Machine Learning Part 2
As of January 15 (today) I've had a look at the AI players' battle statistics to get an idea of how well they are performing.
In my opinion they are performing admirably well, more so now that one of my users won the campaign.
The AI players in my game, as discussed before, improve over time as human players play the game and the AI learns from the human players.
As such their percentage chance of victory in any battle hovers around 66%, about two thirds of the time the AI player will win the battle. This is great - if a football team had a 66% chance of victory they'd be over the moon.
The reason for the high figures is because the AI in my game learns from other players. When a human player does well the AI player copies the flight behaviours and settings of the quality human players and tweaks the settings slightly to add a bit of variety to the settings.
As a result, with one of my users playing the full campaign and winning easily, the AI has now improved and as I've seen in my back end stats - it is very difficult to win against now.
So the AI for my game learns - and it learns quickly. It has rapidly become a tough cookie to beat.
Well done AI opponents, I am well pleased with your learning. A+
AI Faction | Battles Won | Battles Fought | Percentage Mars | 63 | 91 | 69 AioSenti | 58 | 85 | 68 EarthHope | 44 | 66 | 66 Hushami | 56 | 85 | 65 Outlanders | 51 | 82 | 62