chemavx
63d9f637ff
CI/CD / build-and-push (push) Successful in 2m30s
Phase 1 — Edge neto real (paper.py, bayesian.py, risk/manager.py, db.py):
- Trade records now store edge_gross, edge_net, prior_prob, final_prob,
mid_price, spread_estimate, commission, family_key
- edge_net = edge_gross - SPREAD_ESTIMATE(0.02) - COMMISSION_RATE(0.02)
NOTE: both constants are heuristics, not exact Polymarket exchange costs
- Execution gate changed from edge_gross > MIN_EDGE to edge_net > regime_min_edge
Phase 2 — Market families (polymarket.py):
- market_family_key(market) groups related markets:
texas-republican-2026, fed-april-2026, openai-2026, etc.
- At most 1 trade per family per cycle; occupied_families propagated via main.py
- Family key logged on every TRADE and SKIP line
Phase 3 — GNews priority (news.py, bayesian.py, main.py):
- NewsClient.get_freshness() returns 1.0/0.75/0.40/0.10 by cache age
- gnews_priority(market, news) = uncertainty × volume_score × freshness
- Politics markets sorted by priority DESC before eval so best markets get
the 5-query/cycle GNews budget first
Phase 4 — Regime min-edge by category/horizon (bayesian.py):
- politics >60d → 0.12, 30-60d → 0.10, <30d → 0.08
- tech / crypto/finance → 0.10
- All thresholds applied to edge_net (not edge_gross)
Phase 5 — Observability (bayesian.py, main.py):
- Structured skip labels: SKIP_UNSUPPORTED, SKIP_NO_SIGNALS,
SKIP_PRIOR_EXTREME, SKIP_FAMILY, SKIP_GNEWS_PRIORITY, SKIP_EDGE_NET
- TRADE lines now include family_key, edge_gross, edge_net, regime_min, days
- schema.sql: 8 new cols on trades, 7 new cols on signals (via ALTER TABLE IF NOT EXISTS)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
452 lines
18 KiB
Python
452 lines
18 KiB
Python
"""
|
|
Polymarket CLOB API client.
|
|
Docs: https://docs.polymarket.com
|
|
"""
|
|
import asyncio
|
|
import logging
|
|
import os
|
|
import re
|
|
from dataclasses import dataclass, field
|
|
from datetime import datetime, timezone, timedelta
|
|
from typing import Optional
|
|
import httpx
|
|
|
|
log = logging.getLogger(__name__)
|
|
|
|
POLYMARKET_API = "https://clob.polymarket.com"
|
|
GAMMA_API = "https://gamma-api.polymarket.com"
|
|
|
|
# ─────────────────────────────────────────────────────────────────────────────
|
|
# Phase 2 — Market family classification helpers
|
|
# Used by market_family_key() below.
|
|
# ─────────────────────────────────────────────────────────────────────────────
|
|
|
|
_YEAR_RE = re.compile(r"\b(202\d|203\d)\b")
|
|
_MONTH_RE = re.compile(
|
|
r"\b(january|february|march|april|may|june|july|august|"
|
|
r"september|october|november|december)\b",
|
|
re.IGNORECASE,
|
|
)
|
|
_FED_TRIGGER_RE = re.compile(
|
|
r"\b(federal reserve|interest rate|bps|basis point|fed\s+(rate|meeting|decision))",
|
|
re.IGNORECASE,
|
|
)
|
|
_US_STATE_RE = re.compile(
|
|
r"\b(Alabama|Alaska|Arizona|Arkansas|California|Colorado|Connecticut|"
|
|
r"Delaware|Florida|Georgia|Hawaii|Idaho|Illinois|Indiana|Iowa|Kansas|"
|
|
r"Kentucky|Louisiana|Maine|Maryland|Massachusetts|Michigan|Minnesota|"
|
|
r"Mississippi|Missouri|Montana|Nebraska|Nevada|New\s+Hampshire|"
|
|
r"New\s+Jersey|New\s+Mexico|New\s+York|North\s+Carolina|North\s+Dakota|"
|
|
r"Ohio|Oklahoma|Oregon|Pennsylvania|Rhode\s+Island|South\s+Carolina|"
|
|
r"South\s+Dakota|Tennessee|Texas|Utah|Vermont|Virginia|Washington|"
|
|
r"West\s+Virginia|Wisconsin|Wyoming)\b",
|
|
re.IGNORECASE,
|
|
)
|
|
_PARTY_RE = re.compile(r"\b(Republican|Democrats?|Democratic|GOP)\b", re.IGNORECASE)
|
|
_ELECTION_TYPE_RE = re.compile(
|
|
r"\b(presidential|president|mayoral|mayor|gubernatorial|governor|"
|
|
r"senate|congress(?:ional)?|primary|election)\b",
|
|
re.IGNORECASE,
|
|
)
|
|
|
|
# Ordered list of (pattern, place_slug) for named non-US locations.
|
|
# Checked after US-state patterns so US city/state names don't shadow these.
|
|
_NAMED_PLACES: list[tuple[re.Pattern, str]] = [
|
|
(re.compile(r"\bColomb", re.IGNORECASE), "colombia"),
|
|
(re.compile(r"\bSeoul\b", re.IGNORECASE), "seoul"),
|
|
(re.compile(r"\bBusan\b", re.IGNORECASE), "busan"),
|
|
(re.compile(r"\bGyeonggi\b", re.IGNORECASE), "gyeonggi"),
|
|
(re.compile(r"\bChungcheong", re.IGNORECASE), "chungcheong"),
|
|
(re.compile(r"\bSouth\s+Korean?\b", re.IGNORECASE), "south-korea"),
|
|
(re.compile(r"\bLos\s+Angeles\b", re.IGNORECASE), "los-angeles"),
|
|
(re.compile(r"\bCuba\b", re.IGNORECASE), "cuba"),
|
|
(re.compile(r"\bLebanon\b", re.IGNORECASE), "lebanon"),
|
|
(re.compile(r"\bIsrael\b", re.IGNORECASE), "israel"),
|
|
(re.compile(r"\bUkraine\b", re.IGNORECASE), "ukraine"),
|
|
(re.compile(r"\bRussia\b", re.IGNORECASE), "russia"),
|
|
]
|
|
|
|
# Ordered list of (pattern, company_slug) for tech/company markets.
|
|
_NAMED_COMPANIES: list[tuple[re.Pattern, str]] = [
|
|
(re.compile(r"\bopenai\b", re.IGNORECASE), "openai"),
|
|
(re.compile(r"\banthropic\b", re.IGNORECASE), "anthropic"),
|
|
(re.compile(r"\bnvidia\b", re.IGNORECASE), "nvidia"),
|
|
(re.compile(r"\bapple\b", re.IGNORECASE), "apple"),
|
|
(re.compile(r"\bmicrosoft\b", re.IGNORECASE), "microsoft"),
|
|
(re.compile(r"\bgoogle\b", re.IGNORECASE), "google"),
|
|
(re.compile(r"\btesla\b", re.IGNORECASE), "tesla"),
|
|
# \bmeta\b does NOT match MetaMask (no word boundary mid-compound-word)
|
|
(re.compile(r"\bmeta\b", re.IGNORECASE), "meta"),
|
|
]
|
|
|
|
|
|
def _end_month(market: "Market") -> str:
|
|
"""Return market end_date formatted as YYYY-MM, or '' if unparseable."""
|
|
raw = market.end_date
|
|
if not raw:
|
|
return ""
|
|
try:
|
|
dt = datetime.fromisoformat(raw.replace("Z", "+00:00"))
|
|
return dt.strftime("%Y-%m")
|
|
except (ValueError, TypeError):
|
|
return ""
|
|
|
|
|
|
def market_family_key(market: "Market") -> str:
|
|
"""
|
|
Return a stable slug that groups related markets together.
|
|
|
|
Markets in the same family share an underlying event (same election,
|
|
same Fed meeting decision, same company). The bot allows at most one
|
|
open position per family per cycle to avoid correlated exposure.
|
|
|
|
Priority order (first match wins):
|
|
1. Fed / interest-rate decision → fed-{month}-{year}
|
|
2. US state + party election → {state}-{party}-{year}
|
|
3. Named non-US city/country → {place}-{event_type}-{year}
|
|
4. Named tech company → {company}-{year}
|
|
5. Fallback → {category}-{end_YYYY-MM}
|
|
|
|
Examples:
|
|
"Will Ken Paxton win the 2026 Texas Republican Primary"
|
|
→ texas-republican-2026
|
|
"Will the Fed decrease rates by 25 bps after April 2026 meeting"
|
|
→ fed-april-2026
|
|
"Will OpenAI IPO by December 31 2026?"
|
|
→ openai-2026
|
|
"""
|
|
q = market.question
|
|
|
|
# Prefer year from question text; fall back to end_date year if absent
|
|
year_m = _YEAR_RE.search(q)
|
|
if year_m:
|
|
year = year_m.group(1)
|
|
else:
|
|
end_m = _end_month(market) # e.g. "2026-06"
|
|
year = end_m[:4] if end_m else "unknown"
|
|
|
|
# 1. Fed / interest-rate meeting
|
|
if _FED_TRIGGER_RE.search(q):
|
|
month_m = _MONTH_RE.search(q)
|
|
if month_m:
|
|
return f"fed-{month_m.group(1).lower()}-{year}"
|
|
return f"fed-{year}"
|
|
|
|
# 2. US state + party (primary, senate, governor, etc.)
|
|
state_m = _US_STATE_RE.search(q)
|
|
party_m = _PARTY_RE.search(q)
|
|
if state_m and party_m:
|
|
state = re.sub(r"\s+", "-", state_m.group(1).lower())
|
|
raw_party = party_m.group(1).lower()
|
|
# "democrat" prefix covers "democrat", "democrats", "democratic"
|
|
party = "democrat" if "democrat" in raw_party else "republican"
|
|
return f"{state}-{party}-{year}"
|
|
|
|
# 3. Named non-US city / country
|
|
for place_re, place_slug in _NAMED_PLACES:
|
|
if place_re.search(q):
|
|
etype_m = _ELECTION_TYPE_RE.search(q)
|
|
if etype_m:
|
|
raw_etype = etype_m.group(1).lower()
|
|
# Normalise synonyms
|
|
etype = {
|
|
"president": "presidential",
|
|
"mayor": "mayoral",
|
|
"governor": "gubernatorial",
|
|
}.get(raw_etype, raw_etype)
|
|
else:
|
|
etype = "event"
|
|
return f"{place_slug}-{etype}-{year}"
|
|
|
|
# 4. Named tech company
|
|
for company_re, company_slug in _NAMED_COMPANIES:
|
|
if company_re.search(q):
|
|
return f"{company_slug}-{year}"
|
|
|
|
# 5. Fallback: category + end_date month
|
|
end_month = _end_month(market)
|
|
base = market.category if market.category else "misc"
|
|
return f"{base}-{end_month}" if end_month else f"{base}-{year}"
|
|
|
|
|
|
@dataclass
|
|
class Market:
|
|
id: str
|
|
condition_id: str
|
|
question: str
|
|
yes_token_id: str
|
|
no_token_id: str
|
|
yes_price: float # 0-1, current best ask for YES
|
|
no_price: float
|
|
volume_24h: float
|
|
end_date: str
|
|
active: bool
|
|
category: str = ""
|
|
|
|
|
|
@dataclass
|
|
class OrderBook:
|
|
market_id: str
|
|
yes_bids: list[tuple[float, float]] = field(default_factory=list) # (price, size)
|
|
yes_asks: list[tuple[float, float]] = field(default_factory=list)
|
|
mid_price: float = 0.5
|
|
|
|
|
|
class PolymarketClient:
|
|
"""
|
|
Async Polymarket client.
|
|
In paper mode, API key is not needed — only public data.
|
|
API key required for placing real orders.
|
|
"""
|
|
|
|
def __init__(self) -> None:
|
|
self.api_key = os.getenv("POLYMARKET_API_KEY", "")
|
|
self.secret = os.getenv("POLYMARKET_SECRET", "")
|
|
self.passphrase = os.getenv("POLYMARKET_PASSPHRASE", "")
|
|
self._client = httpx.AsyncClient(timeout=30)
|
|
|
|
# Keywords that identify crypto / finance markets.
|
|
# Short tickers are padded with spaces to avoid false substring matches
|
|
# (e.g. " eth " won't match "Hegseth"; " sol " won't match "solar").
|
|
_CRYPTO_FINANCE_KEYWORDS: list[str] = [
|
|
"bitcoin", "btc", " eth ", "ethereum", " sol ", "solana",
|
|
"xrp", "ripple", "dogecoin", "doge", "litecoin", "ltc",
|
|
"coinbase", "binance", "kraken", "bybit", "okx",
|
|
"usdc", "usdt", "stablecoin",
|
|
"defi", "nft", "blockchain", "crypto",
|
|
" fdv", "airdrop", "token launch", "token listing",
|
|
"microstrategy", "mstr", "saylor",
|
|
"nasdaq", "sp500", "s&p 500", "s&p500",
|
|
"federal reserve", "fed rate", "interest rate",
|
|
"inflation", "tariff", "treasury yield",
|
|
"recession", " gdp ", "unemployment", "trade war", "trade deal",
|
|
" ipo ", "sec ", "cftc",
|
|
]
|
|
|
|
_POLITICS_KEYWORDS: list[str] = [
|
|
"election", "president", "congress", "senate", "vote", "war",
|
|
"trump", "biden", "ukraine", "russia", "israel", "nato",
|
|
]
|
|
|
|
_TECH_KEYWORDS: list[str] = [
|
|
" ai ", "openai", "apple", "google", "microsoft", "meta",
|
|
"nvidia", "regulation", "antitrust",
|
|
"tesla", "elon", "nuclear", "quantum", "chip",
|
|
]
|
|
|
|
_EVENTS_KEYWORDS: list[str] = [
|
|
"world cup", "oscar", "nobel", "spacex", "nasa",
|
|
]
