Open-source LLM and dataset for sports forecasting (Pro Golf)
Get picks like this daily. Today's launch numbers, the top launches, AI and tech news, and a daily opportunity pick, straight to your inbox.
Want to reach these readers? Sponsor an issue.
Hey HN, I fine-tuned a small open-source model on golf forecasting and it beats GPT-5 at predicting golf outcomes. The same approach can be used to build a specialized model in any domain, you just need to update a few search queries.We fine-tuned gpt-oss-120b with LoRA on 3,178 golf forecasting questions, using GRPO with Brier score as the reward.Our model outperformed GPT-5 on Brier Skill (17% vs 12.8%) and ECE (6% vs 10.6%) on 855 held-out questions.How to try it: the model and dataset are open-source, with code, on Hugging Face.How to build your own specialized model: Update the search queries and instructions in the Lightning Rod SDK to generate a new forecasting dataset, then run the same GRPO + LoRA recipe.SDK link: https://github.com/lightning-rod-labs/lightningrod-python-sd... Dataset: https://huggingface.co/datasets/LightningRodLabs/GolfForecas... Model: https://huggingface.co/LightningRodLabs/Golf-ForecasterQuestions, feedback on the SDK, suggestions for new domains to try this on - all are welcome.
HN · launched Feb 24, 2026 · 7 upvotes (top 62%) · 0 views
Niche: large language models · 191 launches, 4 in the last 7 days ▼ -33% · Explore this niche →
Category and tags
- Media & entertainment
- Model & infra
- Developer
- AI-native
- Hobby / open-source project
Visit website ↗ Source listing ↗
More like this: all 191 launches in large language models →
Could you build this?
No Fine-tuning 120B parameter models using LoRA/GRPO reinforcement learning to beat state-of-the-art generalist models requires large-scale multi-GPU compute clusters, domain-specific forecasting benchmarks, and deep RL/post-training expertise.
What it would actually take: Building this requires a distributed GPU cluster (e.g., 8x or 16x H100s) running DeepSpeed, Megatron-LM, or vLLM with Ray. The critical bottlenecks are designing the reward model and RL loop (GRPO / PPO), scraping historical golf tournament statistics and weather data to construct reliable ground-truth forecasting benchmarks, and managing memory/sharding for 120B parameter models. This requires senior ML research engineers and thousands of dollars in dedicated GPU compute.
Discussion
No comments on this launch.
Competitors
Other products that read as similar to this one : 57 launches clear the similarity bar, closest 8 shown.
Attention rank: #32 of 58 (itself plus its competitors, highest first, normalized so YC and Product Hunt are compared fairly).
Launched 110 days after the earliest competitor.
- Trained an LLM to predict "What will Trump do?" · hn · 2026-02-20 · 10 upvotes · similarity 0.53
- GPT-5.5 by OpenAI · ph · 2026-04-26 · 476 upvotes · similarity 0.42
- T0: We trained a 256M SOTA open-weights time-series forecasting model · hn · 2026-10-14 · 14 upvotes · similarity 0.39
- I mapped every US golf course · hn · 2026-07-26 · 223 upvotes · similarity 0.39
- Hindsight Models and Inference API · ph · 2026-11-04 · 73 upvotes · similarity 0.38
- WebGCM · hn · 2026-09-21 · 6 upvotes · similarity 0.38
- TabPFN-2.5 · hn · 2025-11-06 · 73 upvotes · similarity 0.37
- lepoch · ph · 2026-09-16 · 1 upvotes · similarity 0.37
The niche, labels, vibe code rating and competitor comparison on this page are analysis by Nicheloom. If you use them, please credit nicheloom.com. How you may use our data.