Optical character recognition was my bread and butter for automating games. I have been looking into algo trading for a bit, but I can lose my money just fine on my own for the moment. I haven't seen one person mention OCR as a method to circumvent API restrictions.

I'm not taking about training an RNN on image recognition. I'm talking using high level scripting to do pixel region color detection, OCR and keyboard/mouse hooks to buy directly from the commission free app and not the api.

I used to make bots for games, and I would use color detection on specific parts of the screen. Knowing the exact flow of user interface allowed me to infer what was happening in very complex situations with just a color code, at a specific coordinate. When some conditions were met it became beneficial to do one OCR pass on the current frame. That last bit turned said bots from crap to gold.

Maybe I'm overthinking it all, but at 60hz you'll beat any standard issue API in regards to server calls per minute. Including processing time. This would allow a sufficiently advanced setup to watch 120 different markets every second for no extra money.