EzOrderFlow · Details: datafeed, metrics, naming

Disclaimer

EzOrderFlow is a descriptive product. It shows public market data (order books, executed trades, liquidations, open interest, venue announcements) and derived views computed from them. It contains no trading signals, no recommendations, and no financial advice.

Only the liquidation zones are model output, and they are always labeled "estimated". Everything else draws what was recorded.

Data limits to know

Bitget fills-history caps at 100 fills per call, so Trades/Prints are not offered for tail-tier coins; the cells would be silently partial.
Depth and aggregate-heat rows are 5-minute snapshots: a level that appeared and vanished inside one slot leaves no record.
Band kinds (depth, xdepth, prints, fp, liqs, oi) keep ~1 week hot before archiving.
Liquidations come from other venues' public feeds: a proxy for this market, not its complete set.
Liq zones assume positions entered near the price at the moment OI rose; entry-price distribution is not directly observable.

Free to use. Charting library: TradingView Lightweight Charts™.

Metrics: what each panel computes

∿ Volume bars, Delta, CVD

Aggressive (taker) volume per candle, split buy/sell. History replays from our collector archive (Bitget taker-buy-sell rows, per-minute footprint cells on tracked symbols, and Binance kline taker split where the pair exists), summed over the selected venue set (default: all venues). Where no split was measured, the bar still paints the candle's total volume in its own direction (green on up, red on down): Delta adds nothing there, while CVD counts that volume by the inferred side.

V_i^buy, V_i^sell: taker volume lifted from offers / hitting bids in candle i
Δ_i = V_i^buy − V_i^sell (0 when no split is measured)
CVD_i = Σ_j≤i Δ_j, with unsplit candles counted as +V/−V by direction

What it is not: aggressor side, not position opening/closing: a positive delta means buyers lifted offers, not that price must rise.

≡ Depth

Live order-book histogram beside the price axis. The book buckets into rows at a configurable price step; each row sums resting quantity.

bucket(b) = Σ_{i : p_i ∈ [b·s, (b+1)·s)} q_i, s = step% × visible range
wall(b) ⟺ bucket(b) ≥ k · median(bucket), k = 3 by default

What it is not: resting size, not a queue position; orders can be pulled before price reaches them.

△ Walls

The Bitget wall lifecycle trail. Bands open at levels that qualify as walls (bucket ≥ threshold × median, top-percentile and minimum-size gates), persist while the level keeps resting, and close when a candle trades through the price or the level is pulled. Replays from recorded depth rows; a stored band only opens at wall grade (≥3× median).

Screenshot of the app: order-book wall bands above and below ETHUSDT candles

What it is not: a wall is resting size, not intent; it can be pulled at any moment.

▦ Heat

Cross-venue aggregate liquidity cloud. The collector merges book snapshots from all reporting venues per 5-minute slot into one aggregate book (levels beyond ±50% of the reference mid are dropped as implausible), bucketed at a fixed 0.05% step. Brightness is each level's share of the largest level that minute; stored Bitget depth rows fill slots the aggregate missed.

book_agg(p) = Σ_{venues v} book_v(p), per 5-min slot
intensity(p) = book_agg(p) / max_p book_agg(p)

What it is not: not one executable book: venues trade at their own prices; the cloud records where size rested.

◉ Trades

Bubbles at (time, price) for large executed fills. Dense-tier sweep keeps fills ≥ $25k notional per venue, deduped by venue+trade id; the view then draws the top share of loaded dots (configurable, plus per-candle merge and minimum-size filters).

Screenshot of the app: size-labeled trade boxes on ETHUSDT candles, one hovered to show fills

stored ⟺ p·q ≥ $25,000
drawn ⟺ q ≥ top-% quantile of loaded dots

What it is not: a big fill is a fact about the past; it does not say the same actor will trade again.

◫ Prints (footprint)

Per-price lanes inside each candle: every deduped fill aggregates into minute × price cells as {buy, sell}; cells tile into the candle they belong to. Stored fp rows cover history; the live trade stream owns the current minute (the boundary minute's stored tail is skipped to avoid double-counting).

cell(m, p) = { Σ q_buy, Σ q_sell } over fills with minute(m) = m, price(f) = p

What it is not: buy/sell is aggressor side: it says who lifted or hit, not who opened or closed.

