Catch market-moving news before it hits the crowd.

Second-Order Chain

When a big story hits one company, the company economically attached to it often has not reacted yet. This board reads the platform's 38,000-edge relationship graph — typed, directional, evidence-backed supplier / customer / partner / competitor links extracted from news — and finds those un-reacted neighbours. The hard part is not finding connections; it is the three ways this strategy loses, each of which is a filter here rather than a judgement call: 1. THE NEIGHBOUR HAS ALREADY MOVED. This is a PRICE question, not a news question — sympathy selling moves a stock without generating a single article, so quiet coverage proves nothing. Each candidate's return over the window is compared with the headline's, and anything that has already collected a meaningful share of the move (in either direction) is dropped. 2. THE SIGN IS BACKWARDS. Edge direction in the graph is unambiguous — `from -[supplier]-> to` means from SUPPLIES to — so the rule is arithmetic: everyone in a value chain moves WITH the headline, a competitor moves AGAINST it. 3. THE EDGE IS TOO SMALL TO MATTER. A real relationship touching 2% of revenue is a fact, not a trade. Edges are gated on strength and on how many independent articles asserted them. Pipeline: 1. Headline names: enough coverage in the last few days to be a real story, and either a strongly one-directional sentiment reading or a large price move. Both are reported, because they disagree more often than you would expect — one recent story ran at +0.31 average sentiment across 75 articles while the stock fell 15%, and a sentiment-only screen would have pointed the wrong way. 2. Neighbours: every graph edge above a strength and mention floor, deduplicated to ONE edge per pair (the graph stores both directions and sometimes several types for the same pair). 3. Tradeable: US-listed and liquid enough to enter and exit; the raw graph contains foreign listings that no order could fill. 4. Sign, then the not-yet-moved test. 5. Contradictions dropped: a name that two live stories push in OPPOSITE directions is removed rather than resolved, because it has no second-order edge left. Several stories pushing the same way are kept as corroboration. Each row carries the headline and its move, the neighbour and its move, the share of the move already captured, the edge type / strength / evidence count, and the resulting side.