- July 30, 2020 at 1:11 pm #9453martyfParticipant
One idea I had bouncing around is that the recent move was so out of bounds for the ML that it may signify what that literally means – this new phase is different than what the machine has seen before.
again, this is why i was surprised things went from being adaptive and always learning to using a static model – if i remember that correctly.
not to say that a dynamic model would have snagged it.
right now im watching the bot behavior to see how it handles things. I mean, capturing ranging activity can be good enough to make a living if it does it flawlessly. So presumably, if this is just your standard market with standard ranging activity, the bot should catch em. Maybe the transition wasnt caught, but the ranging might.
- August 1, 2020 at 8:49 am #9459matthew55Participant
I’d like to know what the admins think about it, as well 🙂
- August 1, 2020 at 11:43 am #9463JustinModerator
@martyf, you actually have nailed it pretty well.
In technical analysis, there are two major classes of indicators:
1. Momentum indicators (like MACD) that hop in on existing moves. These are typically great in trending markets but bad in sideways markets.
2. Reversal indicators (that issue overbought/oversold) that counter existing moves. These are typically great in range-bound or slowly decaying markets.
Over the last couple of years (2018 through now) price movement has fallen much more into the 2nd category. As the ML continues to look at “what works,” it’s leaning toward that 2nd type of behavior–counter trading.
Our system is designed to be flexible enough to recognize the large pattern and move between these modes as appropriate, but clearly it didn’t happen fast enough to catch this rally.
A couple of fixes we’re working:
1. Identifying additional data points to help identify these changes. That may be looking at more complex trade setups.
2. Continuing to feed new data into the learning process. So whereas coming into this, a strong trending market was a distant memory, now it’s upfront and strong for the ML to consider.
Does that make sense?
- You must be logged in to reply to this topic.