- March 27, 2020 at 1:38 pm #8051adminKeymaster
As of 3:30pm EST, March 27, the BCH signal system has been upgraded to release 5.0, which uses Deep Neural Network machine learning.
More detail about the change here: https://crypto-ml.com/blog/machine-learning-upgrade-to-5-0-deep-neural-networks/
The precision of the model is 0.92, which is the highest of the models.
Both the 4.0 and 5.0 signals were in a “SELL” status.
We expect BCH 5.0 to issue 4.5 trades per month on average. This can and will change, but can stand as a guideline.
No action is needed on your end.
Please let us know of any questions.
- March 31, 2020 at 7:05 pm #8108memecosapoParticipant
Hi, I’m considering migrating to auto-trader. 0.92 is very impressive, so I’m wondering if this would be a better pair than BTC. But, again, we’re approaching the halving…
Any reasons why BCH has a better result than other coins? Is it more predictable or the model is improved compared to other pairs? If you had applied this model (back testing) since the beginning of auto trader, what would the results be?
Thanks in advance,
- April 3, 2020 at 12:26 pm #8157JustinModerator
Yes, the .92 is extremely good. However, we have since found that we can reduce the averaging of the inputs (in terms of comparing current price to the predicted price 24-hours from now as described here), which increases the speed at which we can react to changes.
With that change, the models actually drop in precision a bit (since there is now more noise and fluctuation) but it translates better in terms of backtested trade results.
That’s a long-winded way of saying the signals are currently .81 for BCH, .85 for BTC, and .88 for ETH. Despite being lower than .92 we still expect better trade results when we consider the entire chain.
Ultimately, when we look at the end-to-end system, we will only roll out an update if it can get better results even if the precision drops a bit.
Let me know if that helps or if I managed to confuse things. 🙂
- You must be logged in to reply to this topic.