We show the boundary between evidence and a decision.
OrbSentry cannot see the current in-game order ladder. It organizes what the connected sources really published, reconstructs whole-item outcomes, and gives you a private browser worksheet for the missing live inputs. It never guarantees a price, quantity, fill, Gold cost, or profit.
Observed
- Exact source response bytes, retrieval time, source time, and content hash.
- GGG completed-hour volume traded plus low/high stock and ratio aggregates.
- poe2scout representative pair snapshots and their published summary fields.
- Private in-game quote amounts entered by the signed-in player.
Calculated
- Cross-rates, same-hour GGG envelopes, and source disagreement.
- Whole-item flooring after every leg—never fractional inventory.
- Direct changes over named windows, only after observations span at least 98.0% of that window.
- Changed-state recurrence, five-minute persistence checks, later-state survival, and historical route drift.
Interpretation
- A pattern may be worth checking in-game.
- A missing pairing may be worth researching as a market-making opening.
- An unusual or fractional rate may carry execution risk.
- Repeated historical evidence can raise confidence in the pattern, not in a current fill.
Unknown now
- Current quantity at each price level, order count, queue position, and cancellations.
- How long each leg may take and whether every leg can fill before the market changes.
- The Gold cost for your exact character, amounts, and directed pairs.
- Player intent, future price movement, latency, and handling time.
Aggregate activity is not live order-book depth.
GGG summarizes trades from completed hours. poe2scout publishes volume, stock, and representative-price fields, but does not document the exact interval or transformation behind every summary. Neither source publishes the current price-level ladder, quantity waiting at each ratio, queue position, cancellations, or expected fill time.
OrbSentry may use aggregate activity to describe historical participation. It must never present that value as current executable capacity or as proof that a route will fill quickly.
| Source | What it contributes | What it cannot establish |
|---|---|---|
| GGG | Canonical completed-hour volume traded plus low/high stock and ratio aggregates. OrbSentry stores exact responses and never rewrites those source facts. | The current incomplete hour, current price-level depth, queue position, expected fill time, or simultaneous route fillability. |
| poe2scout | Frequently checked representative pair rates, summary volume and stock fields, and market catalog coverage. | The exact interval and transformation behind every summary, or an independently verified live order book. Fields that match the same GGG hour are treated as shared upstream data, not independent corroboration. |
| poe.ninja | Normalized price history and a fallback comparison model where direct pairs are missing. | A direct offer or executable pair. Modelled edges are labelled as inferred. |
| You | Current whole-number quote ratios, visible quantity, intended bankroll, and Gold shown in-game. | A future fill. Your worksheet proves its arithmetic only for the values entered. |
Inspect the source contracts directly: GGG developer reference · poe2scout API schema.
- Use Opportunities, History, Risk flags, and Market gaps to find a pattern worth checking.
- Open every directed pair in-game and copy the current whole-number give/receive amounts.
- Enter visible input quantity and the Gold for the exact planned step when known.
- Let the worksheet floor every conversion, then reject any route with a missing quote, insufficient quantity, impractical handling, or unacceptable Gold.
- Treat the result as a decision aid for that observation—not a promise that the market will wait.
Automated research currently studies up to 7 legs and 7-day mature trends from 169 hours of retained analysis history. On a dense market the automated route search is broad but not exhaustive over every permutation. The public scratchpad supports up to 12 hand-built legs so a player can test a longer route the automated search did not surface.
Open the live worksheet →