Sign into an existing account

Unless his LTC or Avalanche bots turn a profit. Getting them trained right is rather difficult though.

Interesting. So an RL approach. Maybe just like the GPTs all used supervised and are now fine-tuning with RL, maybe I could do something like that.

I guess maybe that's a better scoring system. Don't just perform. Perform better than what other bots can get me in the same period.

It's an interesting idea to put several separated supervise learning AIs into a zoo.

Honestly I'm trying to figure out if an RL approach is better even from the ground up for this. The issue is that predicting future prices and acting on it is not as clear cut as it seems. There is intricacy in measuring the performance of a trade in that sure it's good if something you bought is more valuable a day from now, but what of 48 hours from then, what of an hour from then. How does a trade affect outcomes within a concert of trades. Did it compete with the capacity to make another more valuable trade. Did it aid in making other trades.

It gets gritty when you actually try to measure performance of each trade and generate a real number for it and you realize some of the slight logical flaws on simply acting on what a projected price will be. So an outcome based approach might be better completely. That's RL territory.