Is it worth investing in training 10mil+ T1 troops to make a trap account? Test how many troops get severely wounded for regular vs trap situations: (1) 1000 T5 vs 1000 T5 - control (2) 1000 T5 vs 1000 T5 + 9000 T1 - trap (3) 1000 T5 vs 100 T5 + 9000 T1 - same power as control (4) 1000 T5 vs 1900 T5 - same power as trap

(1) Time series EDA How does the aggregated stats fluctuate over time? : Aggregated stats converges towards mean. Use coefficient of variance? Year-to-year variations in the stats? : Sampling starts fresh every year. Can independence be assumed? Does a team's stats fluctuate year-to-year? By how much? : Repeated measures tests? (2) Stats comparisons between conferences Is WP affected by conference? Stats? : manova? (3) Per game model Needs multiple output model to predict scores for both winning and losing teams.