Lesson

01 / Chart first, indicators when needed

Use the top controls for symbol, timeframe and price zoom. 1s, 5s and 2m change candle duration; 1x–500x change the vertical price span, not trading leverage. Log / Linear only changes the price scale.

Dragging, wheel zoom and pinch zoom freeze only the camera, not live candles, trade counts or indicator updates. New candles keep arriving without moving or refitting your view; candles beyond its right edge do not pull you forward. Reset chart restores default time zoom, an automatic logarithmic price scale and camera following, keeping your symbol, timeframe, indicators and login. Auto-follow keeps the current time zoom width. Use History to jump to a saved period. Beyond the 12,000-bar view capacity or a long data gap, the historical view stays intact while the latest price above keeps updating; reset to see the latest candles. Expand Indicators for studies or sound. Closing the panel does not disable a study.

🔥 Trending / Server TPS acceleration watchlist

AWS continuously scans up to 80 high-turnover Bybit USDT markets and shows up to three candidates ranked by acceleration, traded value and symbol. This is separate from Combined Top 10 divergence, not a whole-market ranking or trade advice. Closing the page does not stop server discovery or Auto capture; Safari tab changes, reloads and symbol changes do not control it.

Each server sample compares rolling one-second trades against 30 prior one-second samples, excluding the current sample. Baseline ≥2 TPS, current ≥25 TPS, acceleration ≥3× and value ≥25,000 USDT must hold for two consecutive samples. Disconnects, long sampling gaps, universe changes and stale trades clear candidates or restart warmup; missing observations are never invented.

TPS uses server receipt time, not gap-free exchange-clock candles. Network batching can affect counts, and live candidate timing need not match the former browser receiver. These thresholds are not win-rate validated and do not infer price direction.

Auto capture is a saved server-wide administrator control. Only the top candidate is considered, at least 15 minutes between automatic admissions. BTC, ETH, ZEC, SOL, XAU, CL and XRP are already permanent; when one is top, #2 is not substituted. Each new admission receives a 24-hour recording lease and continues after rank exit. Up to 8 extra coins are retained, with no eviction when full. Ranking and polling never renew an active lease; only explicitly reopening the coin as administrator renews 24 hours. Admission after expiry begins a new recording period, without restoring expired history.

Trending toggles only this page display, never the server scanner. Only an administrator can change Auto capture; its shared state is stored on the server and read by other tabs. Old browser preferences cannot overwrite it. Auto off stops new automatic admissions, not existing leases or data. Explicit symbol-reopen renewal is unchanged, and Reset indicators does not alter the server control.

The page reads results about every 2 seconds; it never nominates symbols. The background controller rechecks the fresh top candidate, instrument, Auto revision, 15-minute hold and remaining capacity; stale or conflicting requests do not write. Requested is not Recording, which requires fresh collector evidence. Outages show unavailable rather than claiming capture; the saved control intent survives and recovery considers only current candidates.

02 / Trade count is not average TPS

When signed in, your candle timeframe is saved to your account. Choose 2m and it returns after refresh, sign-in, symbol changes or signing in on another device. Reset chart and Reset indicators do not change it. Guests remember their timeframe in this browser only. If Timeframe not synced appears, click Retry. A new account without a saved choice starts at 5s. Candle timeframe and display refresh speed are separate settings.

The histogram sums trade prints in each candle; it does not average them. One-second TPS is the count in that original second. They coincide at 1s, but differ on 5s, 1m and longer candles.

Faded bars are unfinished or have incomplete coverage. Blank space means no record, not zero trades. Changing timeframe cannot fill missing observations.

03 / ⚡ TPS & directional streaks

⚡ requires a complete original second with more than 3,000 trades. Exactly 3,000 does not qualify. A 5s total above 3,000 does not mean any individual second qualified. Buy / Sell is set by net taker value, not candle colour.

🎯 Sniper projects immediately on receipt of a complete lightning second: its closing price ×1.007 for net taker buying, ×0.993 for selling. It does not wait for a price touch and never uses the enclosing 5s / 1m candle close. Example: from 80,000, the buy target is 80,560 and the sell target 79,440.

If the legacy BTC bridge omits the close, it is filled only from matching original seconds in the same recorder. Time, source, high/low, count and net taker value must all agree and coverage checks must pass. Unmatched prices are never inferred; the chart warns about missing targets.

The ⚡ label shows the final trigger’s directional target in that candle; each evidenced trigger second keeps its own dashed price level, starting after that second ends. Timeframe changes do not recalculate its price. Draws the latest 120 targets, prioritizing newer overlapping labels; crosshair details show the last 8 in a candle. Targets never move the price scale; zoom out vertically to see off-screen levels. 🎯 shares the saved ⚡ toggle and reset behavior.

