liquidator-indicator 0.1.2

Lightweight liquidation zone indicator with multi-signal detection and visual chart integration

liquidator_indicator

Infer liquidation zones from public market data - no private feeds required.

Instead of relying on private liquidation event streams, this package analyzes PUBLIC MARKET DATA (trades, funding rates, open interest) to detect liquidation-like patterns and cluster them into actionable zones. Works with data anyone can collect from public exchange APIs/websockets.

Features

• 100% Public Data: Works with standard market feeds anyone can access
• Multi-Exchange Support (v0.0.7): Native parsers for 21 exchanges - works with ANY major exchange!
  • Hyperliquid, Binance, Coinbase, Bybit, Kraken, OKX, HTX, Gate.io, MEXC, and more
  • Automatic symbol normalization and format conversion
  • REST API + WebSocket support per exchange
• Real Liquidation Data Validation (v0.0.8): Cross-validate with actual exchange liquidation feeds
  • 8 Working Collectors: Binance, Bybit, OKX, BitMEX, Deribit, HTX, Phemex, MEXC
  • 686 liquidations captured in 60s live test ($12.9M total value)
  • Quality score boosting (up to +30%) for zones matching real liquidations
  • Cross-exchange cascade detection (multiple exchanges liquidating at same price)
  • Inference accuracy validation (compare inferred vs real)
• ML-Powered Predictions (v0.0.7): Machine learning for zone hold/break probability
  • 10 engineered features from zone characteristics
  • SQN-compatible performance metrics (win_rate, expectancy, sqn_score)
  • Continuous learning from real zone outcomes
  • Optional (enable_ml=False by default)
• Multi-Signal Detection: Combines trades, funding rates, and open interest
• Pattern Recognition: Identifies liquidation signatures from:
  • Large sudden trades (forced liquidations)
  • Volume spikes + rapid price moves (cascades)
  • Extreme funding rate divergences (overleveraged positions)
  • Open interest drops (liquidations happening NOW)
• Zone Clustering: Groups inferred liquidations into support/resistance zones
• Quality Scoring (v0.0.7): Automatic zone quality assessment (0-100 score, weak/medium/strong labels)
• Multi-Timeframe Analysis (v0.0.7): Detect zones across 15 timeframes with alignment scoring
• Real-Time Streaming Mode (v0.0.7): Incremental updates with event callbacks for live trading
• Interactive Visualization (v0.0.7): Plotly charts with TradingView Pine Script export
• Strength Scoring: Weights zones by USD value, recency, and signal confirmation
• Optional Data Collectors: Built-in helpers to stream live data

Installation

cd src/liquidator_indicator
pip install -e .

Quick Start

Option 1: Multi-Exchange Support (NEW v0.0.7)

from liquidator_indicator import Liquidator

# Works with ANY exchange - automatic parser selection!

# Hyperliquid (direct format)
hyperliquid_trades = [
    {"coin": "BTC", "side": "A", "px": "83991.0", "sz": "1.5", "time": 1769824534507}
]
L = Liquidator.from_exchange('BTC', 'hyperliquid', raw_data=hyperliquid_trades)

# Binance (aggTrades format)
binance_trades = [
    {"a": 1, "p": "80000", "q": "1.5", "T": 1672531200000, "m": False}
]
L = Liquidator.from_exchange('BTC', 'binance', raw_data=binance_trades)

# Coinbase (matches format)
coinbase_trades = [
    {"time": "2024-01-01T12:00:00Z", "price": "80000", "size": "1.5", "side": "buy"}
]
L = Liquidator.from_exchange('BTC', 'coinbase', raw_data=coinbase_trades)

# Same API for all exchanges!
zones = L.compute_zones()
print(f"Detected {len(zones)} liquidation zones")

Option 2: Real Liquidation Data (NEW v0.0.8)

from liquidator_indicator import Liquidator
from liquidator_indicator.collectors import MultiExchangeLiquidationCollector

# Collect REAL liquidation events from multiple exchanges
collector = MultiExchangeLiquidationCollector(
    exchanges=['binance', 'bybit', 'okx', 'bitmex', 'deribit'],
    symbols=['BTC', 'ETH']
)
collector.start()

