Competition link

My approach was to take data from high-level agent games and train a keras regressor to predict each bandit's true payout rate.
My agent then picked the bandit with the highest expected payout at each step.

This challenge gives very little data to work with, but I improved my results by supplementing the available data with pull and win streaks.

• step number
• number of pulls
• number of opponent pulls
• reward - total reward for the agent
• streak - number of times the agent has pulled this in a row
• opponent streak - number of times the opponent has pulled in a row
• win streak - number of times this agent has given a reward in a row

• Thompson Sampling
• Bayesian Upper Confidence Bound
• sklearn regression
• keras regression
• sklearn/keras ensemble
• keras classifier
• pull vegas, only used for training. source

My main notebook file, this contains all the code to train agents and run simulations.

This notebook downloads agents, parses them into data frames, and saves them.

Based on https://www.kaggle.com/masatomatsui/santa-episode-scraper

Some basic statistics functions to analyze game results. Given a list of (p1, p2) scores it will calculate the following:

• Wins for each agent
• 95% confidence interval for player 1 win probability
• Likelihood of superiority

I added my own simulator.py code to run games in parallel and print out agent statistics.

compare_agents will play two agents head-to-head and stop early if one agent is clearly better rank_agents efficiently ranks a list of agents using merge sort round_robin takes a list of agents, plays them all against each other, and prints a grid of win percentages.

Bayesian UCB - https://lilianweng.github.io/lil-log/2018/01/23/the-multi-armed-bandit-problem-and-its-solutions.html
Thompson Sampling - https://www.kaggle.com/ilialar/simple-multi-armed-bandit
Training data collection - https://www.kaggle.com/masatomatsui/santa-episode-scraper
Decision Tree Regressor - https://www.kaggle.com/lebroschar/1000-greedy-decision-tree-model
Ray support - https://www.kaggle.com/nigelcarpenter/parallel-processing-agent-trials-using-ray
Pull Vegas - https://www.kaggle.com/a763337092/pull-vegas-slot-machines-add-weaken-rate-continue5