Requires the original one-second close and net taker value, with matching raw events and candle summaries. Balanced flow, missing prices, incomplete or mismatched records do not project; later prices never fill missing anchors. Shares the 5s signal poll, so delivery is not zero-latency. Revisions may update targets; a historical marker does not prove contemporaneous receipt. 0.7% is a price distance, not win rate, account return, touch confirmation or automatic take-profit. No changes to speech or Signal API.

Each qualifying same-side second adds one; an opposite qualifying second resets to #1. Sub-threshold, exactly 3,000 and balanced-value seconds neither add nor reset. There is no fixed four-hour reset. ≥ means the start of the run is unknown; ↔ counts direction changes within a candle.

This is a sequence of recorded signals, not proof that no opposite event was missed. It is not an entry instruction.

▭ TPS boxes extend the high / low of complete seconds above 1,000 trades to the right. Cyan is net taker buying; magenta is net selling. Touches do not remove a box, and boxes do not guarantee support, resistance or reversal.

04 / Trade-count VPVR

Bars show cumulative trade counts at each price, not counts divided by seconds. Only matched complete seconds are included.

The right-side bars group covered trades in the visible range into price bins. They show print counts, not traded value or resting depth. 24 / 48 / 96 bins control granularity. Yellow POC is the bin with the most trades.

Panning and zooming change the visible range, so the profile and POC can change. Partial coverage includes only recorded data, without estimating missing seconds.

05 / Session VWAP

Off on first use or after Reset indicators; subsequent toggles and settings are remembered. Choose HLC3, HL2 or Close as the minute price source, then accumulate with 1m volume weights. This is not an exact tick-by-tick VWAP. It resets daily at your chosen UTC time.

σ uses deviation bands; % uses percentage bands. Set bands 1–3 from 0 to 10; zero hides a band. Apply saves the parameters. Values are confirmed after minute close; gaps are not estimated.

Auto Cycle / 4–5d automatic cycle fit

06 / Order flow, pressure & regime

Taker percentages sum buy and sell values in the selected window before dividing; they do not average per-second percentages. Taker delta / window CVD is buy value minus sell value.

Taker flow is already-executed trading; bid / ask depth is visible resting liquidity. They are different. Regime is a rule-based classification of flow, price response and coverage, not proof of market-maker intent. Incomplete data should be read as undetermined.

Regime v1: absolute value imbalance ≥60% and a same-direction price move ≥2 bps indicate flow-driven activity. An absolute price move ≤1 bp suggests possible absorption, not confirmed absorption or a reversal prediction. Relative activity compares rates with the previous 300s; trade-count bars are totals. Price response describes observed moves, not future results. Shares are blank when no trades occur.

Order book data uses one-second snapshots, not full event replay. Book age and snapshot lag are each at most 1s; live trades allow 3s to complete. Books use known time; trades use exchange time. Missing or stale values appear as —. Liquidations lag the trade window by 1s.

07 / 🧽 Absorption & 🧊 Iceberg

Absorption candidate: strong net selling with little downward response maps to a buyer candidate; strong net buying with little upward response maps to a seller candidate. It uses a baseline on the selected timeframe, so changing timeframe recalculates it.

Iceberg uses a separate BTC / ETH order-book and trade observer to find executions, depletion and later replenishment at the same best price. Markers appear after detection, not back at the first touch.

🧊 identifies Bybit / OKX / Hyperliquid evidence. Binance Iceberg detection is removed; archived records are preserved. BTC / ETH perpetual books and trades are evaluated separately, never merged across venues. Candles remain Bybit; other-venue annotations use detection time, with the actual venue price in details. Not ready does not mean no iceberg.

OKX uses source=0 normal aggregate messages and official contract values; missing classification fails closed. Hyperliquid uses native USDC perpetual l2Book snapshots and trades: a baseline precedes execution, visible depletion and later refill; duplicate or stale snapshots cannot count as refills. The USDT symbol is a chart-routing alias, not a quote conversion. Sampling differs, so event frequencies are not directly comparable. Signal API Iceberg remains Bybit-only.

Both are suspected candidates. Anonymous L2 cannot prove a hidden order, its owner or remaining reserve. A level decrease cannot simply be called a cancellation.

🧹 Sweep · Suspected multi-level sweeps

Green buy sweeps lift asks; pink sell sweeps hit bids. They describe observed multi-level executions and book changes, not Long / Short instructions. Available for Bybit BTC / ETH, sharing the public L2 order-book and trade capture used by Iceberg.

