Why do I want to have more model slots?
Although it is not an exact science how long you have to gather live performance to assess, it takes a significant amount of time of live predictions to judge if a model is worth keeping or not.
When you look at my top TC model you notice a gap and a comment that says that something changed in round 305. I have discontinued that model a few weeks before TC became a thing. So I am unable to judge which model makes sense based on computable metrics performance ==> I want to be able to test many ideas live
At any given point in time there is a myriad of tabs open with papers, github repos, blog posts about innovation which I want to try out. Now, in all honesty, the biggest reason I have not yet processed all of these ideas is not my lack of free model slots, but it would definitely help.
- I have a quite diverse staking policy, which fills slots with production models rather than testing
- A few of my slots are investments from family & friends in my predictions
- I have a hard time “killing” a model
- I wonder how many other people a full since months like I am. Maybe I am so queer that my problem is unique. Afaik there is no way from the API to get account-specific information to make a statistic, but if anybody from the team where to share stats on average model slots I would be interested
How many slots would I want? I remember Anson saying that he wants to remove the limit entirely. That is not what I am asking for! I can imagine Numerai would have to safeguard for a lot of trolling and attacks. I would be happy with a lot less than infinite slots.