[Proposal] Numerai Community Marketplace

Well, that would be fine. Or just have Numerai keep a history of submissions (timestamp & checksum) viewable by API so it can be seen when submissions are replaced.

1 Like

@wigglemuse Actually you are right, it’s sufficient. The platform can compare the checksum of the buyer submission to the seller submission.

What kind of leaderboard attacks are you specifically talking about?

The incentive would be to get a model to reach high on the leaderboard so it appears attractive to buyers and can be sold. The attack would be to create many otherwise useless models that overfit in different ways.

The “skin in the game” shifts from the creator to a buyer. The creator makes profits without taking risk. The buyer puts the skin in the game but stakes without important knowledge.

If buying/selling models became popular, then Numerai would be relying on evaluation skills of buyers, rather than the confidence of creators in their models.


Reasonable buyers would take into account owner’s stake when forming the buying decision. If the owner is well-staked then this would not be an issue. If owner of such models do not stake much, few would be willing to buy it given other options.

1 Like

Yes, I agree that would be a good indication.

1 Like

If CoE can’t fund you for any reason, you can ask for donations… Indicate an address and a progress bar and I’ll be happy to collaborate

1 Like

Possibly another way to make things a bit simpler is using something like Gumroad for individual listings. This way each individual could create what they want. Then all that would really need to happen is have some sort of webpage parser to get the pricing data from each persons gumroad account and display the data on a centralized website. This could also safeguard against fraud and security concerns around APIs etc because everything would be conducted on other peoples individual listings through gumroad and the transactions would be person to person. Could also save a lot of time in the development of a site from scratch. Sort of the inspiration behind this was a twitter post by Jordi with a link to his gumroad site where he just started listing his weekly prediction files on a tiered structure on a monthly or quarterly subscription based plan. See link https://jrdi.gumroad.com/. Im sort of thinking things might get so complicated with how each person wants to sell(scrips/raw predictions/tiered pricing based on stake size/lockup periods/etc) that this thing will never really kick off or take a very long time for development. Just a suggestion


I am sort of envisioning a central website that will have a cool home page. Tabs could include links to leader board, rocketchat, forums, numerai docs and a tab for account creation with submission instructions. The marketplace tab could be filled with everyone’s individual gumroad listings with pricing and subscription structures. We could also use the Numerapi to get other data from accounts/models to sort of “show off” like previous payouts, reputations/ranking, 3 mo returns on models being sold, current amount staked on models, performance charts and model stake value charts etc. Gumroad upon subscribing allows the owners to send posts to the individuals/newsletters with attachments such as scrips or CSV submission files.


@aventurine Great suggestions, I think Gumroad can be used for listings that do not require platform verification or automated submissions, or for people who prefer to transact in fiat. I will see if any organic integration would be possible.

@restrading So we discussed this a bit on the Office Hours live last week and seemed like we have consensus to get going on this! Is 69.420 NMR the amount that would be requested upfront? Is there a smaller chunk you’d like to begin with that has a smaller scope of work and lower cost? We talked about some kind of conversion into an hourly rate, but we also want a cap and a manageable short-term scope to get started. Ideas for what piece to do first?

Hi @hb_scout I am currently drafting the updated features list, work plan as well as funding mode, please allow me 1 or 2 more hours. Thanks.

Hi @hb_scout and everyone interested in this project. Below is the updated work plan and funding request. If you have any feature request, or want to see some features prioritized over others or anything else, feel free to comment below. Thanks all!

I have already done some prelimiary work on tech stack evaluation and decided to go with FastAPI-VueJS instead of Flask-React. I will be able to commit to this project full-time from Jul 15 for at least a month.

I will keep the table up-to-date as the project goes along. Recorded hours will always be less than actual time spent as I will exclude any unproductive time spent.

