Mint. I’d tried to estimate the performance of sets of features over time but hadn’t thought of weighting them per se. I’m going to give it a try.
1 Like
Top models are simply lucky! Most of them…
Model’s rise, when the current era is favourable and then drop from top100.
You can’t show me one, that sticked in the top100 for half a year long.
That’s not a problem, but you have to be aware of this fact.
If you stay continously in the top 1000, you will get good annual profits. And that’s what really counts.
But it’s not possible to be continously be at the top of all chars.
20 Likes
Yeah now that you say it we can see the Guitargeek has dropped out of the top 100. Wild. However we see the nopaix models continue the great performance for quite a while now.
We’ve seen that over and over. Someone gets to the top and is dominating for a while and they seem unstoppable and then they drop and we haven’t yet seen of any them return to their throne once deposed. Not that I’d mind doing that myself. I’m sure guitargeek’s model will probably still do pretty good even if it isn’t #1.
7 Likes
Nopaixx finished with a flourish today alright. Impressive stuff.
1 Like
100%
High variance, luck.
If the model doesn’t make too much difference, then the key contribution factor will be remaining as feature and training data filtering (ERA).
1 Like