Same strangers, same stillness, same hours — opposite outcome.
Conditions are published figures or derived from them. Output is measured from a standard run — 1,200 people, 4 hours, fixed seed, identical for both rooms. A highlighted cell is only marked where “more” is actually better; density and volume are conditions, not scores.
Strip both situations to their physical description and they are the same: strangers, held close, not moving, for a long period, sober. A crush-loaded carriage is even denser — 4.0 people per square metre against a queue's 1.5 — and it is quiet enough that speech carries 2.4 metres.
The carriage produces 74 lasting ties and a largest connected group of 24. The queue produces 633 and 634. Roughly nine times the ties and twenty-six times the network, from the sparser room.
The difference is what the two crowds are doing. A queue is a set of people waiting together for the same thing; a carriage is a set of people whose journeys happen to overlap for a while. In the model that is the attention parameter — 0.85 against 0.10 — and it is the one number in this whole system with no citation behind it. It is an assertion. This pair is the case that makes the assertion hard to avoid.
