So I’m running 6-7 different experiments, each of which loop over 250 tickers and pull a price for each ticker.
Is there a limit to how many experiments I can run simultaneously (from the API side)? I’ve read inconsistent reports on limitations on the API so I’m not really sure.
@redcoatwright The number of account REST calls one can make is limited to 200 calls/min. This is per account. If one wanted to test 6-7 different strategies/algos, the simplest is to create 6-7 separate paper accounts. However, one can have multiple algos/computers/instances use the same account, but the sum of all the calls would need to stay below 200 call/min.
If you are referring to the market data APIs, the REST calls are limited to 200 calls/min with a Free subscription and 10,000 calls/min with the Unlimited subscription. Similar to the account APIs, the calls can be from a single or multiple algos.
For both the account and market data APIs, the number of streaming websocket connections is limited to one so streaming becomes problematic if running multiple algos.
I would suggest not ‘looping’ over 250 symbols. All the market data endpoints allow fetching multiple symbols (with roughly a max of 4500 symbols/call) which is much more efficient than fetching one at a time.
Ohhh right, I forgot I’m doing 1 call for the lump 250 tickers then looping over the prices, that happens every 3-4s so I’m doing about 17 calls every min which means I have a LOT more bandwidth there.
Awesome! Wait can we have multiple paper accounts now? I thought that wasn’t available yet.
@redcoatwright One can have multiple paper accounts but they must each have a unique email associated with it. They aren’t sub-accounts. They would be literally separate accounts.
Oh oh I see what you mean.
Honestly, I wonder if the play here is to inculcate the historical data into the experimentation so I can just backtest, I’m guessing the granularity for it isn’t as high, though.
Edit: nvm granularity of historical data is insanely high