Omni Predictor v4.4.1

Cloudflare Pages → MEXC EDGE direct · REST-only Worker · persistent market archive · interactive chart · hash-bound native indicator adapters · recovery-safe A100 ledger and market core

Pages ready Worker REST waiting MEXC WS connecting EDGE packets waiting L2 waiting Inputs waiting Model blocked Production unverified History waiting Adapter loading Storage starting Indicator engine starting A100 starting

Open Worker

Worker REST endpoint—Waiting for configuration

Health unknownPages WS + Worker REST

Index price

—

Last price

—

Best bid / ask

—

L2 quality

Waiting

EATTK

Waiting

A100 success model

NO TRADE

Probability engine waiting

Data gate

BLOCKED

Waiting for atomic state

Emergency data recovery

This release opens the existing v4.4.0 database without downgrading it, merges legacy modelEvents and new modelLedger records, and does not delete either store.

RECOVERY HOLD

Do not clear site data

Close other Omni Predictor tabs if storage says upgrade blocked. Export the recovery JSON before resuming automatic captures.

Hold all automatic prediction capture

Recovery state

WAITING

Existing database is never downgraded

Recovered records

0 legacy · 0 ledger · 0 merged

Deduplicated by prediction ID

Automatic backup

Not created

Stored inside IndexedDB settings

Recovery error

None

Close other tabs when upgrade is blocked

A100 probability-first 10-minute research and calibration engine

The primary output is the estimated probability that the selected direction settles successfully against the MEXC BTC index price at +600 seconds. The current estimator is intentionally marked uncalibrated until enough resolved MEXC-specific samples exist.

CALIBRATION PENDINGA100-10M v1.3

Recovery and shadow testing only

Raw probability is a deterministic research estimate used to collect labelled outcomes. It is not yet an out-of-sample calibrated success probability and cannot authorize live-capital execution.

Enable A100 evaluation Collect continuous non-overlapping 10m research samples Auto-capture calibrated A100 candidates UP payout % DOWN payout % Manual target Use current index as shadow target

Selected direction

NO TRADE

Waiting for data

Raw directional estimate

—

Score-derived input to calibration

Calibrated success probability

PENDING

Chronological Platt calibration; provisional after sufficient class-balanced outcomes

Required probability

85.00%

Absolute A100 minimum

A100 decision

NO TRADE

Waiting for evaluation

Shadow target / settlement

—

No active prediction

Expected value proxy

—

Available only from calibrated probability

Resolved samples

0 wins · 0 losses · 0 draws

Observed hit rate

—

Descriptive only until sample requirements pass

Raw Brier score

—

Score-derived probabilities

Calibrated OOS Brier

—

Uses probability stored at prediction capture

Calibration evidence

0 / 1000

Provisional model starts after 100 class-balanced outcomes

Qualified A100 candidates

0 / 300

Calibrated ≥85% and every core gate passed

Directional performance

L 0/0 · S 0/0

Separate LONG and SHORT resolved outcomes

Settlement data quality

0/0 exact-window

No current-price fallback is permitted

Model mode

RESEARCH COLLECTION

Production gate remains blocked

Evidence group	LONG	SHORT	Selected interpretation

A100 prediction ledger

Each entry is resolved only from a stored MEXC index sample within ±3 seconds of the exact +600-second timestamp. Missing settlement data is marked unresolved; current price is never substituted.

Time	Type	Direction	Raw P	Cal P	Target	Settlement	Result	Index at settle	Sample Δ ms

Native OmniTrader master chart

The chart renders registered native indicator adapters from synchronized MEXC data. Pine sources are hash-bound to adapters; unregistered scripts cannot influence calculations.

No chart dataPine fingerprint

DemandSupplyOrder blockFVGVWAP familyRealtime 15m predictionPredicted LONG tradePredicted SHORT trade

Multi-timeframe reset VWAP matrix

Each VWAP resets at its own UTC bucket and develops continuously from volume-weighted typical price.

TF	Value	Price position	Slope	Distance	Reset	Next reset	Effect

Live order-flow evidence

Automatically calculated from MEXC L2 and raw aggressive trades. No manual heatmap fields are used in the live score.

60s aggressive buy

—

60s aggressive sell

—

60s delta

—

Top-20 imbalance

—

Bid / ask wall

—

Absorption

—

Replenishment

—

Pulling

—

Sweep

—

Approved OmniTraderPro v28 heatmap

This exact OmniTrader deployment is the only authorized heatmap and order-flow visualization for this project.

AUTHORIZED VISUALIndependent shared MEXC feedOpen full screen

The approved OmniTrader chart and Omni Predictor consume MEXC market data independently. The predictor does not read pixels, scrape the iframe, or require a bridge to the heatmap. Predictor calculations continue from its own synchronized MEXC feed.

Active structural objects

Supply/demand are exact RC4.1 pivot/ATR logic. OB and FVG are native extension modules and remain separately labelled.

Type	Direction	TF	Proximal	Distal	State	Touches	Strength	Age

Historical Data Integrity Matrix

Loaded is not the same as verified. Gaps, duplicates, invalid OHLC and forming candles are reported separately.

TF	Loaded	Coverage	Missing	Duplicates	Invalid	Continuity	Last closed	Forming	Status

Browser-authoritative L2 diagnostics

The Pages browser owns the direct MEXC EDGE connection. Frames are decoded through one serialized queue. The 1,000-level book uses a single-flight state machine for Worker REST snapshot bootstrap, exact incremental versions, depth_commits recovery and snapshot fallback. Book-derived calculations freeze during recovery.

