3 ms·

Did you code the stuff yourself? How did you start? I am looking forward to writing my own stuff. I found that some exchanges like OKX/OKEX have APIs that allow

by fb03 5y ago

Did you code the stuff yourself? How did you start? I am looking forward to writing my own stuff. I found that some exchanges like OKX/OKEX have APIs that allow you to algorithmically trade.

samtho 5y ago
I was using a smaller exchange’s API but now I use alpaca.markets. I wrote the program in Node.js that just runs a callback function every N second interval and decides if it’s a good time to buy, sell, or do nothing. My best advice to getting started is write your trading algorithm or decision making logic in a parameterized way so you can feed in snapshots of time for back testing. I’m sure there are frameworks out there but I literally wrote this hot garbage in a weekend with not a lot of research. I’ll probably rewrite it someday and optimize it a bit more.
- fb03 5y ago
  Wow, I just read your comment but even without doing so, this is exactly what I did last weekend! Loaded 4+ years of crypto market data into a database using python+some binance api, so that I can feed the historic data for backtesting/development purposes!
  - samtho 5y ago
    I’ve looked at one or two frameworks (that I cannot even recall now) and they all seemed to over complicates thing. I’m not smart (or educated on the specifics of) enough to invent complicated trading algos, so why not just make it simple, observable, and testable. Great to see you came up with this independently as well, I’m not in finance so its a little validating that I’m not totally crazy!
    - fb03 5y ago
      Same thing for me. Both frameworks I looked (freqtrade and some other with a fancy name) had a insurmountable amount of context needed just to grok the concepts and how to make the complicated machinery go. I would probably spend months just trying to understand it. It got me so uneasy to the point I just decided to go straight at it. Load the values and hey, let's try some backtesting. Simple stuff. If at the end of the day I end with more money than I began, shit is working. tweak that into oblivion. That's my plan. So yeah, you also validated me (because I was thinking "this is just going to be stupid stuff. that can't work"). Thank you for existing and lucky me to be able to have this timely exchange in such a noisy communication world heh o/

by fb03 5y ago

Did you code the stuff yourself? How did you start? I am looking forward to writing my own stuff. I found that some exchanges like OKX/OKEX have APIs that allow you to algorithmically trade.

samtho 5y ago

I was using a smaller exchange’s API but now I use alpaca.markets. I wrote the program in Node.js that just runs a callback function every N second interval and decides if it’s a good time to buy, sell, or do nothing. My best advice to getting started is write your trading algorithm or decision making logic in a parameterized way so you can feed in snapshots of time for back testing. I’m sure there are frameworks out there but I literally wrote this hot garbage in a weekend with not a lot of research. I’ll probably rewrite it someday and optimize it a bit more.

fb03 5y ago
Wow, I just read your comment but even without doing so, this is exactly what I did last weekend! Loaded 4+ years of crypto market data into a database using python+some binance api, so that I can feed the historic data for backtesting/development purposes!
- samtho 5y ago
  I’ve looked at one or two frameworks (that I cannot even recall now) and they all seemed to over complicates thing. I’m not smart (or educated on the specifics of) enough to invent complicated trading algos, so why not just make it simple, observable, and testable. Great to see you came up with this independently as well, I’m not in finance so its a little validating that I’m not totally crazy!
  - fb03 5y ago
    Same thing for me. Both frameworks I looked (freqtrade and some other with a fancy name) had a insurmountable amount of context needed just to grok the concepts and how to make the complicated machinery go. I would probably spend months just trying to understand it. It got me so uneasy to the point I just decided to go straight at it. Load the values and hey, let's try some backtesting. Simple stuff. If at the end of the day I end with more money than I began, shit is working. tweak that into oblivion. That's my plan. So yeah, you also validated me (because I was thinking "this is just going to be stupid stuff. that can't work"). Thank you for existing and lucky me to be able to have this timely exchange in such a noisy communication world heh o/