Market Anomaly Scanner
Flags cross-market deviations across spot, perps, options, ETFs, chain flows, macro events, and news catalysts — with z-scores, source attribution, and timestamps.
Research copilot for trading desks
An AI research analyst for quant teams
that never stops reading the market.
SignalForge scans price action, funding, open interest, ETF flows, on-chain data, news, filings, and macro events to generate testable trading hypotheses, Python backtest drafts, risk notes, and daily research briefings.
SignalForge turns market noise into
testable trading hypotheses,
backtest-ready Python, and
risk-aware daily briefings.
For
Crypto funds
Prop desks
Quant pods
PMs & research leads
scanner
hypotheses (3)
backtests (1)
briefs
BTC funding divergence
+2.8σ
Funding ↑ while ETF flow ↓ · 30d band
src: coinglass + sosovalue
ETH OI change (24h)
+14.2%
Leverage building into CPI window
src: binance + bybit + okx
SOL exchange inflow
$186M
Above 30d upper band (2.1σ)
src: glassnode + nansen
HYPOTHESIS · #h-2026-0511-04
BTC short-term long squeeze risk
conf 0.71
If perp funding stays elevated while spot ETF inflow decelerates, crowded long exposure may unwind on weak volume confirmation.
backtest_draft.py · auto-generated
# 2026-05-11T07:10Z — funding/ETF divergence
signal = funding_z > 2.0 and etf_flow_3d < ma_20
exit_if = spot_volume_z > 1.5 or basis_norm < 0
# invalidation: breakout confirmed by volume
Research pipeline
From market noise to desk-ready research in five steps.
Every step is auditable. Every output is something a human PM, trader, or quant can challenge, edit, and run themselves.
-
01 SCAN
Scan
Continuously watches anomalies across price, funding, OI, ETF flows, chain data, news, filings, and macro releases.
-
02 HYPOTHESIZE
Hypothesize
Generates falsifiable trading hypotheses around funding, OI, chain flow, volatility, liquidity, and event catalysts.
-
03 BACKTEST
Backtest
Drafts Python notebooks with data assumptions, factor definitions, entry logic, exits, and invalidation checks.
-
04 EXPLAIN
Explain
Summarizes why a strategy worked or failed, what regime matters, and where the risk assumptions break.
-
05 BRIEF
Brief
Delivers morning briefs and risk notes to PMs, traders, and research leads before the desk gets busy.
Core modules
Built for research velocity, not generic chat.
Five focused modules that fit how a real trading desk works — scanning, hypothesizing, verifying, explaining, and shipping briefings.
Strategy Hypothesis Engine
Turns anomalies into ranked, falsifiable trading hypotheses with required data, expected failure modes, and confidence priors.
Python Backtest Drafts
Produces notebook-ready Python scaffolds — data loaders, factor definitions, entry/exit logic, validation — so quants verify ideas instead of starting from a blank file.
Risk & Failure Explanation
Explains drawdowns, regime dependency, liquidity sensitivity, data leakage risk, and the precise invalidation conditions for each thesis.
Morning Brief / Risk Brief Delivery
Ships concise pre-market briefs and intraday risk notes to PMs, traders, and research leads through email, Slack, or Telegram.
Backtest drafts, not blank pages
Each hypothesis ships with a runnable Python scaffold.
SignalForge writes the first draft so your quants spend their time on validation, not boilerplate. Every draft includes data sources, factor logic, entry/exit rules, and invalidation conditions.
# SignalForge AI — auto-generated backtest draft
# Hypothesis #h-2026-0511-04
import pandas as pd
from signalforge.data import Coinglass, SoSoValue, Binance
funding = Coinglass.funding_rate("BTCUSDT", freq="8h")
etf_flow = SoSoValue.etf_net_flow("BTC")
spot_vol = Binance.spot_volume("BTCUSDT", freq="1d")
# Factor construction
funding_z = (funding - funding.rolling(30).mean()) / funding.rolling(30).std()
etf_flow_3d = etf_flow.rolling(3).sum()
spot_volume_z = (spot_vol - spot_vol.rolling(30).mean()) / spot_vol.rolling(30).std()
basis_norm = (futures_basis() / atr(14)).clip(-5, 5)
# Entry / exit logic
entry = (funding_z > 2.0) & (etf_flow_3d < etf_flow_3d.rolling(20).mean())
exit_ = (spot_volume_z > 1.5) | (basis_norm < 0)
# Invalidation: spot volume confirms breakout, abandon short bias
positions = build_short_positions(entry, exit_, max_hold=10)
# Report: regime-conditional sharpe, max DD, factor stability
report = sf.evaluate(positions, regimes="funding_z > 2")
report.to_html("reports/h-2026-0511-04.html")
Sample morning brief
Research output your desk can challenge, edit, and test.
Every brief separates evidence, hypothesis, backtest plan, and invalidation logic — so PMs and traders decide what deserves attention before the open.
- No vague recommendations
- Every claim has a data source
- Every thesis has a kill condition
- Backtest draft attached when applicable
BTC perp funding divergence while spot demand cools
- Signal
- Funding remains above the 30d upper band while ETF inflow momentum has faded for three sessions (-$184M cumulative).
- Hypothesis
- Short-term long squeeze risk increases if price fails to confirm above the weekly breakout level on rising spot volume.
- Backtest draft
- Python scaffold attached — funding z-score, ETF flow momentum, basis normalization, volume filter, 48-month lookback.
- Invalidation
- Ignore the short-bias setup if spot volume confirms breakout (z > 1.5) or basis normalizes below 60d median.
Built by quant practitioners
An AI tool shaped by how trading desks actually work.
Desk pricing
Priced for teams that already spend heavily on research and data.
All plans include data integrations, daily research briefings, and human-in-the-loop review. Annual contracts available with custom terms.
Research Desk
Independent quant · small pod
$1,500/mo
- Daily morning brief
- Market Anomaly Scanner
- 25 hypothesis runs / day
- Python Backtest Drafts
- Email delivery
Most teams
Fund Team
Crypto fund · prop desk · multi-strat
$5,000/mo
- Team research workspace
- Risk & Failure Explanation
- Custom data source routing
- Unlimited hypothesis runs
- Morning Brief + Risk Brief delivery
- Slack & Telegram integration
Enterprise
Private deployment · strict controls
Custom
- Dedicated model routing
- Private data connectors
- Audit trails & access controls
- Research operations support
- On-prem / VPC deployment
Clear operating boundary
Research assistance, not automated trading.
SignalForge AI does not execute trades, does not manage capital, does not custody assets, and does not promise returns. It supports research, validation, risk review, and decision workflows for professional teams that keep humans in control.
- No execution, no order routing, no broker integrations
- No fund management, no custody, no fiduciary role
- Outputs are research drafts — humans review and decide
- Not investment advice; outputs are not signals to trade
Bring AI-native research coverage to your trading desk.
See how SignalForge turns market data, news, and chain activity into hypotheses your team can test — before the market opens.