No, that’s a great question!
In general, one of the main areas we work on is loss control. If you dig back in our history, you’ll see the losses become less extreme over time. Our 4.x models are the best yet at controlling loss.
A bit more-
In the earlier versions of the models, we let the machine learning decide when to open and close positions. It turns out our ML had pretty high tolerance for losses. It’s possible that approach works in the long run, but it’s hard for us humans to stomach. 🙂
Our 3.0 models introduced stop losses and 3.5 introduced trailing stops to better protect profitable positions. Basically, we added that mechanism into the mix so the ML algorithms can play with them and choose how to set the limits.
Our stop losses are internal to our system and not set on the exchange. There are pros and cons to this. The main pro is it helps keep our system fully in sync with the activity in the account.
Now, with all of that said, if the market is in the midst of a flash drop, it’s possible to blast through a stop loss, which is probably what you saw in the 8% loss. That can happen even with a stop loss set on the exchange.
For more detail-
3.0 info: https://crypto-ml.com/blog/machine-learning-models-3-0-released-major-change/
3.5 info: https://crypto-ml.com/blog/crypto-trading-machine-learning-models-upgraded/