Play this window back the way it happened. The clock walks from the open to the bell, and everything scoped to the window is redrawn from the rows recorded at or before it — the tiles, the charts, the books, the gate that was failing, the fills as they land, and the account as it stood that evening. Nothing stamped later than the clock is on screen, the settlement included: the outcome is withheld until the replay reaches it. The tables that pick a window keep showing the whole store, because that is what they are for. Space plays and pauses, ←/→ seek five seconds, with shift thirty.
One candle per market window, folded from the same Chainlink ticks the window chart above draws — so a candle is a window, and its direction is the one the window resolves on: blue closed up, orange closed down. Select a candle to load that window onto the rest of the page. The axis is wall-clock time, so a stretch nothing was recorded for shows as a gap rather than being closed up.
Chainlink Data Streams benchmark price. The gray rule is the window's TWAP open — the level this market settles against.
Mid price of each outcome, in probability terms (0–1).
Live price minus the window open, in dollars, one point per Chainlink report (about one a second). The zero baseline is the open — which side of it the fill sits on is the whole question the window settles. The dashed line is the settling average against its own open, which is the level this market actually resolves on: it lags the fill by design, so the gap between them is the move spot has already made and the average has yet to pay out. The dotted line is the local reconstruction of it.
(bid quantity − ask quantity) ÷ total, in [−1, +1]. Buying Down at p is selling Up at 1−p, so the consolidated book merges both legs into one synthetic Up book. This is the whole signal: the bot only sweeps a side the consolidated book already leans toward, and only in the last 30 seconds, where that lean is a forecast rather than a position.
Every simulated fill, with the state that justified it. Fills inside the window on the price charts are marked there too. Select a row to load its window.
No fills yet.
Realised P&L, cumulative from zero, running through the settled windows in the order they traded. The right-hand end is the all-time figure in the tile above.
The same figure the meter on the TWAP hero fills, replayed across the window — one simulation every 5 seconds off the tick series, on the axis it is read against everywhere else. The market resolves on twap close above twap open, so a side is only ever lost by the settling average coming back through that one level: the solid line is the share of simulated paths that touch it before the bell, the dashed one the share that finish on the wrong side of it. Settlement is always the lower of the two, and the gap between them is the danger that is real but temporary — paths that cross the open and cross back in time. The rule is the bar the bot is judging this instant against; a fill under it is a position it thought safe.
Top levels; bar length is size relative to the largest shown.
Waiting for book…
Top levels; bar length is size relative to the largest shown.
Waiting for book…
Settled positions, newest first. Select a row to load it.
No windows traded yet.
Every quote the bot placed, moved or pulled, and why.
Nothing quoted yet.
The gate that stopped an entry, counted per second inside the entry band. Standing aside is the normal state; the rows marked signal was clear are not — those are trades the strategy wanted and something else refused.
Nothing recorded yet.
Every window the bot watched, traded or not. Select a row to load it.
No windows recorded yet.