Hacker News Reader

Comment by chollida1 | Hacker News Reader

Bold claim, and funnily enough, you are the 3rd or 4th team I've run into this week that claim to be the fastest trading system on GDAX. The Consensus conference brings out these kind of conversations:)

What are you basing your claim off of, I've seen some pretty impressive use of C++ and one team that is starting to burn simple strategies into FPGA's.

What sort of time from network stack to network stack are you consistently getting? ie time from when your network card receives a message until when your network card gets your reply?

That's a standard HFT measurement.

View on HN

These posts kill me. Anyone who says they are burning strategies onto fpgas when they trade with GDAX has no idea what they or doing or talking about. The platforms are way too immature for this to matter. Source, was in real HFT for 10+ years.

Seriously. I bet the data feeds don't even use FPGAs let alone the matching engines. If I'm not mistaken, GDAX runs on AWS.

Yep, can confirm from the other side. (Someone doing crypto equivalent of “HFT”.) It’s nowhere near real HFT level.

> What are you basing your claim off of, I've seen some pretty impressive use of C++ and one team that is starting to burn simple strategies into FPGA's.

What does low latency buy you in crypto trading? What sort of strategies are people running that requires low latency?

the simplest is to monitor prices on multiple exchanges and buy/sell depending on fluctuations.

You mean Exchange Arbitrage? There's already a 20 something kid doing this and claims to be making millions in profits.[1]. Asian exchanges typically have slightly higher prices than US and European exchanges. They also react a lot worse to bad news and go lower than in US, European exchanges.

[1] Source: Stefan Qin, Virgil Capital founder on CNBC - https://www.cnbc.com/video/2018/02/27/one-hedge-fund-manager...

Plenty of people are doing it - that's why having the fastest bot is helpful.

You're probably thinking of my post from a few weeks ago in this thread https://news.ycombinator.com/item?id=16922538

I'm tracking order positions with an in memory orderbook. When gdax accepts my order I know where I end up. I average spot 1.45 when the spread changes. It's hard to get closer to perfect 1.00 because of limit taker orders which, if placed right, always get 1.00 after the taker portion is calculated and the maker is placed 1 cent from other other side.

> I'm tracking order positions with an in memory orderbook.

Yes, but you still didn't address the question of what are you basing your claim of being the fastest off of.

What?

He asked you how you know your bot is the fastest, and if you can prove it.

I answered it already. When the spread changes my bot is the first one to the new best bid / ask.

Reply on news.ycombinator.com