- July 11, 2020 at 1:10 pm #9277JustinModerator
Hi @Julian, sorry you’ve been disappointed with the results.
In my opinion, the ETH model performance has been very strong. See below for the recent trades.
There is always plenty of opportunity to do better, but this system–and no other system–will perfectly capture every single opportunity. This is about trying to capture the high-probability trades consistently over time. Even though ML can be good, it is by no means perfect and there will always be unpredictable behavior in the markets.
As a last note, we are truly never satisfied with performance either. We are constantly actively developing and seeking ways to drive even better performance.
- July 12, 2020 at 9:40 am #9279LeoTheLionParticipant
Agreed on the performance, I have been happy as well. As noted, there are always improvements that can be made. Unsure on the root cause analysis that is done on your end, bringing up potential points of improvement (as observed on this trade that seemed to manifest from perhaps need for improving that detection of trickling patterns) to facilitate model updates or at least focus things on your end as it makes things immensely easier as I have noted with my own models and system development.
- July 15, 2020 at 11:08 am #9298JustinModerator
Hi @Julian, this is a great point. At minimum, we perform a weekly review of signals performance. When things didn’t go as expected, we do perform an RCA usually that same day. It’s an important practice.
But it also helps immensely when our customers point things out we may not be seeing. So I greatly appreciate the deep dives and feedback from your end. Please provide your thoughts at any time.
- July 8, 2020 at 10:37 am #9247LeoTheLionParticipant
I had a concern regarding the trade signals. Earlier this week there was a time where the currency was at 243 that provided around 4% gains. the price then trickled down to 235 at a loss 0.5% with sell at that point. The price then rose to 245 as of today to likely drop down rapidly over the next coming few days. I was trying to get additional information about the trade signals given the new feature of selling during a trickling down. The feature did not apply during this trade which perhaps is due to a more rapid trickle then what it is coded for. In addition the trade signals also appeared to have some deficits given that there could have been gains of net 9% versus -0.5%. Then combined with the prior trade with a gain of roughly 4% there could have been a gain of 13% versus roughly 3.5%.
- You must be logged in to reply to this topic.