Hi - I’ve seen quite a few posts on here and Twitter about users putting a lot of work into their models and continuously making improvements to them. I’m curious, what are people doing / what does this entail?
For example, I did a rough analysis on everyone’s results in round 260 of the main tournament and it looks like there was more NMR burned than paid out for the first time in a few months (~2k net NMR burned). Even a huge chunk of the top 100 models lost NMR in round 260. So what does this mean for your model? How would you identify what went wrong and improve on it? And if you change your model, how do you know it will be better than your current one? (all of these questions are assuming you’re already happy with your val diagnostics)
I’m still fairly new, but so far my model has decent diagnostics on val and has done fine on live data. If I have a bad round, does that mean I should go back to try and build a model with better val diagnostics?