Currently, the IT sphere is experiencing a boom in neural networks and machine learning. Machine learning is widely used in various fields and is intended to replace the human brain for solving complex problems of classification and prediction.
TensorBot is a self-learning trading system with built-in adaptive strategy and a neural network that independently studies market patterns, remembers them and trades. This principle allows you to "train" a robot for trading on any financial instrument and any timeframe without any restrictions. You can trade FOREX, CFD, Commodities, Stocks, Futures, Indexes and Crypto currencies.
Training is done in the strategy optimizer using the MQL5 Cloud Network. Thousands of computers around the world will be employed to find the best strategy for you, and the entire learning process will take no more than 20 minutes.
Markets are changing, so you have to constantly modify strategies. The neural network allows you to get rid of the constant change in strategy. Simply re-train the system for the current market situation.
Advantages of the trading system:
- Adapts to any financial instruments and timeframes.
- High learning speed and ease of use.
- Flexible settings that allow you to scalp or trade positively depending on a timeframe.
- Not sensitive to broker's trading conditions, you can start trading with a minimum deposit sufficient to open a deal.
List of parameters to be optimized:
- Tensor stack size
- Tensor returns predictor
- Tensor LARS predictor
- Tensor LARS derivative
* The above parameters are optimized from 2 to 150 in increments of 2
- Trailing is optimized depending on the selected symbol and timeframe. For example, for EURUSD M15 from 50 to 450 with a step of 5 points.
- Break even - transfer of the position to the breakeven, for EURUSD M15 from 50 to 300 with a step of 5 points.
- Progressive lot - use lot from the size of the deposit, not optimized.
- Use fixed lot to - use the fixed position size. If "0", Progressive lot is used. This option is also not optimized.
- Magic for orders not optimized
Weights of neuron:
- 8 weights of the neuron, each need to be optimized from 0 to 100 in steps of 1.
The learning process of the system is shown in detail in the video example. It is recommended to re-train the system in proportion not less than 4\1. For example, for EURUSD M15, it is necessary to train a robot for the last 1-1.5 months, after which to trade a week, then again re-train. That is, trade should be conducted, in 4 times less than the training period. It is strongly recommended that you follow the above recommendations. You can experiment with optimization intervals for each parameter and find your unique strategy.