Spaces:
Running
Running
File size: 11,274 Bytes
88d2f2a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 | """Tests for the backtest framework.
The mock-LLM path is exercised end-to-end so the test suite stays fast
(<5s) and offline. Real-LLM behaviour is asserted indirectly via the
LLM-factory swap test.
"""
from __future__ import annotations
import json
from pathlib import Path
import pytest
from polyglot_alpha.backtest.outcome_matcher import (
OutcomeComparison,
compare_questions,
infer_category,
infer_framing,
)
from polyglot_alpha.backtest.roi_estimator import (
BUILDER_FEE_BPS,
CAPTURE_RATE_FAIL,
CAPTURE_RATE_PASS,
CAPTURE_RATE_PASS_HIGH,
HIGH_CONFIDENCE_THRESHOLD,
estimate_roi,
)
from polyglot_alpha.backtest.runner import (
MarketRecord,
_pick_winner,
load_markets,
run_backtest,
)
# --------------------------------------------------------------------------- #
# Fixtures #
# --------------------------------------------------------------------------- #
@pytest.fixture()
def mock_markets() -> list[MarketRecord]:
"""Three deterministic market records spanning YES / NO / dispute."""
return [
MarketRecord(
market_id="bt-1",
question="Will Bitcoin exceed $100k by 2026-12-31?",
category="crypto",
outcome="YES",
total_volume_usdc=50_000.0,
uma_dispute=False,
resolution_source="https://example.com/btc",
),
MarketRecord(
market_id="bt-2",
question="Will the Fed cut rates in March 2026?",
category="economics",
outcome="NO",
total_volume_usdc=20_000.0,
uma_dispute=True,
resolution_source="https://example.com/fed",
),
MarketRecord(
market_id="bt-3",
question="Will Apple announce a foldable iPhone before 2026-12-31?",
category="tech",
outcome="NO",
total_volume_usdc=8_000.0,
uma_dispute=False,
resolution_source="https://example.com/aapl",
),
]
# --------------------------------------------------------------------------- #
# ROI estimator #
# --------------------------------------------------------------------------- #
class TestRoiEstimator:
def test_pass_high_confidence_uses_top_capture_rate(self) -> None:
roi = estimate_roi(100_000.0, "PASS", HIGH_CONFIDENCE_THRESHOLD)
assert roi.capture_rate == pytest.approx(CAPTURE_RATE_PASS_HIGH)
expected_fee = 100_000.0 * CAPTURE_RATE_PASS_HIGH * (BUILDER_FEE_BPS / 10_000.0)
assert roi.builder_fee_usdc == pytest.approx(expected_fee)
# Net = builder_fee - agent_cost; should still be positive on a 100k market.
assert roi.net_roi_usdc > 0
def test_pass_normal_uses_lower_capture_rate(self) -> None:
roi = estimate_roi(100_000.0, "PASS", HIGH_CONFIDENCE_THRESHOLD - 1)
assert roi.capture_rate == pytest.approx(CAPTURE_RATE_PASS)
def test_fail_returns_zero_fee(self) -> None:
roi = estimate_roi(1_000_000.0, "FAIL", 0)
assert roi.capture_rate == pytest.approx(CAPTURE_RATE_FAIL)
assert roi.builder_fee_usdc == 0.0
# Net negative because of agent_cost stub.
assert roi.net_roi_usdc < 0
def test_zero_volume_returns_zero(self) -> None:
roi = estimate_roi(0.0, "PASS", 95.0)
assert roi.builder_fee_usdc == 0.0
def test_negative_volume_clamped_to_zero(self) -> None:
roi = estimate_roi(-500.0, "PASS", 95.0)
assert roi.builder_fee_usdc == 0.0
# --------------------------------------------------------------------------- #
# Outcome matcher #
# --------------------------------------------------------------------------- #
class TestOutcomeMatcher:
def test_identical_questions_match_with_jaccard(self) -> None:
result: OutcomeComparison = compare_questions(
"Will Bitcoin exceed $100k by 2026-12-31?",
"Will Bitcoin exceed $100k by 2026-12-31?",
"YES",
use_embeddings=False,
)
assert result.semantic_similarity == pytest.approx(1.0)
assert result.semantic_match is True
assert result.framing_predicted == "YES"
assert result.outcome_match is True
def test_disjoint_questions_low_similarity(self) -> None:
result = compare_questions(
"Will Apple ship a foldable iPhone?",
"Will the Fed cut interest rates?",
"NO",
use_embeddings=False,
)
assert result.semantic_similarity < 0.3
def test_framing_yes_matches_yes_outcome(self) -> None:
result = compare_questions(
"Will the policy be announced before December?",
"Will the policy be announced before December?",
"YES",
use_embeddings=False,
)
assert result.framing_predicted == "YES"
assert result.outcome_match is True
def test_framing_yes_misses_on_no_resolution(self) -> None:
result = compare_questions(
"Will Apple announce a foldable iPhone?",
"Will Apple announce a foldable iPhone?",
"NO",
use_embeddings=False,
)
# Question framing is YES, actual is NO → miss.
assert result.framing_predicted == "YES"
assert result.outcome_match is False
def test_non_binary_outcome_is_not_matched(self) -> None:
result = compare_questions(
"Will Verstappen win the race?",
"Race winner?",
"Verstappen",
use_embeddings=False,
)
assert result.outcome_match is False
assert "non-binary" in result.notes
def test_infer_framing_yes(self) -> None:
assert infer_framing("Will X reach 100 by year-end?") == "YES"
def test_infer_framing_no(self) -> None:
# "Below" should mark this as a NO-framing.
assert infer_framing("Will X fail to stay below the limit?") == "NO"
def test_infer_framing_unknown(self) -> None:
assert infer_framing("xyz") == "UNKNOWN"
def test_infer_category_crypto(self) -> None:
assert infer_category("Will Bitcoin reach $100k?") == "crypto"
def test_infer_category_other(self) -> None:
assert infer_category("Random unrelated string") == "other"
# --------------------------------------------------------------------------- #
# Auction logic #
# --------------------------------------------------------------------------- #
class TestAuction:
def test_pick_winner_picks_lowest_bid(self) -> None:
import random as _random
rng = _random.Random(0)
bids = {"gemini": 0.30, "deepseek": 0.75, "qwen": 0.40}
assert _pick_winner(bids, rng=rng) == "gemini"
def test_pick_winner_handles_ties_deterministically(self) -> None:
import random as _random
rng = _random.Random(123)
bids = {"a": 0.5, "b": 0.5, "c": 0.5}
first = _pick_winner(bids, rng=rng)
# With the same seed we get the same answer.
rng = _random.Random(123)
second = _pick_winner(bids, rng=rng)
assert first == second
# --------------------------------------------------------------------------- #
# Market loader #
# --------------------------------------------------------------------------- #
class TestLoadMarkets:
def test_loads_from_real_parquet_if_available(self) -> None:
repo_root = Path(__file__).resolve().parents[1]
parquet = repo_root / "corpus" / "polymarket_resolved.parquet"
if not parquet.exists():
pytest.skip("resolved markets parquet not present in this checkout")
markets = load_markets(n=3, parquet_path=parquet, seed=42)
assert len(markets) == 3
assert all(isinstance(m, MarketRecord) for m in markets)
assert all(m.question for m in markets)
def test_falls_back_to_sample_json(self, tmp_path: Path) -> None:
# Point at a non-existent parquet so the loader uses sample_*.json
# via the default ``outputs/`` directory.
markets = load_markets(n=2, parquet_path=tmp_path / "missing.parquet", seed=42)
assert len(markets) >= 1
assert all(isinstance(m, MarketRecord) for m in markets)
# --------------------------------------------------------------------------- #
# End-to-end smoke test (mock LLM) #
# --------------------------------------------------------------------------- #
class TestRunBacktestSmoke:
def test_full_pipeline_with_mock_llm(
self,
mock_markets: list[MarketRecord],
tmp_path: Path,
) -> None:
import asyncio
from polyglot_alpha.backtest.runner import run_backtest_async
summary = asyncio.run(
run_backtest_async(
n=len(mock_markets),
seed=42,
output_dir=tmp_path,
mock_llm=True,
use_embeddings=False, # avoid the sentence-transformers download
markets=mock_markets,
)
)
assert summary["n_markets"] == len(mock_markets)
# Output files should have landed in tmp_path.
jsonl = tmp_path / "per_market_results.jsonl"
summary_path = tmp_path / "summary.json"
report_path = tmp_path / "backtest_report.md"
assert jsonl.exists()
assert summary_path.exists()
assert report_path.exists()
# Re-read JSONL and confirm one row per market.
rows = [json.loads(line) for line in jsonl.read_text().splitlines() if line]
assert len(rows) == len(mock_markets)
row0 = rows[0]
# Sanity-check required fields per the spec.
for key in (
"market_id",
"actual_question",
"actual_outcome",
"actual_volume",
"agent_winner",
"agent_question",
"judge_verdict",
"judge_score",
"semantic_similarity",
"outcome_match",
"estimated_roi_usdc",
"uma_dispute",
"category",
"notes",
):
assert key in row0, f"missing key {key} in row"
# Markdown report is non-empty and labelled.
report_text = report_path.read_text()
assert "PolyglotAlpha v2 Backtest Report" in report_text
assert "Executive summary" in report_text
def test_run_backtest_sync_wrapper(
self,
mock_markets: list[MarketRecord],
tmp_path: Path,
) -> None:
summary = run_backtest(
n=len(mock_markets),
seed=99,
output_dir=tmp_path,
mock_llm=True,
use_embeddings=False,
markets=mock_markets,
)
assert summary["n_markets"] == len(mock_markets)
assert "outcome_accuracy" in summary
assert "estimated_total_roi_usdc" in summary
|