Discussing the article: "MQL5 Wizard Techniques you should know (Part 82): Using Patterns of TRIX and the WPR with DQN Reinforcement Learning"
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Check out the new article: MQL5 Wizard Techniques you should know (Part 82): Using Patterns of TRIX and the WPR with DQN Reinforcement Learning.
In the last article, we examined the pairing of Ichimoku and the ADX under an Inference Learning framework. For this piece we revisit, Reinforcement Learning when used with an indicator pairing we considered last in ‘Part 68’. The TRIX and Williams Percent Range. Our algorithm for this review will be the Quantile Regression DQN. As usual, we present this as a custom signal class designed for implementation with the MQL5 Wizard.
Typically, strategies that have evolved from these indicators tend to depend on fixed rules. For instance, these rules can be ‘buy when TRIX does cross above zero and the WPR is below -80’ or ‘sell when the TRIX peaks while the WPR is above -20’. Deterministic as they are, and being easy to back test, they present a common weakness, which is to assume a static relationship in a highly dynamic market. Because of this, their probability in being correct can deteriorate when prevalent conditions change, and these ‘absolute thresholds’ need to some adjustment.
This quagmire, then, does set the stage for reinforcement learning (RL). RL is a member of machine learning techniques that try to learn the suitable actions from experience or interacting with an environment. Instead of exclusively relying on preset indicator thresholds, an RL agent gets to explore different threshold levels, and thus adjust the trading decision-making process in order to maximize long-term rewards. For most traders, this would translate into a system that does not just follow rules, but rather one that adapts; on paper, at least.
One of the RL methods that seems promising, when engaged in trading, is the Deep Q Network (DQN). As opposed to supervised learning models, that try to map inputs to outputs directly, DQNs assess the value of going through with certain actions when presented with particular states. Within the context of trading, states can be features that are got from indicators, in a transformed pipeline, or binary format or even raw values. For this article, those indicators are the TRIX and WPR. The actions on the other hand can correspond to buying, selling, or remaining neutral. The DQN RL framework does enable these actions to be appraised and fine-tuned from experience as opposed to fixed arbitrary rules.
Author: Stephen Njuki