Example predictions - am i missing something?

Hi Everyone,

Been taking part in the Numerai tournament for a couple of months now and have found it a good learning experience which has already helped me in my work. Nonetheless, I think a large portion of us got into Numerai with the hope of making a bit of passive income whilst doing something pretty cool that we enjoy.

With this in mind… Why shouldn’t someone just submit the example predictions. It appears as though they perform relatively well (above 65th percentile). I know the MMC will obvious be near 0 but if you just stake 1x cor… you’re laughing, no?

Don’t get me wrong, I plan to continue making my own models, like I said, I enjoy it and it has already helped me in my work (I am a systems biologist). I’m just confused as to why the majority of people aren’t just staking and submitting the example predictions…? or is there something I have missed or am unaware of?

In my opinion, Numerai is limiting their meta model performance by supplying these example preds. If they were to only supply example scripts then at least they would be using people for their computing power and randomness may contribute to their meta model. As it is, everyone could submit the example preds, get a relatively good cor, causing a non-existent metamodel for Numerai.

Again, I may have missed something but I would like to hear other peoples thoughts.

P.s. who else codes in R?.. everyone seems to be a pythonian!

1 Like

There was a lot of talk about this on the latest fireside chat. I suggest you watch it as most of your questions are answered there: Numerai Fireside Chat with Richard Craib (founder and CEO of Numerai) and Anson Chu (CTO) | 2021 Q3 - YouTube

Edit: TLDR: Numerai are aware of this issue and don’t want to reward people for only submitting the example predictions and will introduce new payout methods in the future to counteract this.

3 Likes

Very helpful. Thank you

I code in R! feel free to send me a message on the rocket chat, twitter, or here, if you ever wanna talk shop :slight_smile:

FYI, for R discussion on RC:
https://community.numer.ai/channel/cran

1 Like

Numerai might stop providing example files at the end. Or another trying is just 0.5 for all the ticker and check the results in the long run.

1 Like