I was just looking around and found this guy. Back to back max TC 336 337. Amazing TC since 332 and not shy about the stake
I think the sad part is that all other metrics CORR CORRJ60 (why do we even have this?) FNC FNCv3 are no good explanation for TC given a single model. Means that you have to try shit out and hope it works, basically gradient-free reinforcement learning.
4 Likes
Yes, you can buy it: https://numerbay.ai/product/numerai-predictions/paul_the_0ct0pus
I am not the owner or creator, so I have no idea about what ML methods it uses.
2 Likes
Congrats on your good fortune! Thank you!
How did you get it then?
He bought it and so can you. This is the creator of the model: https://numerbay.ai/product/numerai-predictions/paul_the_0ct0pus
1 Like
That model is nuts. I wonder if they did the residual fitting and found a corner of the dataset that the meta model neglects/overemphasizes.
That TC is just so high and consistent for so many rounds… like my TC goes from 5th to 90th to 15th percentile over 3 rounds…
Like if that is random, what are the odds of hitting 99%+ so many times?
It works and the Numerbay stakes will grow until it doesn’t perform. And given the nature of momentum chasing, at that point there will be the most amount of NMR staked on it. When we have screened for # of TC rounds positive, we have not found a consistent relationship to future TC. That said, performance characteristics are the info that’s available and its sure easy to analyze! Given our recent Corr pain and the seeming evolution of TC models, we are joining in the TC hunt and are adding Paul because his TC history is impossible to ignore. Last time we did that we ran smack into that model’s downturn. This time => lower TC leverage and much more diversified. TC offers tremendous diversification risk reduction = something all should strive to use.
R349 Paul stake up to 10,894
5 Likes