Determine "accuracy" of users

[StephanAkkerman]

Now that the new sentiment model is done, we can use this information to find out how good the predictions of the users that we follow are. We could consider multiple timeframes such as: 4h, 1d, 1w or even 1m. We can scrape the tweets with the price at the moment of the tweet + sentiment (newly generated). Then we can see if the mentioned ticker changed over time since the tweet.

[StephanAkkerman]

I think this will be better as a separate repo, also updating the scraper code and maybe publishing that here too

[StephanAkkerman]

Interesting but too hard, how do we know the prices etc. after someone tweeting. How useful would this information be in the end?