Introducing the Shiny Numerati Dashboard

[ia_ai]

Hey fam, as promised, I am writing this post about Shiny Numerati. It is a community dashboard for the main Numerai tournament. You can use it to track your model performance and payouts.

The main objectives of this forum post:

1. Onboarding new users with a simple quick-start guide
2. Collecting feedback from the community (bugs and feature requests)
3. Providing a roadmap (and polls to upvote feature requests from time to time)
4. Providing updates on new features

OK, let’s get started!

1. Links to App:

• App with HF nav bar: Shiny Numerati - a Hugging Face Space by jofaichow
• App without HF nav bar: https://jofaichow-shiny-numerati.hf.space/
• Source code (HF): jofaichow/shiny-numerati at main
• Source code (GitHub): GitHub - woobe/shiny-numerati

2. Quick-start Guide

Step 1: “Start Here” → Use the picker input widget to select your model(s) → Step 2: Download data from Numerai

Step 3: “Performance Summary” → Step 4: Adjust the Era/Round Filter → Step 5: Generate

Explore Other Performance Summary Tabs

KPI Analysis (CORRv2 vs. TC)

Payout Summary (Overview)

Payout Summary Chart (Individual Models)

Download Raw Data

Materials from CoE Sponsored Events

3. Roadmap

I will continue to update this section after collecting some user feedback.

Some feature requests that I remember from various discussions on Discord:

• Around the world with Numeratis Survey results
• Flexible model name inputs (e.g. text, account level, json)
• More KPI charts (with options to pick different KPIs for x and y-axis)
• Shiny Numerati for Signals
• Artificial stakes for payout simulation
• Numerai’s benchmark models (a button to download and show data from benchmark models)

(Polls will come soon after I have collected more feedback)

4. Feedback

Please comment below and let me know what you want

[ia_ai]

2023-11-28 - Version 0.2.5

Added a new Score Multipliers: 0.5 x CORRv2 + 2.0 x MMCv2

[ia_ai]

2023-11-29 - Version 0.2.6

Added Survey Results to Community Events tab

[animalfarm]

Hi, your dashboards are very useful, how often you check for new models, I uploaded some last week, but no first results yet.

[ia_ai]

Hey @animalfarm thanks for your kind words. I usually update my dashboard when new scores become available. I am relying on the get_leaderboard API function to get the list of all models which have at least one resolved round. If your models are “too new” (i.e. no resolved round yet), you will have to wait until they have their first resolved round.

Having said that, there is a quick hack if you don’t want to wait. You can fork my app on HF and then edit this line. Replace choices = ls_username, with choices = c("model_name_1", "model_name_2"). Then you can run the app with specific model names.

[animalfarm]

@ia_ai I really appreciate your answer and will definitely check it out.