And with a lot delay the wrapup of last season. My Bot Aicu actually performed horribly. It finished just in Champion I at 4753. It had pretty extreme Ups and Downs last season. Mainly because I decided to experiment heavily with monster distributions.

As you can see, my bot just touched the 5k level last season. The other seasons it finished around 5.2k This season I set everything back to the basics and as you can see, the incline is a lot more stable. So, how did aicu-chan do?

Something like that, much better than last season. I haven't made big changes to aicu-chan that way it performed more like last season. It reached silver III with it's lvl 1 account.

Debugging

Like with many software development projects, debugging takes a large role in this one as well. I thought that by now, I had fixed most, if not all of the bugs in my Simulation. The simulation is the core part of my bot, and without a good simulation, the bot cant perform optimal. Thanks to @bubke I discovered today two more bugs in my simulation. And not even minor ones at that. I forgot to apply the damage buff from non summoner abilities like "Inspire". Without that, the damage boosting ability was completely ignored in my simulations. The next major bug was the incorrect calculation of the damage reduction from Void and Shield. I was looking for a quick way to calculate it and I noticed that the new damage is always the rounded down value of damage times two thirds. In a moment of derp I decided to apply that directly as the damage modifcation. So a melee damage of 3 was not reduced to 2, but 1. Because 3*(2/3) rounded down equals 2. 3 -2 -> 1. But It should have been 3-1 -> 2. Oh well. sometimes you can make major mistakes without even noticing them. I (hopefully) fixed those two bugs and I've pushed them to my github repository I'm curious if those changes in the simulation reflect themselves in the performance of my two bots. If not, then it can't be changed. But it would be nice if it did. Aside from that, I'll look a bit into deep reinforcement learning. It might actually perform pretty well with a game, where tracking the best teams already guarantees you a spot in the top 10 with a max level deck.

Giveaway

Like last time, this time I'm also doing a giveaway. The rules are as usual. If you comment below within the payout time I enter you in the random pool from which I draw the winners. The cards this time are: