{ "cells": [ { "cell_type": "markdown", "id": "c723d607", "metadata": {}, "source": [ "# 03 — Libro de Ordenes en Tiempo Real (Binance)\n", "\n", "Dos enfoques para mantener un order book local:\n", "\n", "### Enfoque A: Partial Book Depth (simple)\n", "Stream `@depth@100ms` con levels = 5, 10, 20.\n", "Envia snapshot completo del top N en cada update. Sin logica de sync.\n", "\n", "### Enfoque B: Diff Depth + REST Snapshot (completo)\n", "1. Abrir stream `@depth@100ms` (diffs incrementales)\n", "2. Buffear eventos iniciales\n", "3. Pedir snapshot REST: `GET /api/v3/depth?symbol=X&limit=1000`\n", "4. Descartar eventos con `u <= lastUpdateId` del snapshot\n", "5. Primer evento procesado debe tener `U <= lastUpdateId+1` AND `u >= lastUpdateId+1`\n", "6. Aplicar: qty > 0 = update nivel, qty = 0 = eliminar nivel\n", "7. Validar continuidad: cada evento `U` == anterior `u + 1`, si no, re-sync\n", "\n", "### Campos del depth update\n", "```json\n", "{\n", " \"U\": 157, \"u\": 160,\n", " \"b\": [[\"price\", \"qty\"], ...], \n", " \"a\": [[\"price\", \"qty\"], ...] \n", "}\n", "```" ] }, { "cell_type": "code", "execution_count": null, "id": "23b19294", "metadata": {}, "outputs": [], "source": [ "import asyncio\n", "import json\n", "import requests\n", "import websockets\n", "from decimal import Decimal\n", "from collections import deque\n", "import pandas as pd\n", "\n", "BASE = \"https://api.binance.com\"\n", "WS_BASE = \"wss://stream.binance.com:9443/ws\"" ] }, { "cell_type": "markdown", "id": "cff459c9", "metadata": {}, "source": [ "## Enfoque A: Partial Book Depth (simple, sin sync)\n", "\n", "Stream `@depth@100ms` — recibe snapshot completo del top N cada 100ms.\n", "\n", "Ideal para monitoreo rapido sin necesidad de mantener estado." ] }, { "cell_type": "code", "execution_count": null, "id": "bf70712b", "metadata": {}, "outputs": [], "source": [ "async def stream_top_book(symbol: str, levels: int = 10, snapshots: int = 50) -> list[dict]:\n", " \"\"\"Captura N snapshots del top del order book.\"\"\"\n", " url = f\"{WS_BASE}/{symbol.lower()}@depth{levels}@100ms\"\n", " results = []\n", "\n", " async with websockets.connect(url) as ws:\n", " while len(results) < snapshots:\n", " data = json.loads(await ws.recv())\n", " best_bid = (float(data[\"bids\"][0][0]), float(data[\"bids\"][0][1]))\n", " best_ask = (float(data[\"asks\"][0][0]), float(data[\"asks\"][0][1]))\n", " spread = best_ask[0] - best_bid[0]\n", " mid = (best_bid[0] + best_ask[0]) / 2\n", " results.append({\n", " \"time\": pd.Timestamp.now(tz=\"UTC\"),\n", " \"best_bid\": best_bid[0], \"bid_qty\": best_bid[1],\n", " \"best_ask\": best_ask[0], \"ask_qty\": best_ask[1],\n", " \"spread\": spread, \"spread_bps\": (spread / mid) * 10000,\n", " \"bids\": [(float(p), float(q)) for p, q in data[\"bids\"]],\n", " \"asks\": [(float(p), float(q)) for p, q in data[\"asks\"]],\n", " })\n", "\n", " return results" ] }, { "cell_type": "markdown", "id": "99c8d974", "source": "### Ejemplo: Capturar 50 snapshots del top-10 del book de BTCUSDT", "metadata": {} }, { "cell_type": "code", "id": "855fc378", "source": "snapshots = await stream_top_book(\"BTCUSDT\", levels=10, snapshots=50)\ndf_book = pd.DataFrame([{k: v for k, v in s.items() if k not in (\"bids\", \"asks\")} for s in snapshots])\nprint(f\"Capturados {len(df_book)} snapshots del order book\")\nprint(f\"Spread medio: {df_book['spread'].mean():.2f} USD ({df_book['spread_bps'].mean():.2f} bps)\")\nprint(f\"Spread min: {df_book['spread'].min():.2f}, max: {df_book['spread'].max():.2f}\")\ndf_book.head(10)", "metadata": {}, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "id": "271ea442", "source": "## Enfoque B: Order Book completo con diff depth + REST sync\n\nMantiene un order book local completo sincronizado via WebSocket diffs.\nMas complejo pero da acceso a todos los niveles de precio.", "metadata": {} }, { "cell_type": "code", "id": "79701eff", "source": "class LocalOrderBook:\n \"\"\"Order book local sincronizado con Binance via WebSocket diffs.