I run the strategy for the first time and I see that some symbols have not been updated to the current date. When I run it second time, there are even fewer of these symbols. I have rechecked manually via Historical data loading testbed and it loads. It may be because of the limits of the data provider, then the ability to add a pause is needed I guess.
Rename
Which Provider?
What does the Log Viewer say about it? (Ctrl + L)
What does the Log Viewer say about it? (Ctrl + L)
CryptoCompare free account. No entries related to this provider in Log. I've noticed this with other providers as well, but there definitely because of the limits.
CryptoCompare has rate limits, but I don't think we've applied a throttle for it. The problem is that I can't find any documentation that states what the limit is. Maybe someone knows?
Cone, may be user configurable limit for each provider? For example, twelvedata also has limits and they depend on user's plan
Of course it depends on the provider and possibly the user's plan.
From what I can tell, CryptoCompare has a multi-rate limit and you can see how many calls you've made and have left by second, minute, hour, day, and month just by navigating here:
https://min-api.cryptocompare.com/stats/rate/limit
For example, my result looks like this:
calls_made: Object
second: 1
minute: 21
hour: 23
day: 100
month: 132
calls_left: Object
second: 19
minute: 279
hour: 2977
day: 7400
month: 49868
Based on "calls_left" it seems that they'll support bursts of up to 20 requests per second, but not more than 300 per minute, which is 5 requests per second. I'll add a throttle for 5 requests per second, and that should keep from missing data during updates.
From what I can tell, CryptoCompare has a multi-rate limit and you can see how many calls you've made and have left by second, minute, hour, day, and month just by navigating here:
https://min-api.cryptocompare.com/stats/rate/limit
For example, my result looks like this:
calls_made: Object
second: 1
minute: 21
hour: 23
day: 100
month: 132
calls_left: Object
second: 19
minute: 279
hour: 2977
day: 7400
month: 49868
Based on "calls_left" it seems that they'll support bursts of up to 20 requests per second, but not more than 300 per minute, which is 5 requests per second. I'll add a throttle for 5 requests per second, and that should keep from missing data during updates.
First I tried a 5/second limit, but still missed a few updates on a larger DataSet. It looks like 4/second is the sweet spot to avoid missing updates.
This is in place for Build 32.
This is in place for Build 32.
QUOTE:
Cone, may be user configurable limit for each provider? For example, twelvedata also has limits and they depend on user's plan
It's a pity that FMP's, Alphavantage's or TwelveData's API tokens aren't indicative of the user's plan so a provider can't guess what rate limits to apply. Despite it may seem reasonable to introduce user configurable limits, for now we've refrained from doing so because it's error prone and would require constant maintenance. User plans can change quickly, we cannot track vendors' changes to the rate limit policy.
Your Response
Post
Edit Post
Login is required