I didn't think to provide an example of AI work here, my mistake.
Stop loss is applied according to ATR (Average True Range).
Transactions in On-chain data are filtered by median, MAD and 95th percentile... and so on with each agent.
Then all the data is sorted, filtered and processed by different ML models.
Here how the result looks:

Hey OP. Thanks for the information. I'll try to apply it and we'll know how all of these indicators perform with market conditions that more or less make us think.