Do you have a question not answered here? Use the contact form to submit your questions.

F.A.Q. - stock-forecasts.com

Who is behind stock-forecasts.com?

There is no company or investor behind stock-forecasts.com.
This is a purely private project; see the legal notice for more information.

What does stock-forecasts.com cost?

Stock-forecasts.com is completely free to use. There are currently no registration or usage fees.
However, there are plans to add advanced functions or resources which could be subject to fees.

Why is stock-forecasts.com still beta?

The project is currently in active development, both on the front end (this website) and on the back end. Identifying it as beta clearly indicates that there can and will be changes to the design and the functionality over time. The right to change, delete, or reset forecast configurations is also reserved.

F.A.Q. - Forecasts

Who creates the forecasts?

A separate neural network is created for each forecast configuration. Once the neural network is trained, it makes its predictions. This is a fully automated process.

What is predicted?

Share prices fluctuate. stock-forecasts.com currently predicts the closing price, i.e. the last price at which the stock will trade on any given market day. In addition to shares you can also forecast some stock indices, ETF funds, and currencies.

What is the neural network?

For general information about neural networks, please refer to the following Wikipedia article: Artificial neural network.
The neural networks of stock-forecasts.com have one or more stocks as an input layer. These are defined during configuration. Then there are several hidden layers. The output layer represents the prediction.

How does the neural network create its forecasts?

The neural network is trained with historical activity data. This training allows the neural network to learn certain schemata or structures within the data. The network then knows how the share price has behaved in the past and applies this knowledge to its predictions.

Do the predictions work?

Yes and no. The neural network prepares the predictions based on this training and the knowledge obtained. This can be correct, but major events, fluctuations, or even stock market crashes cannot be detected and predicted. In the end, the neural network is not a crystal ball. It cannot look into the future - it can only learn from past data.

How good are the predictions?

This depends on the point of view and the calculation basis. The self-developed "Prediction accuracy" indicator achieves 60%-70%.

Do I have any influence on the predictions?

A direct influence on the predictions is not possible.
However, indirect influence is possible. The neural network has many parameters, such as the number of neurons, the number of hidden layers, the training algorithm, the number of days on which the network is being trained retrospectively, etc. These parameters are defined with the configuration and currently cannot be changed manually.

When are forecasts issued?

New predictions are created from existing configurations as soon as the historical data from the previous day becomes available.
When new configurations are set up, the predictions are made shortly afterwards. Depending on how many configurations are created at the same time, this can take up to one hour. As a rule, new predictions should be available after 15 minutes.

What do the prediction stats mean?

  • Quantity - the number of predictions that have been made for this particular configuration
  • Prediction accuracy - self-developed KPI. The mean indicates how good the prediction is. This calculated from prediction value accuracy and other internal KPIs.
  • Value deviation - average deviation of the predicted value from the actual value.

What does "reset the neural network" mean?

The neural network is continuously trained every day with updated market data. From time to time, it may be useful to reset the network and restart the training. This means that old data is discarded and better predictions may result.

What are the different types of graphs?

  • Average graph - shows the mean of all predictions on a single graph.
  • Average graph of first prediction point - shows a single graph and uses the first prediction point from all predictions.
  • Average graph of latest prediction point - shows a single graph and uses the most recent prediction point from all predictions.
  • Last 3 predictions - shows the last 3 predictions.
  • Last 10 predictions - shows the last 10 predictions.
  • All predictions - shows all predictions.

What do the training data indicate?

The training data show how good the training of the neural network is. The training data graph should ideally match the market data graph. If this is not the case, it can be helpful to reset the neural network and completely redo the training.

F.A.Q. - Account

Forecast configuration: Why can't I find a certain stock?

Unfortunately, not all stocks can be supported at present. Your chosen stock may not be in our database and may not be eligible for forecasts at this time.

What does "Additional stock for the forecast" mean?

The neural network is trained with the historical data of the input stocks and generates the predicts based on these. The predictive stock is always used as the input stock. Other stocks may also be added to the input layer of the neural network. These will then directly influence the training and the predictions. Additional stocks can both improve the predictions but also worsen them. They should have some relationship to the prediction stock. This can be a competing company or a stock index, for example. If, for example, you want to forecast BMW stock, you could add VW and Mercedes stock or the DAX stock index to the mix.

Why can I only set up a limited number of configurations and only input a certain number of stocks per configuration?

The number of configurations and entries per configuration is capped by limited server resources. Depending on the number of users, these caps are raised or lowered. In the future, there may be way to raise these caps for a fee as a way to further finance the project. If you need more configurations or inputs per configuration, submit a request using the contact form.
Back to Top