[ptero]

In a kind-of defense of such methods: such technologies do seem crappy when you look at them today, but building good products with bad tools is pretty much in the job description for engineers and even more for experimental scientists. Check a good experimental scientist record of 100-300 years ago and you will be amazed what they did with the tools of the time.

This is not an apples to apples comparison to the spider fight world of trading, but there would still be a lot of ingenuity there. And a simple way to limit risk is to ruthlessly review the results, both automatically and manually. Keep going while it works as expected and stop/adjust if there is a major glitch.