exchanges you list with low latency are ones that I actually consider low quality API exchanges.
5/ More important than low response time is availability and consistency. If an exchange normally quick to response, but turns unresponsive the moment the market is moving significantly, that exchange is useless to me.
6/ Also, what matters is the quality of data I'm getting when I query an exchange. As an example, Bitstamp caches data that it sends over some endpoints. Pretty useless to get data within 200ms that is actually 5000 ms old.
7/ Another good measure of API quality is the limits it imposes on users. If an exchange only allows a call once every 1 or 2 seconds, what use is it that I get a response within 200ms? I have to wait another 800ms to do anything with that information.
8/ Again, great that you're comparing exchanges. We need to get quality up accross the board, and transparency helps. But more nuance is needed to paint a fair picture. Happy to help if you need any further input on this!