Strict rule v1: at least 5 same-direction prints within 250ms, consuming at least 3 initially displayed price levels, with a span of at least 1bp. Executed value must reach 100,000 USDT for BTC or 25,000 for ETH. The initial book must be no older than 200ms; a later DOM update must confirm multi-level consumption within 1s. Same-side cooldown is 2s. Only eligible public trades qualify; reconnects, old snapshots and unmatched observations are not stitched into events.

Charts default to Large: BTC ≥1,000,000 / ETH ≥1,000,000 USDT. Whale uses BTC ≥2,000,000 / ETH ≥2,000,000; All restores the capture floor (BTC ≥100,000 / ETH ≥25,000 USDT). This setting is remembered; Reset indicators restores Large. Size is filtered before the latest-500 cap so small events cannot crowd out larger ones. Raw candidates are retained. Value is the sweep episode total, not proof of one order. Larger value reduces noise; higher win rate or returns have not been established.

Markers appear at the first interval boundary at or after detection, never back at the first print. Changing timeframe changes placement without recalculating events. Crosshair details show start/end prices, value, prints, levels, span, duration and detection time. Enabled visible charts refresh every 5s, reading up to 500 events from the latest 7 days of the viewport and drawing the last 120 with real candles. The toggle stays in this browser and defaults on at first use and after Reset indicators. Other symbols show unavailable.

This is a suspected-sweep heuristic, not a confirmed single large order. Public L2 cannot prove order ownership; multiple participants and cancellations may look similar. It guarantees no continuation, reversal, entry direction or win rate. No qualifying events can mean no markers. Evidence is recorded only after this feature goes live; old candles and earlier Iceberg records cannot reconstruct sweeps. Continuous capture does not guarantee complete trade coverage. Delayed or not-ready status means insufficient evidence, not proof that no sweep occurred.

⏱️ TWAP · Periodic execution candidates

The server keeps BTC / ETH / SOL / ZEC / XRP and dynamically adds Combined TPS Top 10 exact Bybit / OKX USDT perpetual matches, up to 15 symbols. Ranking is read about every 10s; departures get 15 minutes of grace and newcomers queue when dynamic slots are full. Unsupported contracts are not replaced with lookalikes. Entrants need fresh trade samples: 2m candles cannot backfill TWAP. Thresholds use USDT; OKX contracts are converted to base size and cadence is never combined across venues. Monitoring continues with the page closed; only candidates are saved. Taker direction is not an entry instruction.

Strict mode preserves the original rule: group by venue, symbol, taker side and exactly equal base-coin quantity, requiring at least six consecutive prints within that group, not a fixed dollar amount. Intervals are 5–300 seconds, with each gap within ±10% capped at 1 second and cumulative phase error from the first print no more than 1 second. Every print is at least 10,000 USDT and the six-print total at least 60,000 USDT. Each uninterrupted run is marked once; broken intervals, reconnects or invalid evidence cannot be stitched into one run. Existing Bybit strict events remain available.

Adaptive mode groups same-side short bursts spanning at most 200ms into batches, finalized only after 200ms without another print; overlong continuous bursts are excluded. It requires eight batches, each at least 10,000 USDT and at least 80,000 USDT in total. Each base quantity must stay within ±20% of the median. Cadence is estimated from the median of multiple intervals, which must be 5–300 seconds; each gap may deviate by ±25%, capped at 5 seconds. Regular / Variable describes the observed evidence, not accuracy or probability. Batches may mix different traders or miss orders split over longer periods.

Cadence uses exchange execution timestamps. Strict confirmation occurs only when the sixth print has arrived and become knowable; adaptive confirmation waits until the eighth batch is finalized. Markers go on the first candle boundary at or after confirmation, never back at the first print. Changing timeframe only changes placement, not detection. Crosshair details show venue, mode, base quantity, interval, value, start and confirmation times; adaptive events also show all eight batches. Historical candles and two-minute archives cannot reconstruct this print evidence; no markers are backfilled before launch of each new observer.

Two independent observers save compact events, not the full tape; current health is reported separately. Each event store retains up to 30 days and 100,000 events, removing the oldest records at whichever limit comes first. Signed-in Premium or administrator users with a visible, enabled chart refresh every 5 seconds. Each view reads at most 500 combined events over its latest 7 days and draws the last 40 aligned with real candles. The toggle and All / Strict / Adaptive selection are saved in this browser; defaults On at first use and after Reset indicators, with All selected. Both venues report health separately; one delayed venue cannot make the other appear healthy. Other symbols show unavailable; delayed or not ready is not evidence that no TWAP occurred.