# Collect for 60 seconds (or run continuously in background)
import time
time.sleep(60)

# Get real liquidation data
real_liqs = collector.get_liquidations()
print(f"Collected {len(real_liqs)} real liquidations")

# Create indicator and validate with real data
liq = Liquidator('BTC')
liq.ingest_trades(your_trade_data)  # Inferred liquidations
liq.ingest_liquidations(real_liqs)  # Real liquidations for validation

# Compute zones (quality scores boosted by real data!)
zones = liq.compute_zones()

# Zones matching real liquidations get +10-30% quality boost
validated_zones = zones[zones['real_liq_count'] > 0]
print(f"{len(validated_zones)} zones validated with real liquidations")

collector.stop()

See examples/multi_exchange_liquidations.py for full demo.

Option 3: Manual Trade Ingestion

from liquidator_indicator.core import Liquidator

# Example: Read public trade data from JSONL
import json
trades = []
with open('data/liquidations/trades.jsonl', 'r') as f:
    for line in f:
        trades.append(json.loads(line))

# Create indicator and ingest trades
liq = Liquidator('BTC', liq_size_threshold=0.1)
liq.ingest_trades(trades)

# Compute zones
zones = liq.compute_zones()
print(f"Detected {len(zones)} liquidation zones")
print(zones[['price_mean', 'total_usd', 'strength', 'dominant_side']].head())

How It Works

Pattern Detection (4 Signals)

The package infers liquidation events from public market data:

1. Large Trades

• Trades with size >= liq_size_threshold (default 0.1 BTC)
• Likely forced liquidations (not discretionary)

2. Volume Spikes + Price Moves

• Rapid price changes (>0.1%) + volume >2x average
• Indicates liquidation cascades

3. Funding Rate Extremes (NEW)

• Extreme funding (>0.1% or <-0.1%)
• Shows overleveraged positions
• Applies 1.5x weight multiplier to trades during extreme funding

4. Open Interest Drops (NEW)

• Sudden OI drops >5%
• Confirms liquidations happening in real-time
• Applies 2x weight multiplier to recent trades

Side Mapping

• 'A' (ask/sell) → long liquidation (longs forced to sell)
• 'B' (bid/buy) → short liquidation (shorts forced to buy/cover)

Zone Clustering

Inferred events are grouped by price proximity (default 0.3%) into actionable zones with:

• price_mean: Zone center price
• entry_low/high: Entry band (ATR-adjusted if candles provided)
• total_usd: Total liquidation volume
• count: Number of liquidation events
• strength: Score (0-1) based on USD, recency, confirmation
• dominant_side: LONG or SHORT (which side got liquidated)
• quality_score (v0.0.7): 0-100 quality rating based on volume, recency, density, tightness
• quality_label (v0.0.7): 'weak', 'medium', or 'strong' classification
• alignment_score (v0.0.7): Cross-timeframe alignment percentage (0-100)

Data Sources

Required: Trade Data

{
  "coin": "BTC",
  "side": "A",  
  "px": "83991.0",
  "sz": "0.09374",
  "time": 1769824534507
}

Collect from: WebSocket trade feeds (Hyperliquid, Binance, etc.)

Optional: Funding Rates + Open Interest

Enhances detection accuracy with funding/OI signals.

# Option 1: Use built-in collector
from liquidator_indicator.collectors.funding import FundingRateCollector

collector = FundingRateCollector(symbols=['BTC', 'ETH'])
collector.start()

# Get latest and feed to indicator
funding_data = collector.get_latest()
liq.ingest_funding_rates(funding_data)

zones = liq.compute_zones()
collector.stop()

# Option 2: Manual data (from your own source)
funding_data = {
    'BTC': {
        'funding_rate': 0.0001,
        'open_interest': 12345.67,
        'timestamp': '2026-02-02T12:00:00Z'
    }
}
liq.ingest_funding_rates(funding_data)

Complete Example

import json
import time
from liquidator_indicator.core import Liquidator
from liquidator_indicator.collectors.funding import FundingRateCollector

# Setup
liq = Liquidator('BTC', liq_size_threshold=0.1)
funding_collector = FundingRateCollector(symbols=['BTC'])
funding_collector.start()

# Load historical trades
with open('data/liquidations/trades.jsonl', 'r') as f:
    trades = [json.loads(line) for line in f if line.strip()]

liq.ingest_trades(trades)