Module Feature/Task isOpen Est. Hrs. Recorded Hrs. Completion
Project Tech stack evaluation - 2 2 Y
Project scaffolding and local dev env setup - 2 2 Y
Local rapid frontend prototyping and designs - 12 15 Y
Backend-Seller Data schema, model, profile and authentication API - 3 2
Authentication with Metamask Integration 4 4 Y
Seller onboarding and model ownership verification API - 8
API tests - 2
Frontend-Seller User Middleware - 5 3.5
Authentication with Metamask Integration 11 11 Y
Seller profile UI - 2 2.5 Y
Seller listing / model verification UI - 5
Backend-Listing Data schema and model - 4
Listing details and Numerai integration - 4
API tests - 3
Frontend-Listing Product Middleware - 6
Listing Catalog UI - 4 1
Listing details UI - 8 6
Listing creation UI (Gumroad off platform) - 5
OffPlatform Gumroad integration - 8
Backend-Buyer Data schema, model, profile and authentication API - 2 1
API tests - 2
Frontend-Buyer Purchase Middleware - 4
Buyer profile UI - 1
Buyer purchases UI - 2
Payment-Crypto Support for NMR payment [Tentative]
Backend-Submission Seller submission API - 3
Submission file storage, access control and retention policy - 4
Submission file encryption and security [Open]
Email and other forms of notications to both parties [Open]
Live score monitoring and model consistency check - 4
CI/CD Docker Compose Setup - 2
Set up GitHub CI - 1 1 TBC
CD setup for GCP - 4
Cloud deployment of Minimum Viable Product - 2
Write development guide - 2
[Code handover, transfer to CoE GitHub]
[Start Accepting GitHub Contributions from This Point On]
Backend-OnPlatform Auction and order book API [Tentative] 8
Stripe Connect [Open]
Data schema and model [Tentative] 1
Transaction API [Tentative] 3
Numerai account and stake monitoring [Tentative] 2
API tests [Tentative] 5
Frontend-OnPlatform Auction UI [Tentative] 4
Purchase UI with dummy payment processing [Tentative]
Automated job queue for submissions and delivery [Tentative] 8
Payment-Crypto NMR collateral for buyers and sellers for stake-mode sales [Open]
Erasure support for buyer stake enforcement [Open]
Erasure support for seller model verification enforcement [Open]
Other Support sales for Signals data [Open]
Support sales for Signals files [Open]
Support sales for model files / scripts / notebooks [Open]
Reputation system [Open]
Subscription Sales [Open]
UI Enhancements (Search, filter, etc.) [Open]
Voting App with Account Requirements [Open]

Total Estimated Hours: 162
Recorded Hours Balance: 29 [For CoE: Recorded hours balance: 29 (1015 USD ~= 30 NMR at the time of writing) + 202.69 USD for domain name ~=6 NMR ==> 36 NMR in total]
Cumulative Recorded Hours: 51

Funding Rate: $35 / hr (approx. market rate for junior dev)

Funding Mode: No upfront payment needed, funding to be paid at the end of the project / weekly based on recorded hours in NMR using point-in-time NMR/USD conversion rate. For fairness the CoE can apply discretion at the end to make adjustments to the funding based on assessment of work quality.

Dev Cost: I will bear any infrastructure cost during dev and the initial 3 months of deployment. Going further I might fund the hosting through community donation, small platform fee (after crypto payment is implemented) or via separate proposals for hosting cost.


2021-07-15 — 2021-07-17:
Sneak Peek:

  • Completed framework setup and base e-commerce frontend boilerplating.
  • Updated workplans, prioritized off-platform (Gumroad) integration and CI/CD, which would allow earlier release of code, more collaborations and earlier deployment of minimum viable product.
  • Revised some estimated hours based on recent work. Changes/additions to the workplans are in bold.

2021-07-18 — 2021-07-25:
Sneak Peek:

  • User will be able to login through Metamask (and potentially other providers) by signing a backend generated nonce, paving the way for crypto payment functionality
  • Completed generic user module (both frontend and backend), product listing UI is in-progress
  • Updated workplans and time estimates. Changes are in bold. Deprioritized buyer specific stuff as they are not yet necessary for 3rd party platform listings.
  • Next week to focus on seller onboarding and product related UI, middleware and backend.

An important question which seems to be missing in the marketplace development is Terms of use for products. Is there any plans about it? May be we can create this document together and place it under open-source license somewhere. Otherwise, every seller have to spend some time and money on developing of this document, because it is illegal to use someone elses Terms of use with no permission. I have some draft developed by me which I can share, but I’m not a lawyer and have no experience in that kind of things. May be it would be better if CoE will sponsor professional lawyer for developing of that doc.


Right now I’m focusing on building something that works first. The early product will host 3rd party listings such as Gumroad. We can figure out the legal terms along the way as the product matures. I’m open to suggestions.

1 Like

I’m late to the party but wanted to chime it. I love this idea. I may have missed it, but I didn’t see anything that addressed long-term support and maintenance. Would it make sense once the product is funded and built out to shift financial upkeep off of the COE and push it towards a self-sustaining usage fee? If I were selling models on the marketplace, I’d have zero reservation about losing a few percentage points of the purchase price, back to the marketplace. Put differently, I have no issue with @restrading making money off of this. A lot of money if it works well.

I’d also like to see the purchase price tied back to the seller’s skin in the game on a model via some multiplier cap: i.e. you can’t sell a model for 100 NMR if you only had 1 NMR staked on it for 20 weeks. If you had 25 NMR staked on it for six months and want to sell it for 100, maybe… If that makes sense. I think SITG is still important here.

Great idea, looking forward to seeing it mature.


Thanks for the support. Yes as said above I can think of 3 ways going forward, with the small platform fee being the most self-sustaining method. Such fees can be used to cover hosting costs as well as to fund future devs so as to reduce financial burdens on the CoE.

I like the idea of stake-tied-pricing limits, and will explore how it would make sense. :slight_smile:

1 Like

First 26 NMR funding tranche sent from CoE multisig


Thanks @jrai and CoE for the fast turnaround and support, NMRs received! I will make the next update in a week.


This thread was archived, for new updates please see: [Updates] NumerBay - The Community Marketplace