Common lot sizes, multiple participants and coincidences can produce regular prints; randomized genuine TWAP may be missed. Public print cadence cannot confirm a TWAP algorithm, one account or a parent order. Strict and Adaptive may identify the same pattern, not two parent orders. There is no guaranteed direction, win rate or return, no order connection or automatic trade.

The page-level TWAP alert is separate from chart indicators, defaults On and remembers its sound setting. It receives new events for the actual server roster every 5s, independent of the chart symbol: five permanent coins plus dynamic Top 10. Missing exact contracts never count as ready. First load, reconnect and return from browser suspension never replay old alarms. A compact red warning stays until manual dismissal; Alpha repeats every 3s. Browsers can restrict playback: use Test to check speaker output. The server monitors continuously and sends new candidates to the linked Telegram group; a closed page cannot guarantee an immediate browser sound.

Official references: Bybit public trade fields · Bybit TWAP strategy · OKX API · All trades · OKX TWAP strategy

💥 CL / XAU · Net taker anomalies

The server independently reads recorded Bybit CLUSDT and XAUUSDT seconds and aggregates taker buys and sells into complete 2m windows. Net value = taker buys minus taker sells, in USDT: positive is net buying and negative is net selling. Direction never comes from candle color or TPS. This is not a single oversized print, TWAP or an entry instruction.

Rule v1 uses the mean absolute net value of the previous 20 strictly complete 2m bars within 6 hours, excluding the candidate itself. Its absolute net value must be at least 3× baseline and its dominant side at least 65% of gross traded value. Same-symbol/same-side events are separated by at least 10 minutes. CL and XAU are evaluated independently.

Every candidate needs 120 strictly covered seconds. Missing seconds are not filled with zero, and price-only candles cannot supply taker value. First evaluation waits at least 15 seconds and occurs no later than 90 seconds after close; stale backfilled data does not generate catch-up alerts. Insufficient baseline shows Warming; incomplete candidate coverage shows Incomplete, not Ready merely because a collector is connected. Details expose baseline count, latest closed coverage, mean absolute net value and the threshold.

💥 Taker CL/XAU defaults On; its independent toggle and volume stay in this browser, not across devices. Off disables new page alerts but still shows the latest 20 server records and never stops detection. A small red warning shows symbol, net buy/sell, value, multiple and first-confirmation time. At most four symbol/side groups are retained, showing each group’s newest event. Manual dismissal does not disable future alerts, and switching charts does not dismiss them.

The page receives about every 2 seconds. Initial load, reload, reconnect, service epoch change, returning or long suspension silently establish the latest cursor and never replay history. Sound admits only new events within 15 seconds of first confirmation. Ordinary clicks prepare the player silently; only real alerts or explicit Test play audio. Detection continues with the page closed, but closed/suspended pages or browser audio restrictions prevent guaranteed sound. This is not system push notification.

This describes an observed net-taker anomaly, not proof of opening/closing a position or an imminent reversal. It guarantees no price direction, win rate or profit. Warming, gaps or no records do not prove the market had no anomaly.

🥵 / 🥶 Overextension warning

↔ Three swings → Net-flow divergence → Reversal candidates

3-swing levels use the third high for bearish setups and the third low for bullish setups, extending horizontally right with a price label. They appear only after confirmation in the selected mode, beginning at confirmation time rather than backdating knowledge to the pivot. Gray structures and unconfirmed amber divergence draw no level. The switch defaults on, is remembered in this browser and returns on with Reset indicators.

Only the latest 20 confirmed levels within loaded history are shown. New prices do not move a level, and a price crossing does not delete it. Symbol, timeframe and filter changes recompute references; history eviction, correction or backfill may change them. This is not a permanent signal ledger. Lines do not change chart zoom; out-of-view prices stay off-screen rather than being pinned to an edge. Structure reference only, not an order or automatic stop-loss.

Layered display: gray dashed ①②③ lines show three rising highs / falling lows, not divergence by themselves. Amber lines mean net-flow divergence is present but the selected mode’s confirmation conditions are not fully met. Only all conditions for that mode produce 🔴⬇️ Short? / 🟢⬆️ Long?. Falling trade-count histogram peaks are not a substitute for net taker-value divergence.

Short candidate: three consecutive rising swing highs with progressively weaker net taker buying around each high, elevated third-pivot activity and the second right-side candle closing below the second high. The Long candidate mirrors this with falling lows, weakening net taker selling and a close above the second low. A rejection wick and the third pivot’s own close-back are extra requirements only in All / Strict modes, not the default Volume mode. Strength means taker buy value minus sell value in USDT, not RSI or a direction inferred from high TPS alone.