Book version

—

Depth age

—

Matching-engine data age

Applied depth updates

Waiting for synchronized updates

L2 recoveries

Commit recovery and resnapshot fallback

Local feature calculation

—

Browser L2 and flow cycle

Browser adapter

—

Last RC4.1 cycle

Continuous L2 stability

00:00

Waiting for confirmed synchronization

Quality components

0/0/0/0/0

WS / Book / Depth / Trades / Ticker

Atomic packet generation

No market generation rendered

L2 phase

WAITING

Book-derived calculations frozen

Version bridge

— → —

Current → expected next version

Pending version range

— … —

0 buffered WebSocket versions

Commit range

— … —

0 commit/cache versions applied

Recovery duration

—

Not started

Depth row schema

AUTO

Awaiting snapshot sample

Bid price	Contracts	Orders	BTC	Ask price	Contracts	Orders	BTC

EATTK latency chain

Edge Advanced Time Travel Coefficient quantifies freshness and processing headroom. It does not predict the future or remove network delay.

Exchange → Browser

—

Matching-engine evidence age

Local feature cycle

—

L2 and flow calculations

Proxy delivery

—

Zero for direct WebSocket frames

Browser adapter

—

RC4.1 calculations

Render age

—

Visible output freshness

Total detection age

—

Latest evidence to screen

EATTK waiting

High-frequency calculations begin when the Worker feed and history are ready.

Pine source intake & native adapter engine

Paste and store Pine source, audit its dependencies, bind it to a registered native adapter, and block activation until source hash and parity requirements are satisfied.

NATIVE ADAPTER ONLYPENDING

Execution boundary

TradingView Pine Script is not executed inside this Pages application. Source is preserved exactly, but calculations run only through a matching native JavaScript adapter. Unknown Pine hashes stay blocked rather than being approximated.

Indicator name

Pine source

Load or paste a Pine source. Unknown scripts are stored but cannot execute until an exact native adapter is added and validated.

Source hash

—

Exact source identity

Static audit

Waiting

Version, type and line count

Detected features

—

Inputs, plots and alerts

Registered adapter

—

Hash-bound native implementation

Name	Hash	Type	Adapter	Parity

Persistent market-data archive

Raw L2 updates, aggressive trades, index samples, feature snapshots, candles and model state changes are written to browser IndexedDB in batches. The A100 ledger is persistent.

Storage state

STARTING

IndexedDB and persistent-storage request

Usage / quota

—

Browser-reported estimate

Records written

Batched writes this session

Storage errors

Must remain zero

Raw-event retention

6h rolling

L2/trades/features/index samples

Retention scope

The browser stores every received raw L2 update and aggressive trade during the rolling retention window. Candles and indicator sources remain keyed records. Long-term multi-day raw L2 storage requires exporting or a dedicated backend; browser quota is not unlimited.

L2 rawVersion, exchange timestamp, bid updates and ask updates exactly as received.

Trades rawPrice, contract volume, aggressor side, BTC conversion and notional.

FeaturesAtomic quality, synchronized-book state, imbalance, delta, walls and latency.

Model eventsShort-term state changes use rolling retention; the A100 prediction and resolution ledger is stored persistently.

TradingView visual reference

The official public widget cannot display your private RC4.1 Pine source. Use the native chart above for ported indicator values.

Open original

Operator guide

Fast interpretation of the most important values.

BOOK SYNCHRONIZEDConfirmed only after three consecutive valid frames. Required before book imbalance, walls, pulling or replenishment may influence the model. Any unresolved version gap blocks production use.

VWAP alignedPrice is on the same directional side of weighted reset VWAPs. Full alignment is supportive, but overextension or opposing liquidity can still create danger.

Aggressive deltaExecuted buyer notional minus executed seller notional. Positive supports upward pressure; negative supports downward pressure. Absorption can reverse the interpretation.

Supply / demandRC4.1 pivot zones. Fresh demand can support longs; fresh supply can obstruct them. Penetration, repeated touches and invalidation reduce reliability.

Order blockNative extension tied to displacement and structure break. It is not the same object as a pivot supply/demand zone.

FVGThree-candle price imbalance. Unfilled gaps can act as magnets or reaction areas; mitigation percentage is displayed.

A100 probabilityThe headline probability remains PENDING until the chronological class-balanced calibrator is valid. A100 requires calibrated probability ≥85% plus every pre-registered core gate; raw estimates collect research data only.

ALL INPUTS FRESHDepth, aggressive trades and ticker evidence must all be within their freshness limits. A synchronized book alone is not enough.

MODEL READYRequires synchronized L2, fresh inputs, verified history and an active RC4.1 adapter. Grade is evaluated only after this gate passes.

PRODUCTION VERIFIEDRemains blocked until TradingView parity, confidence calibration and Event Index validation are complete.

EATTKHigher means fresher data and lower processing delay. Low EATTK warns that a manual reaction window may already be degraded.

Omni Predictor v4.4.1

Emergency data recovery

A100 probability-first 10-minute research and calibration engine

A100 prediction ledger

Native OmniTrader master chart

Multi-timeframe reset VWAP matrix

Live order-flow evidence

Approved OmniTraderPro v28 heatmap

Active structural objects

Historical Data Integrity Matrix

Browser-authoritative L2 diagnostics

EATTK latency chain

Pine source intake & native adapter engine

Persistent market-data archive

TradingView visual reference

Operator guide

Runtime log

Bind the Worker REST backend