Trade analytics for pro traders.
🏆 Wave 1 & Wave 2 winner (1st place, both). Earlier submissions are preserved at tags
wave-1-finalandwave-2-final. The currentmainis Wave 3 — the final wave.
Edgework is a performance intelligence tool for serious traders on SoDEX. It turns your raw order history into the one thing PNL doesn't show: conditional edge — the slices of time, regime, and behavior where you actually make money, and the slices where you give it back.
Most traders discover that 60–80% of their losses come from 10–20% of their setups. Edgework finds those setups, benchmarks them against the SoDEX leaderboard, and — new in Wave 3 — gives you a verdict before you take your next trade.
Live app: edgework.streamlit.app Submission: SoSoValue Buildathon
The arc: Wave 1 measured edge → Wave 2 compared it to the leaderboard → Wave 3 turns it into a pre-trade decision you can act on.
Everything below works by just visiting the site — no wallet to connect, nothing to install, no login. Paste any SoDEX address (or hit Use demo data, which loads a real top trader) and it's live.
Thinking about a trade? Pick a symbol + side and Edgework tells you whether to take it — instantly — by combining three signals:
- Your own history on that exact symbol+side (winrate, expectancy, n).
- The live smart-money book — aligned, contrarian, or no clear bias (same thresholds as the Smart Money Watch).
- Your edge in the current BTC regime for that setup.
A colour-coded verdict runs from "🛑 strong skip" to "✅ green light". Flip long/short and it recomputes live. It's the risk-hook the Wave 2 judges asked for, delivered as a tool you actually use — honest about a read-only tool's limit: it won't place the order, it tells you whether to.
Reconstructs the smart-money book at each of your past entries and shows how your aligned vs contrarian trades actually performed — the evidence behind the verdict.
Bootstrap-based verdict confidence (no more decorative %), retries, parallel fetches, low-sample badges, fees decomposition, and a tilt detector — plus a TL;DR card, anchor nav, tooltips, and a full visual redesign: brand logo, a clean two-option sidebar, and an elegant above-the-fold landing.
See CHANGELOG.md for the full list.
Everything from Wave 1, plus six new value props that turn Edgework from a diagnostic dashboard into a decision tool you use every session:
A dashed overlay on your equity chart showing what your PNL would have been if you'd skipped the trades flagged by the risk filters. Quantifies the dollar value of your worst patterns directly on the chart — not buried in a table somewhere.
A real-time snapshot of where the top 20 active+profitable SoDEX traders are positioned right now. We pull the top 50 by 30-day volume, filter to those with positive PNL (so you don't see the lucky one-shot whales), then aggregate their open positions per symbol into LONG / SHORT / NET exposure. Refreshes every 15 min.
Compares each of your open positions to the smart-money consensus on the same symbol. Flags ✓ aligned, ⚠ contrarian (≥3 traders or 2× notional dominance opposite to you), or **~ mixed** — the groundwork the Wave 3 Trade Check turns into a pre-trade verdict.
Replaces the old PNL rank (skewed by 118k dormant wallets) with 30-day volume rank — a real signal of activity, not noise. Tells you where you stand among traders who actually trade.
Replaces the old briefing + Q&A buttons with one "Generate diagnostic"
CTA. A deterministic tool-use loop powered by Claude Sonnet 4.6 that always
calls get_full_breakdown first (one cheap call covers ~all data), then
produces a structured autopsy: the 5 biggest problems destroying your
account + 3-5 specific rules to implement. Every cited number is
backed by a tool call against your data — no hallucinations.
Full UI translation — toggle at the top right. Topbar, slicers, equity curve, verdict, confrontation, waterfall, performance tabs, peer benchmark, smart money watch, your-positions-vs-smart-money, risk filters, diagnostic prompt, bottombar — every label and every dynamic insight string.
- Bookmarkable URL state — slicer filters + wallet survive a refresh / share
- BTC regime tagging (uptrend / downtrend / chop) as a slicer dimension
- 2D risk filter (cross-dimensional anti-pattern detector)
- Heatmap day-of-week × hour-of-day
- Peer benchmark vs top-5 traders with PT/EN dynamic insights
-
Conditional Performance Mapping. Pulls every order from your SoDEX account and slices winrate, expectancy, and time-in-trade across dimensions you don't normally see: time of day, day of week, BTC regime, consecutive losses, position size quartile, hold duration, side bias, and per-symbol performance.
-
Leaderboard-Benchmarked Alpha. Compares your conditional performance against the public SoDEX leaderboard top traders. You don't just see "I'm losing in this slice" — you see "top traders also fade this setup; you're trading their losing setup."
-
AI-Powered Diagnostic. A single-call Claude diagnostic that runs a full autopsy of your history and produces ranked, dollar-quantified problems + specific rules. Prompt-cached for repeat-session efficiency.
- Python 3.13+ for the analytics core
- Streamlit ≥ 1.39 for the live demo (deployable to Streamlit Cloud)
- Pandas 3.x + Plotly for slicing and visualization
- Anthropic Claude Sonnet 4.6 for the diagnostic engine (with prompt caching + tool use)
- SoSoValue API for news, indexes, ETF flows, sector rotation
- SoDEX Public API for account history, leaderboard, live positions
- ValueChain for on-chain proof of trade history (later phase)
edgework/
├── src/edgework/
│ ├── sodex_client.py # SoDEX API client (read-only, account + leaderboard)
│ ├── sosovalue_client.py # SoSoValue API client (news, indexes, ETF flows)
│ ├── slicer.py # Conditional Performance Mapping core
│ ├── smart_money.py # Live top-trader consensus + open-positions fetch
│ ├── risk.py # 2D anti-pattern engine (powers Trade Check)
│ ├── qna.py # Full AI diagnostic engine (Claude tool use)
│ ├── briefing.py # Legacy briefing (kept for back-compat)
│ └── config.py # Settings, API keys via env vars
├── tests/ # slicer, qna, risk, + module tests (39 passing)
├── data/ # Local parquet/CSV cache (gitignored)
├── docs/ # Architecture notes, API usage plan
├── streamlit_app.py # Live app entry point
├── pyproject.toml
└── README.md
git clone https://github.com/nftradercrypto/edgework.git
cd edgework
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
cp .env.example .env
# fill in:
# SODEX_USER_ADDRESS — your EVM wallet (read-only, no auth needed)
# SOSOVALUE_API_KEY — for news / indexes / ETF flows
# ANTHROPIC_API_KEY — for the AI diagnostic
streamlit run streamlit_app.pyThe app loads a wallet's full SoDEX history via the public read API
(mainnet-gw.sodex.dev), so no upload step is required — paste any
SoDEX wallet address and you see the analytics immediately.
Built solo by @nftradercrypto. Wave 1 winner (1st place) of the SoSoValue Buildathon. Top-3 weekly volume trader on SoDEX Season 1, creator of TokenBar on SoSoValue.
MIT.