The model now includes a new engineered stat which measures the highest and lowest values in each stat column and also compared their current stat to their peak and valley. This almost quadruples the model’s total stats. Accuracy increased from 68.8% to 71.1%. Return On Investment has increased from 21% to 28% for the last 8 months of fights.
The model uses a couple dozen stats scraped from ufcstats.com then chains a bunch of modifiers to those stats like average, recent average, change since last fight, etc. A new category of modifier stat was added: stat_peak, stat_valley, current_stat_divided_by_peak, current_stat_divided_by_valley. So a stat like recent_avg_change_takedown_differential will be turned into another column of stats, recent_avg_change_takedown_differential_peak.