This panel is a live read of three independent daemons. Nothing here is calculated in your
browser except WINS; every other value was measured on a server, written to config, and is being read back.
The panel does not trade — it shows what the engine has already decided and why.
WHERE THE DATA COMES FROM
MACRO ANALYZERMacro VPS, every 15 min. Fetches 200 bars of 30-minute candles per pair, finds swing highs and lows, clusters them into support and resistance zones, fits a trend line, and projects one bar ahead. Feeds MACRO, AR, STRUCTURE and S/R.
PERF ANALYZERMacro VPS, every 20 min. Reconstructs completed trades from raw fills and decides each pair's tier, weight and restrictions. Feeds EXP, TIER, STATUS, STRK and the LAST RUN ACTIONS below.
SENTIMENT HUBMacro VPS, every 5 min. Polls Fear & Greed, funding rates and long/short ratios. Feeds SENTIMENT COMPOSITE.
The age badgeThe number beside each daemon is how long since it last published — not since the page refreshed. A number climbing past its cadence means that daemon has stopped.
THE VERDICT — two horizons, deliberately separate
SHORT TERMWhat the 5-minute engines say right now: the direction and how many of the enabled engines agree. The denominator moves — engines can be switched off on the Strategy Tune panel — so it is always written as n of m.
not scoredNot scored is not "no signal". The open monitor stops a pair at the first gate that applies — a full book, a timer, a ban, price sitting on a level — and a pair stopped that way never reaches the engines at all. Measured across three hours on 2026-09-22: of 271 pair decisions only 74 reached the engines, and 95 stopped because the book was at its ceiling. This panel used to print "engines below threshold" for every one of them. It now names the gate instead, and says no signal only when the engines were genuinely asked.
LONG TERMThree reads vote: the regression bias through the swing points, the swing structure itself, and the AR forecast. Where they split, the line says so rather than picking a winner, and a forecast on a losing streak is flagged. Its horizon is 200 bars of 30-minute price — about four days. For 4H, daily and weekly, open the row and read timeframes.
ENVIRONMENT — what stands between the fleet and a trade
n/m scoredHow many of the watched pairs actually reached the engines on the monitor's last pass, and how many of those carry a live signal. A low number here is not a quiet market; it means something stopped the scan earlier.
BOOKOpen positions against the target, with the hard ceiling beside it. At the ceiling no new pair is scored at all — which is the single most common reason this panel looks asleep.
gate chipsEvery gate that stopped a pair this pass, counted. Red is the book, yellow is where price is sitting, orange is a feed problem, indigo is a timer or a rule. Hover a chip for the raw reason name as the monitor records it.
standing rulesTrue whatever the last pass did: pairs with a direction banned by the analyzer, benched pairs, pairs on probation, pairs barred from averaging down, and pairs the cost gate is blocking. Hover for the list. These persist until the analyzer lifts them, so they explain a quiet pair better than any single cycle can.
MARKET REGIME — the fleet in one line
BEAR / BULLA vote, not a forecast: every pair's macro bias counted up. ▲5 ▼10 ●2 reads as five bullish, ten bearish, two ranging.
The tag rowHow many pairs are currently in each structural state — downtrend, uptrend, ranging, expanding, transition, churn — plus how many sit at-resistance or at-support, and how many show equal-highs or equal-lows (repeated rejections at one level, often a stop cluster).
SENTIMENT COMPOSITE
F&G INDEXCrypto Fear & Greed, 0–100. Counts only at the extremes (below 25 or above 75).
ORDER FLOWShare of recent taker volume hitting the ask versus the bid. Above 50% is buy-side pressure.
BOOK DEPTHResting bid size divided by resting ask size. Above 1.0 means more support below than supply above.
L/S RATIOLong accounts per short account on the venue. Read as a contrarian signal: a crowded side is the side that gets liquidated.
FUNDINGPerpetual funding rate. Shown for context and deliberately not counted in the /4 verdict — it moves too slowly to be a fifth vote.
−1/4The net score over the four counted inputs. Small numbers are normal; the thresholds are extreme by design, so this reads neutral most of the time.
PERFORMANCE STRIP
WR · LAST 10
WR · LAST 20Win rate over the most recent 10 and 20 decisive trades fleet-wide. Decisive means the trade made or lost money — exact break-evens are excluded rather than counted as losses. The small n is how many of that window were decisive.
TRADES 24HCompleted round trips reconstructed from fills in the last 24 hours.
FILLS 24HRaw executions. Higher than TRADES because one trade can exit in several partial fills.
MACRO ERRTimes three or more pairs lost inside the same five-minute window — the signature of a macro move going against the whole book rather than a bad pair. Triggers a fleet-wide entry cooldown.
QUIET RUNSConsecutive analyzer runs that found nothing to act on. A high count trips a safety valve that restores the previous thresholds, on the theory that the last change is what silenced it.
WR here and WINS in the matrix answer different questions. This is the fleet's most recent 10 and 20 trades. WINS is per pair, over everything loaded. They can disagree sharply and both be right.
SELF-TUNED LINE
T1 / T2The take-profit ladder, as a percentage move. T1 is the first scale-out and trims about a third; T2 takes half of what is left. T1 < T2 always. There used to be a third rung below T1 called T0, a 25% micro-take. It was retired on 2026-09-16 — not because it lost money on its own, but because firing it was the only thing that armed the stagnation close, and that pairing shut winners at a worse price than the micro-take had already got. The trail took over the space it left.
