⚡

Polymarket Alpha

Smart money discovery + signal generation + paper-validated edge measurement on Polymarket prediction markets.

📌 What this is

Polymarket is a $1B+ decentralized prediction market. Like in equities, a small group of informed traders consistently outperform the crowd — they have edge from research, news access, or domain expertise. This system finds them, tracks them, and tells us what they're betting on, in real-time.

Whales tracked

—

tier S/A/B

Universe scanned

—

candidates analyzed

Alerts (history)

—

sent to Telegram

Paper positions

—

$100 each, tracked

Paper P&L

—

realized + unrealized

🎯 The thesis

Three observations from analyzing thousands of Polymarket wallets:

Top-50 PnL ≠ skill. 38 of the top 50 PnL wallets have win rate <60% — they got lucky on home-run bets, not consistent edge.
Specialization wins. Whales with 75%+ WR in a single category (politics, futures sports) hold edge across all their bets in that vertical.
Consensus is the strongest signal. When 2+ skilled traders enter the same market independently, historical WR is ~10-15pp above their solo trades.

The system codifies these into a discovery engine, scoring algorithm, and signal pipeline.

🏗️ Architecture

┌────────────────────────────────────────────────────────────┐ │ POLYMARKET PUBLIC API │ └────────────────────────────────────────────────────────────┘ ↓ ↓ ┌─────────────────┐ ┌──────────────────┐ │ DISCOVERY │ │ ACTIVITY │ │ every 6h │ │ every 5 min │ │ │ │ │ │ Paginate │ │ For each whale: │ │ /trades $1K+ │ │ pull recent │ │ → ~1500 wallets│ │ trades, score │ │ → analyze each │ │ │ │ → tier S/A/B │ │ │ └─────────────────┘ └──────────────────┘ ↓ ↓ ┌──────────────────────────────────────────────┐ │ SCORING + GROUPING ENGINE │ │ Score = tier + cat + size + price + consensus│ │ Group: multiple trades on same event = 1 alert│ └──────────────────────────────────────────────┘ ↓ ↓ ↓ Telegram Dashboard Paper Trades (live alerts) (web UI) ($100 hypothetical) ↓ ↓ ↓ Your phone Team browsers Edge measurement

📊 Tier system

Wallets are filtered by hard quantitative criteria. No subjective ranking.

Tier	Win Rate	PnL	Sample Size	Treatment
S	≥ 68%	≥ $30K	≥ 30 trades	Elite — high-conviction copy candidates
A	≥ 60%	≥ $10K	≥ 15 trades	Consistent — bulk of alerts
B	≥ 55%	≥ $3K	≥ 12 trades	Promising — watch list, weight lower

🚀 5-Phase roadmap to autonomous trading

Build the intelligence layer

Whale discovery, scoring, Telegram alerts, dashboard. Trade by hand from alerts.

✅ DONE

Paper trading validation

Every alert opens a hypothetical $100 position. Track mark-to-market + resolutions. Measure real edge before risking capital.

▶ CURRENT

Auto-trade tier S consensus only

Once paper shows positive ROI on a subset (likely consensus + tier S), enable real execution with fixed $50-100 sizing. Hard daily/weekly loss limits.

~4-6 WEEKS

Dynamic Kelly sizing + expand signal types

Move from fixed $ to fractional Kelly based on each whale's measured edge. Open more signal types based on per-bucket paper performance.

~8-12 WEEKS

Full auto across all whales

Tier A + B included with proportional sizing. Active position management (entry/exit, rebalancing). Risk caps automated.

~3-6 MONTHS

🛠️ How to use this

Dashboard

icosagon-alpha.duckdns.org shows everything live. Click any whale or market for deep stats. Most useful tabs:

Dashboard — overview, hot signals last 24h, top whales by Sharpe
Opportunities — every recent trade from tracked whales, filterable
Bot Feed — exact mirror of what got sent to Telegram
Paper Trades — hypothetical P&L if we'd auto-traded every alert
Markets — markets ranked by smart money activity

Live alerts arrive in our channel automatically. Interactive commands available via DM with @PolitoMBot:

/status — system health, active whales, alerts today
/whales — top whales by tier with stats
/paper — current paper P&L summary
/help — full command list

👥 Team roles

Person	Focus
Dario	System architect / execution lead. Owns infrastructure, scoring tuning, go/no-go decisions on phase transitions.
Manuel	Capital allocation & risk caps. Decides per-phase position limits and overall capital deployment.
Lucio	Sourcing — market intelligence on Polymarket trends, validating discovered whales via external research.
Ernesto	BD — communicating wins / methodology to investors, demo for partners.
Tz	AWS infra co-owner. Eventual auto-execute engine (CLOB SDK integration, private key custody).
Tomas	PM — coordination across phases, ensures we don't skip validation.

⚠️ What this is NOT (yet)

Not auto-trading. The bot only sends alerts. Real capital is not deployed yet. Phase 2 (paper trading) needs ~2-4 weeks of resolution data before we can responsibly move to Phase 3.

Not financial advice. Sample sizes are still small. Whales can lose. Always size positions according to your own risk tolerance.

Built on public data. Everything visible here comes from the public Polymarket API. No private information, no insider data, no proprietary feeds.

❓ FAQ

Why prediction markets?

They aggregate beliefs into prices, but unlike stocks the universe is narrower (one market per question) and edges are larger (less institutional competition). Polymarket alone moved $9B+ in 2024-2025.

Why copy-trading instead of building our own predictions?

Speed and humility. Building a Trump-vs-Biden model from scratch is months of work. Identifying who already has that model and following them is days. We can always build our own later.

How often do alerts come?

Currently ~15-25 alerts per day in the Telegram channel. The bot caps at 3 alerts per whale per cron run to avoid spam.

Can someone front-run us by seeing the same alerts?

Theoretical risk but limited in practice: the whale already bought before we see it. We're following, not racing. Latency from whale-trade → our-alert is typically 1-5 minutes.

What happens if a whale starts losing?

Every 6 hours discovery re-runs. Whales below tier thresholds drop off automatically. We don't manually maintain the list.

When do we put real money in?

Phase 3, after paper trading shows positive ROI on at least one well-defined signal subset over 100+ resolved trades. That's ~4-6 weeks at current alert volume.