▼ Liqs

Liquidation markers from public venue feeds: Binance forceOrder and Bybit allLiquidation live, OKX REST for recent history. Bitget exposes no public liquidation feed. Events in one candle slot and side merge into a single marker at the quantity-weighted price.

Screenshot of the app: liquidation markers on ETHUSDT candles with a detail tooltip

marker(candle, side) = { Σ q_i at p̄ = Σ q_ip_i / Σ q_i }

What it is not: a proxy: foreign-venue liquidations are venue-reported coverage, not this market's complete set.

▨ Liq zones (estimated)

When open interest rises, the model treats the increase as new positions entered near that moment's price and maps them through standard leverage tiers: a 10x long liquidates ~10% below entry, a 50x ~2% below. Zones erode where later candles sweep through (wicks count) and die on a full cross ('sweep') or when OI drains the position set ('fade'). A single OI jump beyond 50% of the larger side is dropped as a data artifact; each drain event is floored at f ≥ 0.5.

Screenshot of the app: estimated liquidation zones shaded around ETHUSDT candles

liq distance ≈ entry · (1/L), L = leverage tier
zone dies ⟺ candle range crosses the band fully, or OI drains the mapped positions

What it is not: a model's estimate, never presented as measured liquidity.

◧ Imbalance zones

A candle births a zone when its displacement score clears a rarity gate; the zone spans the candle's range, projects forward, erodes as later candles trade back into it, and dies on a full cross. Demand zones sit below price, supply above.

Screenshot of the app: buy and sell imbalance bands around ETHUSDT candles

score = 0.30·bodyFrac + 0.25·volRatio + 0.25·Δshare + 0.20·wickGap
birth ⟺ score ≥ q₉₅(recent scores) ∧ score ≥ 0.25

What it is not: a record of where one-sided flow displaced price, not a forecast of where it will again.

⚑ Events

Exchange announcements and scheduled events plotted as flags on the time axis. An LLM scores each headline's likely market-wide impact (plus a coin-scoped verdict); items awaiting a score render gray ("pending"). Listings and delistings come from the contract-universe diff.

What it is not: the score is a classifier's read of a headline; the chart shows events, never interpretations of direction.

Γ Options levels

Strikes where option open interest and gamma concentrate across Deribit, OKX, Bybit, Binance, Paradex and Delta, merged every 5 minutes: green dashed lines = net positive GEX, red = net negative, amber = the zero-gamma flip; triangle count = strength. Toggle lives in ƒx Indicators → Options levels. Covers BTC, ETH, SOL, XRP, DOGE, BNB, HYPE.

GEX(K) = γ_K·OI_K·S²·0.01 (call−put, dealer sign)
stale venue > 2 sweeps old is excluded and listed, never silently blended

Screenshot of the app: options gamma levels with the zero-gamma line over BTCUSDT candles

What it is not: an options feed on every symbol: only the 7 tokens that have listed options produce levels; everything else shows n/a.

◇ Effort

How much volume each bar needed to move its range: filled diamonds ◆ mark absorbed bars (top-ranked effort, small displacement), hollow diamonds ◇ mark vacuum bars (thin effort, large displacement). Flags sit on the candles; hover one for the effort index, taker delta, OI and result breakdown.

eff = V/(H−L bps); res = |C−O| bps
absorbed: pct(eff) ≥ 90 ∧ pct(res) ≤ 35; vacuum: pct(eff) ≤ 10 ∧ pct(res) ≥ 95 ∧ res ≥ 15bps

What it is not: not a read of who wins next: absorbed bars mark where effort was soaked, vacuum bars mark where the book was thin.

Indicators: definitions and formulas

All indicators compute over the loaded candles only: close = C, high = H, low = L, open = O, volume = V, typical price TP = (H+L+C)/3, HH_n/LL_n = highest high / lowest low over n bars, pct_w(x) = percentile rank of x against the last w bars, OI = open interest, Δ = taker delta (V^buy − V^sell). Overlays draw on the price pane; the rest get their own pane. Defaults in brackets; line width and opacity exist on every indicator.

Overlays on the price pane

SMA [50]

Simple moving average of closes; one line per period entered.