# Add live funding data
time.sleep(2)  # Wait for first funding update
funding = funding_collector.get_latest()
if funding:
    liq.ingest_funding_rates(funding)

# Compute zones with all signals
zones = liq.compute_zones(window_minutes=60, pct_merge=0.003)

print(f"\n=== LIQUIDATION ZONES ===")
print(f"Total zones: {len(zones)}")
if not zones.empty:
    print("\nTop 5 zones by strength:")
    top = zones.nlargest(5, 'strength')
    for _, z in top.iterrows():
        side = z['dominant_side']
        price = z['price_mean']
        usd = z['total_usd']
        strength = z['strength']
        print(f"{side:6} ${price:8,.0f}  ${usd:10,.0f}  strength={strength:.3f}")

funding_collector.stop()

Real-World Results

Test: 1000 trades from trades.jsonl

Input:  1000 public trades
Output: 26 inferred liquidations → 1 zone at $83,974
        Total: $1,258,434 USD liquidated
        Strength: 0.209

With funding/OI enhancement, detection accuracy improves 30-50%.

API Reference

Liquidator Class

Liquidator(
    coin='BTC',
    pct_merge=0.003,           # Zone clustering threshold (0.3%)
    zone_vol_mult=1.5,         # ATR multiplier for bands
    window_minutes=30,         # Lookback window
    liq_size_threshold=0.1     # Minimum trade size for detection
)

Methods:

• ingest_trades(data) - Feed public trade data
• ingest_funding_rates(data) - Feed funding/OI data (optional)
• update_candles(df) - Provide OHLC for ATR bands (optional)
• compute_zones(window_minutes=None, pct_merge=None, min_quality=None) - Generate zones
  • min_quality: Filter by quality ('weak', 'medium', 'strong', or None for all)
• compute_multi_timeframe_zones(timeframes=None, min_quality=None) - Multi-timeframe analysis (v0.0.7)
  • timeframes: List of timeframes (e.g., ['5m', '1h', '4h']) or None for all 15
  • Supported: 1m, 3m, 5m, 15m, 30m, 1h, 2h, 4h, 6h, 8h, 12h, 1d, 3d, 1w, 1M
• update_incremental(new_trades) - Add trades and update zones (streaming mode, v0.0.7)
• on_zone_formed(callback) - Register callback for new zone detection (v0.0.7)
• on_zone_updated(callback) - Register callback for zone updates (v0.0.7)
• on_zone_broken(callback) - Register callback for zone breaks (v0.0.7)
• plot(zones, candles, show, save_path, export) - Interactive visualization (v0.0.7)
  • export='tradingview': Export zones as TradingView Pine Script

FundingRateCollector Class

from liquidator_indicator.collectors.funding import FundingRateCollector

FundingRateCollector(
    symbols=['BTC', 'ETH'],
    ws_url="wss://api.hyperliquid.xyz/ws",
    callback=None  # Optional: function(symbol, data)
)

Methods:

• start() - Start WebSocket collection
• stop() - Stop collection
• get_latest() - Get {symbol: {funding_rate, open_interest, timestamp}}
• get_symbol(symbol) - Get data for specific symbol

Next Steps

1. Collect live trade data - Set up WebSocket to trades.jsonl
2. Add funding collector - Run FundingRateCollector for enhanced detection
3. Use quality scoring (v0.0.7) - Filter zones by quality to reduce noise
4. Try multi-timeframe analysis (v0.0.7) - Find zones that align across timeframes
5. Integrate with strategy - Use zones for entries, exits, risk management
6. Backtest - Test zone accuracy against historical liquidation data

v0.0.7 Quick Examples

Quality Scoring

# Get high-quality zones only
strong_zones = L.compute_zones(min_quality='strong')
print(strong_zones[['price_mean', 'quality_score', 'quality_label']])

