Discussing the article: "Quantum Computing and Gradient Boosting in EURUSD Trading"

New comment
 

Check out the new article: Quantum Computing and Gradient Boosting in EURUSD Trading.

The article describes the practical implementation of a hybrid algorithmic trading system that combines quantum computing (IBM Qiskit) and gradient boosting (CatBoost) to predict movements in the EURUSD pair on the hourly time frame. The system extracts four unique quantum features from a probability distribution across 256 states using eight qubits and, in combination with classical indicators and delta encoding of time categories, achieves 62% accuracy on 15,000 candlesticks.

Alexander closes his laptop at 2:37 a.m. On the screen are numbers that look like a verdict: an accuracy of 51.2%. Three months of work, hundreds of hours of debugging, an LSTM architecture with 128 neurons, three layers, regularization, and state-of-the-art optimizers. The result is hardly any different from flipping a coin, but he still has to make a living somehow.

He adds MACD. Then RSI. Then Bollinger Bands and Stochastic. The model overfits, and accuracy drops to 47.8%. He simplifies the architecture: two layers, 64 neurons. Accuracy returns to 50.3%. A statistical zero. The market is laughing at artificial intelligence.

At four in the morning, he realizes: the problem is not the architecture. The problem lies in the very nature of the data. Traditional features look backward. They see that the price closed at 1.1050, but they do not see that while this candle was forming, there was a probability distribution — a 30% chance of 1.1060, a 25% chance of 1.1040, and a 20% chance of 1.1050. The model is trained on collapsed states, while the market operates on superpositions.


Author: Yevgeniy Koshtenko

[Deleted]  
And how can the laws of quantum mechanics be applied to macroscopic objects – or even to things that aren’t quite objects? 🙄 And who is Alexander – Alexander Schrödinger? 😁
 
Maxim Dmitrievsky #:
And how can the laws of quantum mechanics be applied to macroscopic objects – and even to things that aren’t quite objects? 🙄 And who is Alexander – Alexander Schrödinger? 😁
I reckon you’ll definitely get an answer here:

Don’t go too far
 
Maxim Dmitrievsky #:
And how can the laws of quantum mechanics be applied to macroscopic objects – or even things that aren’t quite objects? 🙄 And who is Alexander – Alexander Schrödinger? 😁
It’s more a matter of probabilistic multidimensional encoding via the simulation of quantum circuits. Alexander is a fictional character from the article – a composite image of a trader. Not Schrödinger))) And not even his cat)))
 
Thank you for the article – it’s very interesting. How do you save the trained models? I’d be interested to see how a model trained on one pair or timeframe would perform in a different market.
 
I ran the model ‘as is’ and got an accuracy of ~60 per cent. I removed the quantum features entirely and ended up with 61 per cent. All in all, it’s a solid CatBoost model, but these quantum features are out of place here. These features are linearly dependent on one another; at the very least, Q_Variance and Q_Dominant have a coefficient of 1, whilst the entropy for both is -1. Just one of them would have been enough as an additional feature to provide further information. A good, informative article – thank you!
 
I take that back. It’s this line that’s turning the whole code into rubbish. The model is learning to predict the past based on the future. The corrected model achieves 52 per cent accuracy. The author was asleep whilst writing this :D 
# Remove NaN values and return the data
    data = data.dropna().reset_index(drop=True)
    data['target'] = target[data.index]
    
    return data.dropna().reset_index(drop=True)
 
Артем Резанов #:
I take that back. It’s this line that’s turning the whole code into rubbish. The model is learning to predict the past based on the future. The corrected model achieves 52 per cent accuracy. The author was asleep whilst writing this :D
Good afternoon. Initially, the plan was just to train the LLM – I assumed that, thanks to the generated dialogues and the dataset being in the form of dialogues, it would recognise and generalise patterns without having to memorise them.) Later, layers in the form of Qiskit and CatBoost were added. Future articles will cover the walk-forward mode separately; I’ve already written it, but haven’t had time to include it in the article yet :)
New comment