The new version of farmertron that I entered into Bingmann's AI league has some serious flaws, and I'll be suprised if it gets anywhere with all the updated competition. That'll teach me to spend only one night updating it for a new tournament. I've spent the last few days seriously upgrading it.

The 1v1 I knew was screwed, it was only thrown in as an experiment, it still thinks it is the shot in a 2v2 game and makes only 50 villagers and researches cartography! This sets it back no end, with a slower castle time and less resources. It rarely makes the 75 champions it needs to make use of the low villager count.

The 2v2 is better than the AI Wars version, but I made a mistake that is going to cost me as so many of the AI scripts in Bingmann's league make use of Aztecs and Mayans. I'd programmed in a simple rush response routine which works very effectively, by starting the attack as soon as it finds units in it's base it will pump out infantry until the opposition has left and then it carries on as normal. This works great, but I put 'knight-line' as a response check instead of 'eagle-warrior-line'. This means that it doesn't respond to eagle warrior rushes with the perfect counter - infantry!

In tests against coteti's Torisan, v1.1 fails miserably with that flaw. Change the statement from knight-line to eagle-warrior-line and it wins consistently - in fact it hasn't lost a single test game since the change.

You can download farmertronv1.1 here, but v1.2 or whatever should be out before the next tournament phase and it will be a lot better. So far I've updated it to deal reasonably effectively with Nasarasu's monk defence as well as working on a new style of attack that works better than 'attack-now'.

Any feedback on farmertron would always be welcomed, and you can email me at farmer@achtung.co.uk if you find it doing anything stupid, which it is guaranteed to do at some point as it's only a dumb AI script .