The plan for the rest of the year seems to be:
- Better tutorials and documentation to help new entrants
- Restore flexibility on staking
- Restore details on forecast performance
To quote the official words, from Slyfox/Anson Chu, Numerai's VP of Engineering (https://community.numer.ai/channel/feedback ,Nov 19 2019):
"I will have a more articulated roadmap towards the end of the year, for next year. For the rest of 2019, our focus is mostly incremental 1) creating a better "new user golden path" for growth (tutorials, docs) 2) simplifying rules around stake changes (allowing updates to stake changes etc) 3) adding back submission "final" scores as a first class citizen in the webpage (and maybe bring back round details page) metamodel contribution is probably going to come in Q1 2020 the thing i am primarily concerned about is how to introduce a new scoring/payout system without disrupting the current system"
"in my spare time, i also like to ponder what a "zero sum" tournament would look like. kinda like a poker tournament where everyone pays into the pot (house puts in big chunk to seed it of course) and then winner(s) take it all. I am thinking maybe we can start introducing "mini games" (kinda like how say League of Legends introduces minigames that only last for a while as experiments)"
"The other thing I am thinking is how to schedule changes/updates in general. Maybe we introduce this idea of "seasons" where we guaruantee that rules dont change within the season"
A few thoughts on the comments:
- For new entrants, the new tournament rules set in November won't reach its full effects until March of 2020 -- i.e., 20 weeks or roughly 100 business days. That is when a new entrant under the new tournament rules will finally be qualified for the leaderboard bonus -- a big part of the compensation. And the odds are good that the tournament rules will change either by then or before then. The official word is that Numerai will be mindful to minimize interruption and perhaps try to use "seasons" to freeze the rules for a set period of time.
- The zero sum tournament as a mini game idea sounds like a Kaggle tournament. One of the draws of Numerai over of Kaggle is that it is not a tournament where winner takes all. So I'm skeptical of a zero sum tournament. Also, even a "good" model on financial data can perform poorly over a large period of time. This was observed over the "burn periods" early in the year. The degree of volatility is large and therefore turns the sort of "zero sum" tournament described above into gambling. Note that Kaggle competition don't require staking -- putting in your own money to participate.
And finally, two bonus thoughts on the state of the tournament:
- Engagement measured by staking is way up for the new iteration of the tournament. It's at 23601.99NMR right now (Nov 25, 2019) or about $132,879, using the exchange rate of $5.63. Though with success I wonder if that's going to invite regulatory scrutiny. $132,879 isn't a lot of money per SEC standards, but that might change if Numerai continues to increase engagement. I have not crunch the number, but my sense is that a lot of the staking is coming from NMR that were given out in the past and that new money coming into the competition is still limited. It's still uncertain whether Numerai can increase engagement.
- Since I do participation in the competition and am doing quite well at the moment (still top 20 as of Nov 2019), I follow the forum to stay up to date. If you read through the forum comments over the past two months, you would get the sense that the Numerai website and the backend are buggy. People would report problems and bugs and then the staff go and fix it. The staff does fix things quickly and I don't remember anything bad resulted. But it highlights that Numerai is very much a startup company at its early stage.