SShortSingh.
Back to feed

Developer grades own ML options forecasts after June expiry, scores worse than coin flip

0
·2 views

A developer publicly committed in May to grading Helium's machine-learning probability forecasts for two AAPL options contracts expiring on June 26, 2026. When the contracts expired, AAPL closed at $283.78 on Nasdaq, leaving the $310 call out of the money and the $295 put in the money. The mean Brier loss across both contracts came out to 0.3846, worse than the 0.25 a naive 50-50 coin-flip model would have scored. The call forecast was closer to the actual outcome than the market-implied probability, but the put forecast suffered a high loss of 0.5929 because the model had significantly understated the chance of the stock falling that far. The developer acknowledged the sample size of just two contracts is too small for calibration conclusions, framing the exercise as a discipline of publicly logging and then honestly grading predictions.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingHacker News ·

Developer Releases Tool to Prevent Swastika Patterns in QR Codes

A developer has published an open-source Rust crate called 'qr-swastika-avoider' on crates.io. The tool is designed to detect and avoid the accidental generation of swastika-like visual patterns that can appear in QR codes due to their black-and-white pixel arrangements. Such unintended patterns can cause reputational or social harm when QR codes are used in public-facing contexts. The library gives developers a way to regenerate or adjust QR codes that would otherwise render offensive shapes. The project was shared on Hacker News, where it received modest early attention.

0
ProgrammingHacker News ·

Researcher Trains JEPA World Model to Learn Super Mario Bros Environment

A developer named Benjamin Bai has built a project called LeMario, which applies a Joint Embedding Predictive Architecture (JEPA) world model to the classic video game Super Mario Bros. JEPA is a self-supervised learning framework originally proposed by AI researcher Yann LeCun, designed to learn abstract representations of how environments work. The project explores whether such a world model can effectively learn the dynamics and structure of a game environment without explicit labels. LeMario was shared on Hacker News, where the original article and technical details are hosted on Bai's personal project page.

0
ProgrammingDEV Community ·

Developer Uses Spec-Driven Development to Build a Rogue-Like Game for Kiro's First Birthday

A developer used spec-driven development (SDD) to build a rogue-like game celebrating the first anniversary of Kiro, an AI-powered IDE. SDD is a structured approach that breaks work into three phases — requirements, design, and task generation — all completed before any code is written, with a human reviewing each step. The game, built using Phaser and TanStack Start, features a 2D Final Fantasy-style world where players explore buildings representing key milestones in Kiro's history. The developer used Kiro's built-in spec mode to generate the planning documents from a single descriptive prompt. While SDD requires more upfront planning time, the developer argues it leads to higher first-pass success rates and more stable outcomes for complex features.

0
ProgrammingDEV Community ·

Why Supabase RLS Errors Are a Feature, Not a Bug — And How to Fix Them

Developers using Supabase often encounter the error 'new row violates row-level security policy,' which signals that Row-Level Security (RLS) is active and correctly blocking unauthorized writes by default. The error typically stems from one of three causes: a missing INSERT policy, a mismatched or null user_id field, or an unauthenticated request where auth.uid() returns null. The recommended fix involves creating an explicit INSERT policy that allows signed-in users to write only their own rows, combined with a server-side column default to automatically assign the correct user ID. Developers are also cautioned that the Supabase SQL editor runs with elevated privileges that bypass RLS, meaning successful writes there do not guarantee the same behavior from a client app. Disabling RLS to silence the error is discouraged, as an unprotected table is the more serious security risk.