- September 8, 2019 at 10:34 am #5736kwParticipant
The ML is a bit of a black box to me. Rather than just pretend it’s “magic” I would prefer to have a deeper understanding of the way it works to gain more confidence in its performance.
For example, is the ML model at all aware of this big triangle that is forming on BTC-USD? Does is realize that once it breaks, it could lead to a big move to the upside or downside?
Are there any articles that explain more deeply how the Crypto-ML model works under the hood? If so, could you please point me to them? If not, would it be possible to write such an article?
- September 9, 2019 at 1:45 pm #5747JustinModerator
Hi @wahnker, we get enough questions along this line that I think it would be best to write up an article in more detail. So I’ve added that to the content calendar and we’ll get something that will hopefully be informative.
What I can give you here is, yes, the ML is aware to traditional technical analysis. One of the early steps was to feed it as many TA indicators as possible. Using ML-oriented statistics, it shortlisted the ones that have value. With that basic framework in place, it now regularly adjusts the values for the indicators to best adapt to the market conditions and recent data.
What ends up happening is it generates its own support and resistance lines. These lines oscillate. What you’re seeing with the Trade Meter is the price action moving closer and further from these lines.
And yes, you definitely see different behavior in these triangle-type formations.
You will see the ML open trades when price hits the resistance line (not support). While this can results in several losing trades in a row (as price bounces back down), the math seems to indicate that it’s worth trying in case price *does* break out positively.
The ML seems to have determined the opposite move is risky (trying to trade off of support lines in a consolidating market). Specifically, while you might get a small bounce up, you might also get a big/fast drop down and face an unacceptable loss.
This approach is more akin to momentum trading than reversal trading.
In fact, there was a recent discussion along these lines you can check out here: https://crypto-ml.com/community/topic/ml-bot-is-not-handling-this-chop-well/#post-5599
Overall, I’m very curious to see how the market will play out. It feels like things are weakening quite a bit.
And stay posted for more detail. Does this help give you some additional detail for the time being?
- September 9, 2019 at 7:35 pm #5760kwParticipant
Yes, what you wrote by itself is very enlightening! Looking forward to the article, thanks!
- You must be logged in to reply to this topic.