\"\"\"\n\n def __init__(self, symbol: str, depth_limit: int = 1000):\n self.symbol = symbol.upper()\n self.depth_limit = depth_limit\n self.bids: dict[Decimal, Decimal] = {} # price -> qty\n self.asks: dict[Decimal, Decimal] = {}\n self.last_update_id: int = 0\n self._synced = False\n\n def _get_snapshot(self) -> dict:\n resp = requests.get(f\"{BASE}/api/v3/depth\",\n params={\"symbol\": self.symbol, \"limit\": self.depth_limit})\n resp.raise_for_status()\n return resp.json()\n\n def _apply_update(self, sides: list[tuple[str, dict]]):\n for price_str, qty_str in sides:\n price, qty = Decimal(price_str), Decimal(qty_str)\n return price, qty\n\n def apply_event(self, event: dict):\n for price_str, qty_str in event.get(\"b\", []):\n p, q = Decimal(price_str), Decimal(qty_str)\n if q == 0:\n self.bids.pop(p, None)\n else:\n self.bids[p] = q\n for price_str, qty_str in event.get(\"a\", []):\n p, q = Decimal(price_str), Decimal(qty_str)\n if q == 0:\n self.asks.pop(p, None)\n else:\n self.asks[p] = q\n self.last_update_id = event[\"u\"]\n\n def best_bid(self) -> tuple[float, float] | None:\n if not self.bids:\n return None\n p = max(self.bids)\n return (float(p), float(self.bids[p]))\n\n def best_ask(self) -> tuple[float, float] | None:\n if not self.asks:\n return None\n p = min(self.asks)\n return (float(p), float(self.asks[p]))\n\n def spread(self) -> float | None:\n b, a = self.best_bid(), self.best_ask()\n return a[0] - b[0] if b and a else None\n\n def top_n(self, n: int = 5) -> dict:\n return {\n \"bids\": [(float(p), float(q)) for p, q in sorted(self.bids.items(), reverse=True)[:n]],\n \"asks\": [(float(p), float(q)) for p, q in sorted(self.asks.items())[:n]],\n }\n\n async def run(self, duration_seconds: int = 30):\n \"\"\"Sincroniza y mantiene el book durante N segundos.\"\"\"\n import time as _time\n url = f\"{WS_BASE}/{self.symbol.lower()}@depth@100ms\"\n buffer = []\n start = _time.time()\n\n async with websockets.connect(url) as ws:\n while _time.time() - start < duration_seconds:\n msg = json.loads(await asyncio.wait_for(ws.recv(), timeout=5))\n\n if not self._synced:\n buffer.append(msg)\n if len(buffer) >= 3:\n snap = self._get_snapshot()\n # Inicializar desde snapshot\n self.bids = {Decimal(p): Decimal(q) for p, q in snap[\"bids\"]}\n self.asks = {Decimal(p): Decimal(q) for p, q in snap[\"asks\"]}\n self.last_update_id = snap[\"lastUpdateId\"]\n\n # Procesar buffer\n for evt in buffer:\n if evt[\"u\"] <= self.last_update_id:\n continue\n if not self._synced and evt[\"U\"] <= self.last_update_id + 1:\n self._synced = True\n if self._synced:\n self.apply_event(evt)\n buffer.clear()\n if self._synced:\n print(f\"Book sincronizado! Levels: {len(self.bids)} bids, {len(self.asks)} asks\")\n else:\n if msg[\"U\"] != self.last_update_id + 1:\n print(\"Gap detectado, re-sync...\")\n self._synced = False\n buffer = [msg]\n continue\n self.apply_event(msg)", "metadata": {}, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "id": "7370d164", "source": "### Ejemplo: Correr el order book 30 segundos y ver el estado", "metadata": {} }, { "cell_type": "code", "id": "ae407462", "source": "book = LocalOrderBook(\"BTCUSDT\")\nawait book.run(duration_seconds=30)\n\nprint(f\"\\nBest bid: {book.best_bid()}\")\nprint(f\"Best ask: {book.best_ask()}\")\nprint(f\"Spread: {book.spread():.2f} USD\")\nprint(f\"\\nTop 5:\")\ntop = book.top_n(5)\nprint(\"BIDS:\", top[\"bids\"])\nprint(\"ASKS:\", top[\"asks\"])", "metadata": {}, "execution_count": null, "outputs": [] }, { "cell_type": "markdown", "id": "c11ab0fb", "source": "## Snapshot REST directo (alternativa simple)\n\nPara consultas puntuales sin mantener estado, un GET directo al depth endpoint.", "metadata": {} }, { "cell_type": "code", "id": "9a31a0ba", "source": "def get_depth_snapshot(symbol: str, limit: int = 20) -> dict:\n \"\"\"Snapshot directo del order book via REST.\"\"\"\n resp = requests.get(f\"{BASE}/api/v3/depth\", params={\"symbol\": symbol, \"limit\": limit})\n resp.raise_for_status()\n data = resp.json()\n return {\n \"bids\": [(float(p), float(q)) for p, q in data[\"bids\"]],\n \"asks\": [(float(p), float(q)) for p, q in data[\"asks\"]],\n \"last_update_id\": data[\"lastUpdateId\"],\n }\n\nsnap = get_depth_snapshot(\"BTCUSDT\", limit=10)\nprint(f\"Top 10 bids y asks (update_id={snap['last_update_id']}):\")\nprint(\"\\nBIDS (compra):\")\nfor p, q in snap[\"bids\"]:\n print(f\" {p:>12.2f} | {q:.6f}\")\nprint(\"\\nASKS (venta):\")\nfor p, q in snap[\"asks\"]:\n print(f\" {p:>12.2f} | {q:.6f}\")", "metadata": {}, "execution_count": null, "outputs": [] } ], "metadata": { "kernelspec": { "display_name": "Python 3 (ipykernel)", "language": "python", "name": "python3" }, "language_info": { "codemirror_mode": { "name": "ipython", "version": 3 }, "file_extension": ".py", "mimetype": "text/x-python", "name": "python", "nbconvert_exporter": "python", "pygments_lexer": "ipython3", "version": "3.13.7" } }, "nbformat": 4, "nbformat_minor": 5 }