TRAILShown as arm→−drop. The position must first reach a peak of arm before the trail switches on; after that, a fall of drop from the high-water mark closes everything remaining. This is the tier that replaced T0, and it is a better instrument for the job: T0 cut at a fixed small number whatever the trade was doing, while the trail lets the move happen first and then keeps most of it. The drop is clamped so it can never fire below the realisable floor — the trailing close cannot bank a loss.
TIME-LIMITHow long a profitable position may sit before it is closed on the clock rather than on a target.
MIN ENGINESHow many of the signal engines must agree before an entry is allowed. Raised automatically after a macro error or a weak stretch.
COOLDOWNCLEAR, or a countdown during which no pair may open.
STRICT GATEWhen ON, entries additionally require sentiment to agree with the signal.
PAIR MATRIX — column by column
PAIRThe perpetual contract. Click any row to expand its structure detail.
WINSRealised win rate for this pair, counted in your browser from settled exits. Break-evens excluded; the small number is the decisive sample. This is the one figure on the panel computed client-side.
EXPExpectancy — average profit or loss per trade, as a percentage, over the analyzer's rolling window of the last 30 trades. This is the number TIER is derived from. Green above zero, red below. Dimmed when the window is too short to have earned the tier. The small number is the sample size.
TIERThe pair's standing, banded on EXP around zero: ELITE ≥ +0.15%/trade, NORMAL ≥ +0.05%, WEAK ≥ −0.05%, POOR below. Zero is the boundary because zero is where a pair stops making money. Moves at most one step per run in either direction.
STRKCurrent consecutive-loss streak, e.g. 13L. Resets on any winning trade.
STATUSCLEAR — unrestricted. RESTRICTED with a countdown — entries reduced or paused until it expires. NO-DCA — may open, may not average down. NO-LONG / NO-SHORT — that direction is banned after repeated losses on it.
MACROTrend bias from the 30-minute regression: bull, bear or ranging. An entry against both MACRO and AR is blocked.
ARAuto-regressive forecast for the next bar, with its confidence. The arrow is direction; the percentage is how often this pair's forecast has been right lately.
STRUCTUREThe detected pattern — downtrend, uptrend, ranging, expanding, transition. A yellow ◉ marks a churn zone, where the stop-loss is deliberately widened because price is chopping.
S/RDistance to the nearest opposing level. RES 0.14% means resistance is 0.14% above — longs are blocked that close. SUP is the mirror for shorts. A plain R1.9% is the distance with no block in force.
WTThe size multiplier the analyzer assigns (POOR 0.4 → ELITE 1.3). Currently has no effect: position sizing rounds every order up to one contract at the present account size, so a 0.4 pair and a 1.3 pair send the same order. Shown so the analyzer's intent is visible.
COSTThis pair's measured round-trip cost in percent — spread plus both fees — published by the cost-gate daemon. Orange when the live T1 does not clear it, because a take-profit under a pair's own cost banks a loss by construction at any win rate, and ⛔ when the cost gate is actively blocking the pair. A dash means the daemon has not priced it.
SIGThe engines' present-tense reading: ▲4/5 is four of five enabled engines long right now. When there is no arrow the cell says why, not nothing — ceil the book is at its ceiling, hold a timer is running, bench the analyzer or you switched the pair off, wall price is sitting on a level, none the engines were asked and did not agree, feed the price data failed. A dash now means only that the monitor did not report this pair at all.
ROW DETAIL — click any pair
resistance
supportThe actual price levels, each with a touch count. More touches is a level the market has respected more often — it also raises entry size when price is reacting there.
price → ARCurrent price and the forecast for the next bar.
slopeGradient of the trend lines through the swing highs and the swing lows. Diverging slopes mean an expanding range.
timeframesThe same structural read at 30m · 4H · 1D · 1W, each with its bias, its structure, and either the level price is sitting on or the distance to the next one. The macro daemon has always computed these; until CL-385 the panel showed only the 30-minute block. The line underneath says whether they agree — a split is not a failure, it is the reason a 30-minute signal can be right and still run into a daily wall.
exhaustionHow stretched the move is, scored per side, with the reasons named — trend age in 4H bars, a forecast that has turned against the trend, a higher timeframe pointing the other way. trend 72h old is the kind of fact that decides whether to take a continuation entry. Orange when the pair is flagged exhausted.
AR recordCorrect and wrong streaks for this pair's forecast, with its confidence.
LAST RUN ACTIONS — what the analyzer just did
tier_X→YThe pair changed tier. Green labels are promotions, red are demotions.
probation_on
probation_releasedA losing pair keeps trading at a reduced cadence instead of being benched, so it can still produce the evidence needed to rate it back up — and released once it does. A benched pair generates no data and can never earn its way out.
bench_lifted_on_recovery
legacy_bench_clearedAn older-style ban removed, either because the pair recovered or because probation replaced it.
escalation_resetThe consecutive-POOR counter zeroed after the pair came back.
dca_ban_liftedThe pair may average into positions again.
min_engines→NA fleet-wide change to how many engines must agree before any entry.
Reading order that works: MARKET REGIME for the weather, the performance strip for whether the system is currently winning, then the matrix sorted by WINS — with EXP beside it to see whether the tier agrees with the record. Anything in LAST RUN ACTIONS happened in the last twenty minutes.