So I know that it tells us how much my predictions correlate with the jerome 60 day targets. Beyond that, I am clueless, and I cannot stake on it so it strikes to me as a rather useless metric.
For some reason a custom loss function I used for some of my models seems to be good for CORRJ60 eventhough I did not use the jerome 60 day target during training, which made me curious.
Are there some future changes planned where this metric might be more useful?
They were moving towards staking on 60 day targets, but then backed off the idea, at least for now. (i.e. J60 was looking to be the new corr20 maybe) But for corr, being good on the J60 (and several other of the targets) is a good sign, and possibly worth training on. For TC, who knows?
Thanks for the reply , what news channel do I have to follow to not miss these kinds of somewhat important information?
I just pay attention to this forum and rocketchat – the rest from the quarterly fireside chats (they always talk a lot about upcoming plans there).