SMA_n(i) = (1/n) · Σ_j=i−n+1..i C_j
n: the averaging window, in bars

Settings: Periods (comma list, one line each)

EMA [21, 50, 200]

Exponential moving average; reacts faster than SMA on the same length.

EMA(i) = α·C_i + (1−α)·EMA(i−1), α = 2/(n+1)
α: smoothing factor derived from n

Settings: Periods (comma list, one line each)

Bollinger Bands [20, ×2]

A mean band whose width follows close-price volatility.

mid = SMA_n; upper/lower = mid ± k·σ_n
σ_n: population stddev of closes over n; k: band width in sigmas

Settings: Period n, Multiplier k

VWAP [15 rolling, 0 = session]

Volume-weighted average price: where the average traded coin changed hands.

VWAP = Σ(TP_i·V_i) / Σ V_i
period > 0 rolls over n bars; 0 anchors at each UTC day

Settings: Period (0 = session), Candle-tint on/off

Supertrend [10, ×3]

ATR-based trailing band that flips direction when a close breaks through it.

bands = (H+L)/2 ± m·ATR_n
m: ATR multiplier; a ratcheted band flips when a close crosses it

Settings: ATR period, Multiplier m

Parabolic SAR [0.02, 0.2]

Stop-and-reverse dots that trail the trend and accelerate on new extremes.

SAR_i = SAR_i−1 + AF·(EP − SAR_i−1)
AF: acceleration factor, steps up per new extreme, capped at max; EP: extreme price

Settings: AF step, AF max

Ichimoku [9, 26, 52]

Japanese cloud system: fast/slow midlines, projected cloud, lagging close.

Tenkan = mid(HH_a, LL_a); Kijun = mid(HH_b, LL_b); Senkou A = (Tenkan+Kijun)/2 + b bars; Senkou B = mid(HH_c, LL_c) + b bars; Chikou = C − b bars
a, b, c: Tenkan, Kijun, Senkou B periods

Settings: Tenkan, Kijun, Senkou B periods

S/R levels [3 pivots, 30%]

One support and one resistance derived from the visible-window volume profile, snapped to nearby swing points.

boundary sits where bucket volume < q quantile of buckets
then snapped to the nearest swing high/low on the correct side of price

Settings: Pivot bars, Volume quantile %

Swing points

Marks local highs/lows: the bars that extremes were printed on.

swing high at i ⟺ H_i = max(H_i−k..i+k); mirror for lows
k: pivot lookback on each side

Settings: none beyond width/opacity

Options levels [GEX · all · 4]

Where option open interest and gamma concentrate across six venues (Deribit, OKX, Bybit, Binance, Paradex, Delta). Green dashed = net positive GEX, red = net negative, amber = zero-gamma flip; triangles count = level strength. Available on BTC, ETH, SOL, XRP, DOGE, BNB, HYPE: the tokens with listed options.

GEX(K) = γ_K·OI_K·S²·0.01 (call−put, dealer sign convention)
zero gamma: spot where ΣGEX crosses zero; walls: max call-OI / put-OI strikes
γ: Black-76 gamma from mark IV when the venue does not quote one

Settings: Source (GEX call−put / |GEX| / OI / volume), Expiry (all / 0dte / ≤1d / ≤7d / ≤31d / ≤91d), Levels (1–8), Zero-gamma on/off

Effort [100 · 90% · 15bp]

How much volume a bar needed to produce its move: filled ◆ marks absorbed bars (heavy effort, tiny displacement), hollow ◇ marks vacuum bars (tiny effort, big displacement). Diamonds anchor on the candle: ◆ above, ◇ below; red = sell-side effort, green = buy-side. Hover a flagged bar for the breakdown (EFF index, Δ, OI, result bps).

eff_i = V_i / (H_i−L_i in bps + 0.5); res_i = |C_i−O_i| in bps
absorbed ⟺ pct_w(eff) ≥ P ∧ pct_w(res) ≤ (100−P)+25
vacuum ⟺ pct_w(eff) ≤ 100−P ∧ pct_w(res) ≥ 95 ∧ res ≥ floor

Settings: Rank window w, Flag pct P, Min move bps (vacuum floor)

Lower panes

RSI [14]

Wilder's relative strength index: momentum between 0 and 100.

