Trident

• 

  shots on target evolution

• 

  kicks evolution over time

• 

  long passes evolution over time

• 

  player risk evolution over time

• 

  successful free kick vs penalty over time

• 

  Heat map Weakness of team

Trident

A Three pronged solution to US Soccer

Inspiration

Inspired by creating a digital twin for the manager to analyze players and create strategies.

What it does

Trident offers data-driven solutions to make tough decision on the field with live data, past trends and dynamic learned AI models. The assistant helps identify the important relevant attributes that define a player's performance on the field. It also observes and learns trends in playing styles over time as it evolves. This can be used to improve the squad by bringing in fresh faces and experienced players according to the team requirements. The digital twin is capable of analyzing the opponent's weakness and strengths helping us determine right strategies to exploit them. Finally, it can also understand how a player performs in a match and recommend performance training for this player while identifying potential weaker players in the team.

How we built it

After much discussion, we identified the core features that would be useful in solving our problem statements. One of the key features was the overall ratings of a player calculated for each match. This enabled a more wholesome metric to determine the player's form through the matches. Quantifying this attribute allowed us to use it to determine his/her performance over the seasons. We did this by:

1. Creating models using algorithms like MARS, RFRegressors, Lasso Regression, XGBoost and LSTM-FCN ( LSTM skipped due to data size and being a black box model) where we learned what constitutes a good defender, good midfielder etc.
2. Using this feature weights, we also were able to visualize how the attributes' performance varies over the years
3. We used the feature weights for these attributes to calculate performance for each player in each match
4. We also analyzed the trends of these attribute over the seasons.
5. Using the event parameters, we were able to visualize the succesful goals against opponents and the steps to exploit these weaknesses.
6. Scraping the FIFA index website

Accomplishments that we are proud of

As a team, we were able to create new features that formed the crux of our solution allowing us to solve multiple challenging problems. Working hard independently but with clear focus, we were able to complete and integrate our codes effectively making great strides towards realizing our broader vision. Using various machine learning, deep learning and fuzzy logic algorithms, we were able to create various models in a short span of time to develop interesting solutions.

What we learned

Using analytics in Sports is difficult due to lack of open source data and this hackathon solved exaclty this problem. Working on this project pushed us to our limits to think beyond the usual data science solutions and develop interesting insights from data we've never worked on before. We also learned web scraping and time series techniques in detail to link data and analyze trends respectively.

What's next for Trident

Trident can evolve better over time integrating with soccer datasets assisting the manager to create intelligent decisions. Down the roadmap, we plan to integerage automatic player substitutions, real time dynamic strategy and effective scouting.

Built With

• python