Inspiration
Our idea came from the pressure chess players face as the clock runs down. Even in a strong position, limited time can make it harder to assess a move and avoid a costly mistake.
What it does
We make the pressure visible by visualizing the probability of a player making a blunder, using a thermometer-styled bar.
How we built it
We found a large dataset of chess games and trained a neural network on it to predict when a person playing is likely to blunder based on the position, time on their clock, their elo, etc. 1 person worked on the extension UI. 1 person worked on the server. 1 person worked on training the neural network.
Challenges we ran into
Building efficient communication.
Accomplishments that we're proud of
UI is cool. Neural network actually works (somehow)
What we learned
Chess databases are huge. Download takes 2 hours (at least). Learnt to create web extensions.
What's next for Blundometer
Predict brilliant moves in addition to blunders! Add support for more chess platforms.
