Much to my surprise, Jerome has spent a weekend alone in his flat researching some trading ideas. He is terribly proud of his hard work, and no doubt recognises my surprise when he calls me Sunday afternoon to have a run through it.
By his own admission, his approach is heavily data-mined and the system spec is full of blatantly ad-hoc rules to avoid outcomes he doesn't want. Also he appears to have to work the system 24 hours a day, which is a bit of a downside. But there is a germ of an interesting idea inside the work.
Over the subsequent week, I have had a close look at what he has come up with. Most of his ad-hoc rules are virtually untestable, being in place to handle some specific problem on the day. Other bits lack harmony with the whole. I have probably got rid of more than 90% of his original system spec, but a core idea does remain. And on first sight, there may well be something in this.
So perhaps we will get a slow system to compliment our faster, current approach. that could be extremely valuable.