Neural network. How to predict the victory of one of the three teams?

Question

Python

Hello.
I have been trying to make a neural network for a long time that will predict the victory of one of the three teams.
So, imagine that there are matches in which 3 teams take part, and only one can win (there are no draws).
We need to predict the victory of one of the 3 teams.
To train the neural network, the characteristics of each team will be introduced (for example, the percentage of the team winning, what place in the top, etc.; some text values?), as well as the number of the team that will win.
After reviewing many sites, I did not find such a performance of the neural network. Usually, something is simply determined from the input data (for example, we enter the height, weight, temperature of the animal, and it shows us what kind of animal it is). And we should have some kind of comparison between 3 teams.
I beg you to help and do not enter complex terminology in response, since I am new to this. If there are ready-made implementations (open libraries), then it would be nice if you dropped a link to them.
Thanks in advance

Answer 1

The task of predicting the winner of a match is no different from the task of classifying an animal. Only in the sports problem the object of classification is a "match" which includes a description of the teams that took part in it. The result of the classification - respectively, not "cats" / "dogs", but "team 1 victory" / "team 2 victory".
Corresponding "ready-made implementations" are in Python, and in R, and in almost any library for Machine Learning.

Answer 2

if you take it small , it’s
better to take the stock exchange right away,
or at least learn the theory, independent events there, finite automata, etc.

Answer 3

The simplest option with neural networks is to simply concatenate the characteristic vectors of these three commands, and then, as usual, fully connected layers, at the end of softmax and 3 outputs 0/1 for each command.
But in general, such tasks seem to be much better solved by gradient boosting.

Answer 4

In my opinion, it is most correct in this setting to solve the problem of restoring regression. That is, determining the probability of winning one of the participants.
Send me a link to the data and I'll see how I can help.