What Investors Want in a Data Provider’s Backtest
The bar for a successful package is having the buyer feel “this is worth a serious look,” not “this is proven alpha.”
Read MorevBase Blog
Recent Posts
How Good Are LLMs at Sentiment Scoring?
We find that open-weight models, especially in an ensemble, approach frontier models’ performance. As with labeling, sentiment scoring is now accurate and cheap.
Local LLMs can now handle ticker mapping for alternative data
A benchmark of open-weight LLMs reveals how capable they've become at ticker mapping — and what that opens up for alternative data pipelines.
How good are LLMs at tickerization?
A reproducible benchmark comparing flagship LLMs and light LLMs in extracting asset tickers from unstructured text. Comparing performance of leading LLMs, including GPT, Claude, Gemini and Grok.
If data is oil, AI is the shale revolution
LLMs cut the cost of turning unstructured data into investor-ready signals by 5-100x. Data that was previously stranded may now be a valuable asset.
Your LLM’s alpha might be mere memorization
Commercial LLMs can generate impressive backtest alpha — until the test window moves past their training cutoff. What looks like edge may just be memorization.
Why alternative data trials fail
Most alt data evaluations stall on packaging, not signal quality. Here's why, and what providers can do about it.