RS = avgGain / avgLoss (Wilder-smoothed); RSI = 100 − 100/(1 + RS)

Settings: Period

MACD [12, 26, 9]

Trend momentum from two EMAs: fast minus slow line, its averaged DEA line, histogram.

DIF = EMA_f − EMA_s; DEA = EMA_g(DIF); histogram = DIF − DEA
f, s, g: fast, slow, DEA periods

Settings: Fast, Slow, DEA

OBV + MA [30]

On-balance volume: cumulative signed volume, with a moving average on top.

OBV_i = OBV_i−1 + sign(C_i − C_i−1)·V_i; MA = SMA_n(OBV)

Settings: MA period

Volume MA [5, 10]

Moving average of raw volume, drawn inside the volume pane.

VMA_n = SMA_n(V)

Settings: Periods (comma list)

ATR [14]

Average true range: the bar's real size including gaps, in price units.

TR = max(H−L, |H−C′|, |L−C′|); ATR = Wilder-smoothed TR
C′: previous close

Settings: Period

MFI [14]

Money flow index: RSI weighted by volume instead of close-only.

MF = TP·V; MFI = 100 − 100/(1 + ΣMF⁺/ΣMF⁻) over n bars, split by TP direction

Settings: Period

KDJ [9, 3, 3]

Stochastic oscillator with the J line: where close sits inside the recent range.

RSV = (C − LL_n)/(HH_n − LL_n)·100; K = (2K + RSV)/sK; D = (2D + K)/sD; J = 3K − 2D
sK, sD: smoothing periods

Settings: Period, sK, sD

StochRSI [14, 14, 3, 3]

Stochastic applied to RSI: sharper, mean-reverting oscillator.

stoch = (RSI − min RSI_p)/(max RSI_p − min RSI_p); %K = SMA_kP(stoch); %D = SMA_dP(%K)

Settings: RSI period, Stoch period, %K, %D

DMI / ADX [14]

Directional movement index: +DI/−DI measure sidedness, ADX measures trend strength.

±DI = 100·sm(±DM)/sm(TR); DX = |+DI − −DI|/(+DI + −DI)·100; ADX = RMA_n(DX)
sm: Wilder smoothing; DM: directional movement

Settings: Period

CCI [20]

Commodity channel index: deviation of typical price from its mean.

CCI = (TP − SMA_n(TP)) / (0.015 · mean |TP − SMA_n(TP)|)

Settings: Period

Williams %R [14]

Where the close sits inside the recent high-low range, flipped to −100..0.

%R = (HH_n − C)/(HH_n − LL_n) · (−100)

Settings: Period

BIAS [20]

Distance of the close from its moving average, in percent.

BIAS = (C − SMA_n)/SMA_n · 100

Settings: Period

Open interest

Total open interest across venues at 5-minute slots, from the collector pipeline.

OI(slot) = Σ_venues OI_v(slot)

Settings: none beyond width/opacity

What they are not: indicators are arithmetic over the same candles you see; they describe shape and momentum, they say nothing about future direction.

Filter: the symbol screener

The symbol panel's Filters button opens the screener. Matches land in the Filtered tab; the other tabs are fixed views: All, Watching (your ★ list), Top rising and Falling by 24h change.

Market filters: last-24h tickers

Filter	Reads
24h move ≥ x %	Absolute 24h change in either direction
24h volume ≥ $x M	USDT turnover over the last 24h
vs avg day ≥ x×	Today's 24h volume against that symbol's own daily baseline

Candle filters: computed per symbol

Filter	Reads
Price move over N bars	Net \|Δ\|, up-only or down-only percent move across the last N candles of the chosen timeframe
Volume spike ≥ x× avg	Recent N-bar volume against the per-bar average
MACD cross within N bars	DIF × DEA crossing, bullish / bearish / either
RSI(14) below / above level	Classic oversold / overbought gate
Scan depth: top N	How many top-volume symbols the candle filters fetch and test

Candle filters only scan the top symbols by 24h volume; the scan-depth control bounds how deep the scan goes.

Settings reference

Every panel's settings live in its ⚙ block; all values persist per browser. Shared controls: opacity (peak transparency of the strongest element, weaker ones dim proportionally), price step (% of the visible range per bucket row; coarser merges levels, not just thickens lines), pane height and scale margin for lower panes.

