Market Signals
Small models label stock headlines positive, negative or neutral. Then we check what the stock did.
Does the label line up with the move?
Average return vs Nifty over the 3 trading days after the headline, grouped by the network's label. A bar to the right means the stocks beat the index.
Judge
The judge only steps in when the 3 voters disagree. It gives one hint, never the answer; 3 fresh votes must then all agree.
What the hint achieved
Headlines that went to a hint round and finished it. Without the hint, all of them would have been thrown away.
By kind of headline
Which kinds of news moved the stock? Average 3-day return vs Nifty.
How to read this
This is a test of label quality, not a trading tip.
Nodes see only the headline and the company name, never the price. Returns are filled in from real market data after a label is settled.
If the consensus labels are meaningful, positive headlines should beat the index more often than negative ones. A gap near zero means the labels carry no price signal.
Up / down = beat or lagged Nifty by more than 1%. Past headlines only; earlier trading tests on this data never beat random picks.
Latest labelled headlines
Newest settled labels first, with what the stock did next.
| Date | Headline | Signal | Kind | Decided by | 3-day vs Nifty | Match |
|---|