2 Challenging Reasons why AI is not Adopted by a Lot of People for Trading yet
Couple of weeks ago, I develop an AI model to automate trading processes. The performance is quite satisfying to give it a try.
It is open source, everyone can use it, and it’s easy to set up, but still there’s a problem that AI can’t solve, here’s why:
1. The data is oil and oil is crazy expensive
AI needs data for learning, and not all exchanges provide crypto or stock price historical data. Fortunately, there’s an incredible python package called yfinance, it’s near to saving us.
Netrade is an AI image-based model, it’s learned from the chart pattern and candle stick pattern images like this
But how do we create that image from real-time data? Alright let’s find out.
First we need to grab the historical data. In this example, we will be using yfinance. In this example below, I grab the bitcoin historical data. Netrade is an AI image-based model, it’s learned from the chart pattern and candle stick pattern images like this
Second, we need to visualize that data. To do that, we need to install netrade. First, run this commandpip3 and install netrade. For the chart pattern, let’s take the last 50 candle stick bar because I think it’s enough to capture the pattern.
And to create a candlestick image let’s take just the last 3 candlestick bar
Alright the data is ready. What we need to do is just follow the instruction below
But what exactly the problem is? Well, the bitcoin data that I take using yfinance is from coinmarketcap, not from our desired exchange. In my country, there are a couple of exchanges like indodax, pintu, luno, etc. in your country, maybe, there’s binance or another exchange.
If we take a look at the price from each exchange, the price is very different, and also, for the chart pattern, it’s different too. All of it because the amount of user from each exchange are different, so the purchasing power is also different. But why don’t we just take from our desire exchange? Not all exchanges provide historical data to the public
2. To build a bridge is also crazy expensive
If you have software engineering background, I believe you know the term of API (application programming interface), yeah, that’s the bridge between exchange and us as a user to use their data. If we implement this AI model for real-time data, there will be exploded request to the exchange’s server, and maybe, I don’t know, that’s the reason why not all exchange provide the API for make transaction trough our app or just use their data.
The challenging part is there’s an exchange that provides the historical data, but they don’t provide us API to make a transaction through our app. On the other side, there’s an exchange that provides API to make a transaction through our app, but they doesn’t provide the historical data. As I mentioned above, the price between exchange is the difference, so we can not use historical data. Let’s say we use API to grab the price data from Binance and use API to sell bitcoin from Kraken exchange. We will lose our money if we do that. I’ve tried it, by the way.
Conclusion
I think that’s the problem why AI can’t apply yet by a lot of people in the field of trading, at least for now, but I know there are some platforms out there that already connect their AI to the exchange and do a transaction ( I don’t know how ), AI is great to recognize the price pattern and predict is the price will going up or down AI great for doing that, but AI can’t solve the problem that correlated to the regularization.
