{"method":{"id":"quant-candidates-llm-ranked-v2","name":"Quant candidate screen, model-ranked","summary":"A fixed, rules-based pre-screen builds a short candidate list each session from two documented setups — post-earnings-announcement drift (a positive EPS surprise reported in the last three sessions) and intermediate momentum (top fifth of the universe by 13-week return, trading within 10% of its 52-week high). Names picked in the previous five sessions are excluded. A large language model with live web search then checks each candidate for news the data does not yet reflect and ranks five, stating which setup each pick expresses. Two independent model passes are merged, with names both passes agreed on ranked first.","is_quantitative_model":false,"universe":"S&P 500 constituents (current published list), filtered to a rules-based candidate set of at most 25 names per session","cadence":"Once per weekday, after the US close","horizon":"Five trading sessions. Picks are framed for that window and the primary performance figure is the 5-session excess return over SPY.","determinism":"The candidate set is deterministic and stored for every session. The ranking within it is not: the model may order the same candidates differently on a rerun. Every run is logged with its candidate set, model passes and cited sources.","validation":["Every ticker must be in that session's stored candidate set or the run is rejected; the model cannot add names.","Every pick must name the setup it expresses and carry at least one cited source URL returned by the search tool.","Exactly five distinct tickers, each with a stated key risk and a confidence level; a name only one of the two model passes chose cannot be marked high confidence."]},"disclaimer":"This is a ranked screen output, not financial advice and not a recommendation to buy, hold or sell any security. It is generated by a language model and may be wrong, out of date, or incomplete. No assessment has been made of any recipient’s objectives, financial situation or needs. Do your own research and consider consulting a licensed financial adviser.","live_method_key":"v2","methods":{"v1":{"id":"llm-research-screen-v1","name":"LLM research screen","live":false},"v2":{"id":"quant-candidates-llm-ranked-v2","name":"Quant candidate screen, model-ranked","live":true}}}