paper: Augur: FRC Rankings Predictor
 

Thread created automatically to discuss a document in CD-Media.

Augur: FRC Rankings Predictor by brennonbrimhall

Re: paper: Augur: FRC Rankings Predictor
 

Why do we need a rankings predictor?

A rankings predictor helps teams to visualize what teams will be seeded in the top 8. This helps teams cater their strategies and decide what features to showcase the next day, increasing their chances of being picked. Alternatively, teams that are predicted to seed can use this to predict other teams who may seed higher, predict whether or not 'scorched earth' tactics could be a possibility for the next day, and more.

Theory

A typical FRC match can have three distinct outcomes:

1. Red wins
2. Blue wins
3. Red and Blue tie.

Therefore, for n matches, we have 3^n possible combinations of outcomes. However, since the probability of a red and blue tie is so low, we choose to ignore it. This reduces the number of combinations to 2^n. Essentially, what we do is calculate all 2^n possibilities, compute the first 12 rankings for those possibilities (again, to save computation time), and then analyze them.

What about Tiebreakers?

Tiebreakers, like Assist Points, Auto Points, Truss and Catch Points, and Teleop Points are manually entered. Since there's no really good way to predict how these numbers will end up, we leave it to the human to predict these values.

Example Output

If you run main.py, you should get this output, which used data collected on April 4th, 2014 at 10:54. It's a simulation of the Silicon Valley Regional, with some matches omitted for the sake of computation time (I tried to judge which matches would have no affect on the top 8 without looking at how everything ended up shaking out on TBA).