There should be a stingy official text about the D2PREDICT service, but it will not be. When a project is done not out of mercantile considerations, but out of love for e-sports and IT technologies, with a certain amount of curiosity and fun, it turns out cool, fast and mega efficiently.
The D2PREDICT project is a powerful analytical tool for predicting the results of Dota 2 matches. Yes, machine learning technology is cool, and we used it.
The idea of the service is simple - generating predictions for Dota 2 matches before the game and in real-time, based on an analysis of data from previous tournaments. But behind the scenes remains the most difficult work of the prediction algorithm with its continuous training, so that the predictions of Dota 2 games are as accurate as possible.
D2PREDICT gives 2 types of predictions: prediction for pre-game and forecast for live matches. Moreover, the set of analyzed parameters for each of them is different, because our project is focused on the result.
For pre-game predictions, for example, folowing parameters are taken in calculation: team win rate, player win rate, total number of team matches and etc. For a live matches - the number of kills of the hero, increasing a gold per min, increasing XP per min and many others. We will not reveal all the cards here, of course. The high speed of data processing and analysis allows us to adjust the prediction during few minutes itself.
And now numbers:
- Over 35,000 Dota 2 matches analyzed for algorithm learning;
- About 100 factors are used to predict the outcome of the game;
- Predictions with 80% accuracy. This is wildest cool, and you know that.
Profit for users (for those who have read to the end):
Obvious: freebie - we absolutely free provide the most likely predictions for Dota 2 matches.
Profitable: how do you handle the received info - your choice. A just fun for your favorite team during a stream or the correct bet at a bookmaker with an increased chance of winning thanks to the D2PREDICT project - it is fully your decision.