|
|
|
|
# Sports markets are excluded entirely — BTC/F&G/GNews have no edge there.
|
|
# Checked before any category match so sports don't bleed into politics/events.
|
|
_SPORTS_EXCLUSIONS: list[str] = [
|
|
" nba ", " nfl ", " mlb ", " nhl ",
|
|
"basketball", "football", "baseball", "hockey", "soccer",
|
|
" mvp ", "rookie of the", "championship", "super bowl", "world series",
|
|
"playoffs", "playoff", "tournament",
|
|
"tennis", " golf ", " ufc ", "boxing", "wrestler", "wrestling",
|
|
"slam dunk", "home run", "touchdown",
|
|
# European / international football leagues
|
|
"la liga", "premier league", "bundesliga", "serie a", "ligue 1",
|
|
"champions league", "europa league", "conference league",
|
|
"copa del rey", "fa cup", "dfb pokal",
|
|
"relegation", "golden boot", "top scorer",
|
|
" liga ", "eredivisie", "primeira liga",
|
|
]
|
|
|
|
@classmethod
|
|
def _is_sports(cls, question: str) -> bool:
|
|
q = f" {question.lower()} "
|
|
return any(kw in q for kw in cls._SPORTS_EXCLUSIONS)
|
|
|
|
@classmethod
|
|
def _is_crypto_finance(cls, question: str) -> bool:
|
|
q = f" {question.lower()} " # pad so edge keywords match cleanly
|
|
return any(kw in q for kw in cls._CRYPTO_FINANCE_KEYWORDS)
|
|
|
|
@classmethod
|
|
def _is_politics(cls, question: str) -> bool:
|
|
q = f" {question.lower()} "
|
|
return any(kw in q for kw in cls._POLITICS_KEYWORDS)
|
|
|
|
@classmethod
|
|
def _is_tech(cls, question: str) -> bool:
|
|
q = f" {question.lower()} "
|
|
return any(kw in q for kw in cls._TECH_KEYWORDS)
|
|
|
|
@classmethod
|
|
def _is_events(cls, question: str) -> bool:
|
|
q = f" {question.lower()} "
|
|
return any(kw in q for kw in cls._EVENTS_KEYWORDS)
|
|
|
|
@classmethod
|
|
def _detect_category(cls, question: str) -> str:
|
|
"""Return the category label for a market question, or '' if unsupported."""
|
|
if cls._is_sports(question):
|
|
return "" # exclude sports regardless of other keyword matches
|
|
if cls._is_politics(question):
|
|
return "politics"
|
|
# Tech checked before crypto/finance: company-specific markets (OpenAI IPO,
|
|
# NVIDIA earnings, Apple antitrust) should be "tech" even when they contain
|
|
# generic finance keywords like "ipo" or "sec".
|
|
if cls._is_tech(question):
|
|
return "tech"
|
|
if cls._is_crypto_finance(question):
|
|
return "crypto/finance"
|
|
if cls._is_events(question):
|
|
return "events"
|
|
return ""
|
|
|
|
async def get_active_markets(
|
|
self,
|
|
min_volume: float = 500,
|
|
pages: int = 3,
|
|
page_size: int = 200,
|
|
max_days_to_resolution: int = 90,
|
|
) -> list[Market]:
|
|
"""Fetch active markets from Gamma API (no auth needed).
