Why does my data collecting algo work so well?

So i wanted to look into ml for terminal.to do so i needed some data. i coded an algo which does very little:it places 3 walls and 3 turrets in the middle of the board on the last paths it got scored on. after that it spends all left sp on a basic pre determent structure of 3 turrest 3 walls and a lot of factories.every time it has a random amount of 10 to 15 Mps it stacks on the location thesy will take the least damage with nearly the same function as in the starterkit. i the uploaded it to see if it was ok. after a week or 2 it had surpassed al my other algos by at least 150 points. it also has the recording elements in it which slow it down quite a bit. If anyone had an idea why i would be very happy if you could tell me

I’m curious to how you are getting the data from your algo. Are you able to retrieve the recorded match from your algo after you’ve uploaded it?

no i just uploaded it to see if everything was working.my plan was to have it run on my computer against a random algo, and then exporting it as txt.then train some sort of tensorflow model on that.

I mean, if you want to do this, Terminal has some great api possibilities.
Just get you algos ID:

Then get your algos matches:

And then you can look at the single matches:

thank you that makes it easyer

so i just copy the list of maches into a json file, then use some for lopps to get the individual replays and train on that,and that works?