This is honestly the most fun I've ever had coding. The possibilities are endless. FML if I can't make this shit work.
Past couple of days that I've been collecting this information, it's incredibly interesting how the information moves from market -> twitter -> market Today, I played $CAG during the morning's first consolidation and got out for $1.25/share profit but the interesting story is not that it was successful, but that the fucker kept going (+.50 from my exit). I began to analyze the twats and it looks like the second wind came around the time twatter picked up. So there is more work to do here in data collection. Something to the effect of: - Observe market scanner relative strength - I enter on first consolidation after relative strength established (been doing past couple of days - no idea whether this will work long term) - Keep an eye on the twatter - As twatter picks up price movement, sell into the strength In order for me to do this quantifiably, I need to automate the "keep an eye on the twatter" step. This can be done as follows: - For each symbol returned from the IB scanners, count the # of tweets for a given symbol per minute. I can probably do this with a query like ($SYMBOL1 or $SYMBOL2 or $SYMBOL3) and manually collate, to avoid going over twatter limits Yay, I found more fun stuff to do!
As an aside, I found an issue with ib_insync or TWS where the scanner stops working overnight. So it needs to be restarted in the AM. Not happy with that but it is what it is. I've set a cron job to do this around 4am.
Why SQLite is better than anything else in the world: Code: update scanner set parameters = (select json_set(parameters,"$.abovePrice",15) from scanner) where type = "IB";
Twitter search was proving to be a pain in the butt due to limitations on query complexity so I decided to go with the firehose. This is all the code for tracking symbols post IB-scan below. The other issue I had was that Twitter did not seem to let me search for so-called "cashtags" through the stream. A cashtag looks like $NKE. So I had to do a little hacking to get it to work. But, keeping things as simple as possible, it's doing the job. The only problem I can foresee going forward is that this stuff needs to kind of be realtime which means storing in the DB is "ok" but not ideal because the data flow will be world -> process -> database -> clients. However, in the spirit of doing things as simply as possible, I will just hit the database every time I need to do something until it sucks so hard I need to fix it.
No specific model recommendations but as a brand, I like Audio-Technica. No fuss Japanese engineering and very good value.