Calculated on the selected timeframe; start with 2m for comparison, but 2m is not forced. Place the crosshair on a numbered pivot or confirmation candle to inspect the three times, prices, net taker values, activity ratio, wick and unmet conditions for the selected mode. Structure, flow-divergence and confirmed-candidate counts are separate.

①②③ and connecting lines are retrospective annotations: only after the third pivot’s two right-hand bars close are the past extrema drawn. They do not mean the pattern was known at ①, ② or ③. Long/Short emojis stay at the earliest knowable boundary, never back at the third extreme, and require an actual candle there. At 2m this is at least 4 minutes after the third candle closes (6 minutes after it opens), plus delivery/refresh latency. Structure and final qualification are assessed at that same boundary using closed evidence; future prices do not upgrade old signals.

The overlay draws at most 12 visible line groups, prioritizes flow-divergence/confirmed groups and merges overlapping labels; at most the latest 60 confirmation emojis are shown. Settings counts cover loaded data, not necessarily simultaneously drawn groups. Pan/zoom and symbol/timeframe changes reposition annotations; lines do not move the price scale or change trade totals.

Only contiguous, complete, verified strict_ws / archive evidence is evaluated; gaps, unfinished bars or ineligible sources fail closed. Late, backfilled or corrected data can revise historical candidates. Chart time does not prove actual receipt and is not a saved live-order ledger. The screenshot lacks full numerical rules, so this is an explicit price-pattern + trade-flow research proxy, not an exact video-strategy replica. POC reversion targets are not implemented; stop hunting, manipulation, hidden liquidity and a normal distribution cannot be established. Zero candidates is valid. No win-rate or return guarantee, orders, existing TPS English counts or Signal API changes. The divergence bell below is a separate, optional notification.

08 / 🔊 English voice counts

Use Voice count in Indicators to choose a voice, try 1 2 3 4 and adjust volume. On on first use or reset; a saved Off choice is preserved, but each new page needs a click on Activate voice or Try. Known English female voice names are preferred; availability varies by browser. Speaking appears only after the browser reports playback started. Failure to start or interrupted playback prompts a retry, never an automatic replay. A successful preview restores remembered On alerts; remembered Off stays off.

Volume 0 is labelled Muted. If Speaking appears but you hear nothing, check tab muting, system volume and the selected audio output. Try works while browsing history, but real-time counts still require live-follow, ⚡ TPS and fresh data; use Reset chart after dragging. Old signals are never replayed, and counting rules are unchanged.

Indicator toggles, VWAP parameters, VPVR bins, order-flow window, English counts and divergence bell settings stay in this browser only, not across devices. Clearing site data or private browsing may discard them. Reset indicators turns VWAP off and English voice counts, the divergence bell and other indicators on, restores 48 VPVR bins, VWAP HLC3 / σ / 1, 2, 3 / UTC 00:00, 30s order flow, the preferred English voice at 70% volume and the bell at 30%. Audio still needs activation on this page. The bell attempts recovery through ordinary page interactions; English counts still use Activate voice or Try. Symbol, timeframe, zoom, login and trading rules are unchanged. Membership and supported-symbol restrictions still apply.

Speech uses the ⚡ counter: One, Two, Three… for the same side, then Buy / Sell, One after a reversal. Unknown-start counts say At least. Only a visible live-follow chart with fresh data speaks. Historical browsing, late repairs, timeframe changes and baseline snapshots do not replay.

Reads sync every 5s, with at most the latest four events and no long queue. Expired alerts are dropped. It is not complete per-event delivery or an order prompt. Collapsing settings keeps speech enabled; closing or hiding the page stops it.

09 / Sources & full detection rules

Market Chart: single-symbol historical candles and their trade counts come from Bybit public trades and historical archives, grouped by exchange timestamps. Missing records remain blank, are not treated as zero trades and are not filled with another venue’s data. Built with TradingView Lightweight Charts; full TradingView drawing and Pine features are not included, and the chart-library attribution is retained.

Venue screens show the exchange and market type indicated by their labels. Spot and perpetual contracts are distinct markets; prices, trade units and message aggregation can differ. Combined presents statistics from enabled venues, not a consolidated price across exchanges; the single-symbol Market Chart is not an all-venue composite price either. Cross-venue events retain their original venue labels and provenance; evidence must not be stitched across sources.

Legacy live records are marked separately and have not been reverified second by second. Recorded coverage does not guarantee a gap-free trade feed.