|
|
|
|
Fetches events without tag filtering (tag= param is unreliable),
|
|
then keeps only markets whose question matches any supported category
|
|
(crypto/finance, politics, tech, events) and that:
|
|
- have NOT already expired (end_dt >= now)
|
|
- resolve within max_days_to_resolution days
|
|
"""
|
|
seen: set[str] = set()
|
|
markets: list[Market] = []
|
|
now = datetime.now(timezone.utc)
|
|
cutoff = now + timedelta(days=max_days_to_resolution)
|
|
|
|
for page in range(pages):
|
|
try:
|
|
resp = await self._client.get(
|
|
f"{GAMMA_API}/events",
|
|
params={
|
|
"active": True,
|
|
"closed": False,
|
|
"limit": page_size,
|
|
"offset": page * page_size,
|
|
},
|
|
)
|
|
resp.raise_for_status()
|
|
events = resp.json()
|
|
|
|
if not events:
|
|
break # no more pages
|
|
|
|
for event in events:
|
|
event_title = event.get("title", "")
|
|
for m in event.get("markets", []):
|
|
try:
|
|
if not m.get("active") or m.get("closed"):
|
|
continue
|
|
|
|
question = m.get("question", "")
|
|
|
|
# Detect category from question or event title
|
|
category = self._detect_category(question)
|
|
if not category:
|
|
category = self._detect_category(event_title)
|
|
if not category:
|
|
continue
|
|
|
|
# Filter: skip already-expired and far-future markets
|
|
# Gamma API may return endDate or end_date (snake_case)
|
|
raw_end = m.get("endDate") or m.get("end_date") or m.get("endDateIso", "")
|
|
if raw_end:
|
|
try:
|
|
end_dt = datetime.fromisoformat(
|
|
raw_end.replace("Z", "+00:00")
|
|
)
|
|
# Make naive datetimes UTC-aware before comparing
|
|
if end_dt.tzinfo is None:
|
|
end_dt = end_dt.replace(tzinfo=timezone.utc)
|
|
if end_dt < now:
|
|
log.debug("Skipping expired market: %s", question[:60])
|
|
continue
|
|
if end_dt > cutoff:
|
|
continue
|
|
except (ValueError, TypeError):
|
|
pass # keep market if date unparseable
|
|
|
|
market_id = str(m["id"])
|
|
if market_id in seen:
|
|
continue
|
|
|
|
vol = float(m.get("volume24hr", 0))
|
|
if vol < min_volume:
|
|
continue
|
|
|
|
raw_prices = m.get("outcomePrices", ["0.5", "0.5"])
|
|
if isinstance(raw_prices, str):
|
|
import json as _json
|
|
raw_prices = _json.loads(raw_prices)
|
|
yes_price = float(raw_prices[0])
|
|
|
|
raw_tokens = m.get("clobTokenIds", ["", ""])
|
|
if isinstance(raw_tokens, str):
|
|
import json as _json
|
|
raw_tokens = _json.loads(raw_tokens)
|
|
|
|
seen.add(market_id)
|
|
markets.append(Market(
|
|
id=market_id,
|
|
condition_id=m.get("conditionId", ""),
|
|
question=question,
|
|
yes_token_id=raw_tokens[0] if raw_tokens else "",
|
|
no_token_id=raw_tokens[1] if len(raw_tokens) > 1 else "",
|
|
yes_price=yes_price,
|
|
no_price=1 - yes_price,
|
|
volume_24h=vol,
|
|
end_date=m.get("endDate", ""),
|
|
active=True,
|
|
category=category,
|
|
))
|
|
except (KeyError, ValueError, IndexError) as e:
|
|
log.debug("Skipping malformed market: %s", e)
|
|
|
|
except httpx.HTTPError as e:
|
|
log.error("Polymarket API error (page=%d): %s", page, e)
|
|
break
|
|
|
|
by_cat: dict[str, int] = {}
|
|
for mkt in markets:
|
|
by_cat[mkt.category] = by_cat.get(mkt.category, 0) + 1
|
|
log.info(
|
|
"Loaded %d markets (min_vol=%.0f, resolving within %dd): %s",
|
|
len(markets), min_volume, max_days_to_resolution,
|
|
", ".join(f"{k}={v}" for k, v in sorted(by_cat.items())),
|
|
)
|
|
return markets
|
|
|
|
async def get_order_book(self, token_id: str) -> Optional[OrderBook]:
|
|
"""Get order book for a specific token."""
|
|
try:
|
|
resp = await self._client.get(
|
|
f"{POLYMARKET_API}/book",
|
|
params={"token_id": token_id},
|
|
)
|
|
resp.raise_for_status()
|
|
data = resp.json()
|
|
|
|
bids = [(float(b["price"]), float(b["size"])) for b in data.get("bids", [])]
|
|
asks = [(float(a["price"]), float(a["size"])) for a in data.get("asks", [])]
|
|
|
|
mid = 0.5
|
|
if bids and asks:
|
|
mid = (bids[0][0] + asks[0][0]) / 2
|
|
|
|
return OrderBook(
|
|
market_id=token_id,
|
|
yes_bids=bids,
|
|
yes_asks=asks,
|
|
mid_price=mid,
|
|
)
|
|
except Exception as e:
|
|
log.warning("Order book fetch failed for %s: %s", token_id, e)
|
|
return None
|
|
|
|
async def close(self) -> None:
|
|
await self._client.aclose()
|