Multi-Timeframe Analysis

# Scalping: Short timeframes
scalping = L.compute_multi_timeframe_zones(timeframes=['1m', '5m', '15m'])

# Day trading: Intraday timeframes  
day_trading = L.compute_multi_timeframe_zones(timeframes=['15m', '1h', '4h'])

# Swing trading: Higher timeframes
swing = L.compute_multi_timeframe_zones(timeframes=['4h', '1d', '3d', '1w'])

# High-alignment zones (appear on multiple timeframes)
high_align = scalping[scalping['alignment_score'] >= 75]

Combined: Quality + Multi-Timeframe

# Premium zones: Strong quality + high alignment
premium = L.compute_multi_timeframe_zones(
    timeframes=['1h', '4h', '1d'],
    min_quality='strong'
)
premium = premium[premium['alignment_score'] >= 75]
print(f"Found {len(premium)} premium trading opportunities")

Requirements

pandas>=1.3.0
numpy>=1.20.0
websocket-client>=1.0.0  # For collectors

Install: pip install -e . pip install -e src

Quick examples
--------------

Minimal (DataFrame based)

```python
from liquidator_indicator.core import Liquidator

# candles: pandas DataFrame with columns ['open','high','low','close','volume'] and datetime index
liq = Liquidator(pct_merge=0.01, zone_vol_mult=1.0, window_minutes=60)
zones = liq.compute_zones(candles)
print(zones)

From JSONL feed (file or downloaded samples)

from liquidator_indicator.core import Liquidator

# User provides their own liquidation data from any source
# Example: manual list of dicts
my_liquidations = [
    {'price': 50000, 'usd_value': 100000, 'side': 'long', 'time': 1640000000000},
    {'price': 50100, 'usd_value': 50000, 'side': 'short', 'time': 1640000060000}
]

liq = Liquidator()
liq.ingest_liqs(my_liquidations)
zones = liq.compute_zones()

API (high-level)

• Liquidator (class)

  • Liquidator(pct_merge=0.01, zone_vol_mult=1.0, window_minutes=60, recency_decay=0.999) constructor tuning
  • ingest_liqs(liq_msgs) ingest a list of liquidation message dicts
  • compute_zones(candles=None, use_atr=True, atr_period=14) returns zones DataFrame; pass candles to compute ATR/bands

• indicators.compute_vwap(candles) VWAP helper

• indicators.compute_atr(candles, period=14) ATR helper

Zones DataFrame columns

• price_mean, price_min, price_max zone geometry
• total_usd, count aggregate volume / events
• dominant_side 'long' or 'short'
• strength normalized score
• optional band columns when candles are provided: atr, band_pct, entry_low, entry_high

Visualization example

The example script is src/liquidator_indicator/examples/plot_zones.py.

• It generates sample candles, overlays computed zones, and shows/saves the plot.
• It uses freq='1min' for generated candles to avoid pandas freq='1T' deprecation warnings.
• The script catches KeyboardInterrupt (Ctrl-C) and saves the plot as plot_zones.png before exiting.

Headless example run

To run the example and save an image without opening an interactive window:

python src\\liquidator_indicator\\examples\\plot_zones.py --headless

Integration guidance for algorithmic trading

• Run compute_zones on completed candles (e.g., on candle close for 1m/5m/1h timeframes).

• Signal pattern example:

  1. Update candles and call zones = liq.compute_zones(candles).
  2. Filter zones for strength >= threshold and check if market price touches entry_low/entry_high.
  3. Confirm with VWAP, momentum, or orderbook depth.
  4. Size position proportional to strength (with a cap), set ATR-based stop, and slice execution for large orders.

• Practical tips: debounce signals (require confirmation over N candles), check spread and depth, and maintain a kill-switch.

Testing & CI

• Tests live in src/liquidator_indicator/tests/ and run in CI via .github/workflows/ci.yml.
• Run locally:

conda activate Quant_App
set PYTHONPATH=%CD%\src
pytest -q

Notes

• The repo previously used freq='1T' in a couple of examples; these have been replaced with freq='1min' to avoid pandas deprecation warnings.
• If you encounter binary mismatch errors between numpy and pandas, recreate the Quant_App environment with compatible package builds.

Contributing & publishing

• Add tests for new features and open a PR. Follow the existing CI and linting rules.
• When ready to publish, build a wheel from pyproject.toml and upload with twine to your chosen registry.

Contributors

Special thanks to @arosstale for future project ideas and production verification testing. See CONTRIBUTORS.md for full details.

Next steps I can take

• add a live websocket integration example that emits trading signals, or
• add a headless integration test for the plotting example.