Panel	Setting	Default	Effect
Volume / Delta / CVD	Pane height, margin, CVD line width	8%, 2.5%, 2px	Layout and autoscale padding of each lower pane
Volume profile	Resolution, split buy/sell	48 rows	Price levels 12–1000; split stacks sell/buy inside each row
Depth	Bucket step, wall threshold	1%, 3×	Row size; multiple of median liquidity that highlights a wall row
Trades	Trade size, per candle, candle share, min volume, merge range, backfill, max trades	top ~1%, 1, 10%, 5, 0.02%, 30 min, 300	Which dots draw: top-% quantile (log slider), merge cap per candle, share of candle volume, absolute floor, merge cell width, history reach, density bound
Liqs	Min size	$10k	Notional floor for markers
Heat	Opacity, gap tolerance, price step, min rest	45%, 6 min, 1%, off	Peak transparency; minutes a level may vanish before its band closes; band grain; minimum observed lifetime
Walls	Opacity, price step, top volume, depth reach, price gap, min cell, max cells/side, min rest	50%, 1%, top 1%, ±0.5%, 0.1%, 25, 5, 2 min	Band grade gates: share of book shown, per-side depth, dead zone around price, absolute cell floor, per-candle cap, lifetime floor
Imbalance	Scope, rarity, absolute floor, opacity, max zones, history depth	global, top 5%, 25%, 30%, 80, 24h	Per-timeframe/symbol override scope; score quantile gate; junk floor; fill strength; zone count cap; how far back zones are built
Prints	Cell filters and lane controls	per panel	Per-price lane filtering and footprint density
Liq zones	Leverage band, window	standard tiers, ≤24h	How OI changes map to estimated liquidation bands
Panes	Pane order	flow → delta → cvd → indicators	Lower-pane stacking order; also via ▲▼ on each pane
Toolbar	Timeframe chips	3+ picked	Quick-access chips; the rest stay under the ▾ menu
Options levels	Source, expiry, levels, zero-gamma	GEX · all · 4 · on	Ranking source (GEX/\|GEX\|/OI/volume), expiry window, top-N strikes, amber flip line
Effort	Rank window, flag pct, min move bps	100, 90, 15	Percentile lookback; effort-rank threshold; absolute bps floor for vacuum flags
Indicators	Per-indicator period(s), width, opacity	see Indicators tab	Each indicator exposes its own parameters, line width and opacity

Datafeed contract

What lands on the server per collection tier, at what cadence, and how it ages out. Panel names follow the toolbar; kind names are the data/API vocabulary.

Top-50: dense tier (by 24h turnover, re-ranked hourly)

Panel	Kind	Writer	Cadence	Hot	Archive	Bound
△ Walls + Depth history	`depth`	live-book + far-REST snapshot and REST round-robin	~2 rows/5-min slot	7d	gzip move	365d
▦ Heat	`xdepth`	xdepthSweep: all ~13 venues merged	every 5-min slot	7d	gzip move	365d
◉ Trades	`prints`	warmSweep all venues, ≥$25k notional + 30-min backfill on tier entry	every 5-min sweep	7d	gzip move	365d
◫ Prints	`fp`	fpAccumulate: every deduped fill → minute×price cells	sealed per minute	7d	gzip move	365d
Volume/Delta/CVD	`flow`	flowStep: Bitget taker + Binance kline split	~once/5-min slot	365d	gzip copy	365d
Price marks	`px`	flowStep kline + warmSweep real-trade marks	~5-min (per-minute inside sweeps)	365d	gzip copy	365d
▼ Liqs	`liqs`	Binance+Bybit all-symbol streams, OKX 5-min poll	live events	7d	gzip move	365d
Zones/Imbalance feed	`oi`	warmSweep all venues + Bitget round-robin	5-min slots	7d	gzip move	365d
n/a	`funding`	warmSweep per venue	every 5-min sweep	7d	gzip move	365d
n/a	`ratios`	warmSweep per venue	every 5-min sweep	7d	gzip move	365d
24h turnover / Filter	`volume`	snapTickers: Bitget+Binance+Bybit	hourly	365d	gzip copy	365d
Γ Options levels	`optlevels`, `optchain`	optSweep: 6 option venues merged, 7 bases only	5-min / hourly + settlement	365d	gzip copy	365d
⚑ Events	`news`, `newsq`, `newsout`	adapters hourly/daily audit	hourly/daily	365d	gzip copy	365d
n/a	`listings`	contract-universe diff	hourly	365d	gzip copy	365d

