Monte Carlo simulation? Have your CI server run millions of games in fast forward, and collect as many statistics as you can. When you make a change, check that

Monte Carlo simulation? Have your CI server run millions of games in fast forward, and collect as many statistics as you can. When you make a change, check that the stats that you expected to change did change, and that others stayed the same. Wouldn't detect extremely improbable things, though, like having a spear with a super-rare effect + having armour with another super-rare effect, since those would show up so infrequently that they wouldn't skew the overall stats much.