Tiny changes do not always stay tiny. This lab lets you start two or more forecasts almost exactly together, then watch whether feedback keeps them close or pulls them apart.
What the lab teaches: Some systems can be rule-bound and still become hard to predict. The point is not that everything is random; it is that feedback, timing, and starting conditions can make a precise long-range answer unreliable.
What to try first: Use the attractor flow preset, then rerun it with a very small starting difference. Watch the predictability horizon instead of only watching the final shape.
Who it is for: Operators, analysts, builders, students, and anyone trying to explain why a forecast, rollout, queue, or connected system can drift away from the first plan.
Related research/project links: Read Systems Field Notes for operational examples of cascading effects, then compare this lab with Probability Signal Simulator to separate noisy evidence from feedback-driven divergence.
What this tests
The chart shows near-identical paths moving through the same kind of system. If the paths separate, the lesson is simple: knowing the rule does not always mean you can trust one exact long-range forecast.
Simulation modes
- Attractor flow: the lines stay inside one familiar shape, but the exact turn becomes hard to call.
- Feedback map: each step feeds into the next, so a tiny mismatch can compound.
- Local-rule cascade: one small local change can spread across a connected pattern.
- Gravity slingshot: a near miss can change timing and direction.
- Basin boundary: two close starts can land in different final outcomes.
How to read it
Start with a preset, press Run model, and watch the bright lead point move across the chart. The predictability horizon is the first step where the paths are no longer close enough to treat as one answer.
That horizon is a threshold crossing, not a property of the system. It is the first step where the spread between traces passes a tolerance set per simulation mode, and the tolerances differ because the modes measure spread in different units — a gap between values for the feedback map, a distance on the plane for the attractor and the slingshot, a fraction of disagreeing cells for the cascade. Where the horizon lands is partly a statement about how much divergence you are still willing to call one answer.
Why it matters
Markets, queues, networks, launches, and operations can follow real rules and still surprise you. Small measurement errors, hidden differences, or repeated feedback can grow until a forecast that looked precise becomes the wrong story.
Deeper note
This is a teaching model, not a production forecasting engine. The deeper idea is sensitivity: in some systems, the next step depends so strongly on the current state that a microscopic difference can be amplified over repeated steps.
It is also seeded rather than live: the same settings and the same seed reproduce a run exactly. That is what makes the dials mean anything. If the noise were redrawn on every press, a change after moving one slider could always be the new draw rather than the slider, and nothing could be attributed to anything.
Related
Use Probability Signal Simulator for probability updating, or Systems Field Notes for operational examples where small changes can cascade.