Others: sweep tier (every other contract, ~750)

Panel	Kind	Writer	Cadence	Hot	Archive	Bound
△ Walls	`depth`	flowStep REST round-robin (Bitget only)	~1 row/5-min slot	7d	gzip move	365d
▦ Heat	`xdepth`	tail rotation: 1/12 of tail per slot, learned venue coverage	~hourly	7d	gzip move	365d
◉ Trades	`prints`	n/a	none today	n/a	n/a	n/a
◫ Prints	`fp`	n/a	none today	n/a	n/a	n/a
Volume/Delta/CVD	`flow`	flowStep: Bitget taker + Binance kline split	~once/5-min slot	365d	gzip copy	365d
Price marks	`px`	flowStep kline marks	~5-min	365d	gzip copy	365d
▼ Liqs	`liqs`	same venue-wide streams, identical coverage	live events	7d	gzip move	365d
Zones/Imbalance feed	`oi`	Bitget round-robin only	~5-min	7d	gzip move	365d
n/a	`funding`, `ratios`	n/a	none	n/a	n/a	n/a
24h turnover / Filter	`volume`	snapTickers	hourly	365d	gzip copy	365d
⚑ Events	`news`, `newsq`, `newsout`	global `_NEWS` bucket	hourly/daily	365d	gzip copy	365d
n/a	`listings`	`_universe` diff	hourly	365d	gzip copy	365d

Deferred: tail-coin Trades/Prints

Trades (prints kind): feasible, parked. One Bitget fills-history call inside the sweep for non-dense symbols, at a ~$100k tail floor. Caveat: the endpoint caps at 100 fills per call; a coin with more than 100 fills per 5 minutes loses the oldest fills in that window (thin coins unaffected).

Prints (fp kind): not honest, skipped. Footprint needs every fill; the 100-fill cap truncates silently, producing cells that understate volume.

Watchlist browser streams

The watchlist is unlimited and client-side only. One shared Bitget WebSocket in your browser carries the trade channel for up to 8 watchlisted symbols (the currently open symbol excluded; its slot frees for the next member). Fills buffer per symbol and replay through the standard ingest path when a symbol opens, so Trades dots and Prints cells arrive warm instead of empty. The server does no per-user collection.

Retention

flow, px, volume, news/newsq/newsout, listings: 365 days hot plus archive. Everything else (depth, xdepth, prints, fp, oi, liqs, walls, funding, ratios): 7 days hot, then gzipped into archive with a 365-day bound. A day-file is deleted only at retention+1 day, so archive always precedes delete.

Disclaimer

Data limits to know

Metrics: what each panel computes

∿ Volume bars, Delta, CVD

≡ Depth

△ Walls

▦ Heat

◉ Trades

◫ Prints (footprint)

▼ Liqs

▨ Liq zones (estimated)

◧ Imbalance zones

⚑ Events

Γ Options levels

◇ Effort

Indicators: definitions and formulas

Overlays on the price pane

SMA [50]

EMA [21, 50, 200]

Bollinger Bands [20, ×2]

VWAP [15 rolling, 0 = session]

Supertrend [10, ×3]

Parabolic SAR [0.02, 0.2]

Ichimoku [9, 26, 52]

S/R levels [3 pivots, 30%]

Swing points

Options levels [GEX · all · 4]

Effort [100 · 90% · 15bp]

Lower panes

RSI [14]

MACD [12, 26, 9]

OBV + MA [30]

Volume MA [5, 10]

ATR [14]

MFI [14]

KDJ [9, 3, 3]

StochRSI [14, 14, 3, 3]

DMI / ADX [14]

CCI [20]

Williams %R [14]

BIAS [20]

Open interest

Filter: the symbol screener

Market filters: last-24h tickers

Candle filters: computed per symbol

Settings reference

Datafeed contract

Top-50: dense tier (by 24h turnover, re-ranked hourly)

Others: sweep tier (every other contract, ~750)

Deferred: tail-coin Trades/Prints

Watchlist browser streams

Retention

Who we are

How it runs

Report a bug or suggest a feature