5 min read

I Reverse-Engineered a $5.27M Polymarket Account. Found a Pricing Bug That Still Exists.

Source: http://x.com/i/article/2026308616203259904

Polymarket

Top 12. $56.6M volume in 6 months. 35,000+ predictions.

Most people see the $5.27M profit and assume this is a prediction genius. I pulled 580,000 on-chain trades from Polymarket's full transaction history and rebuilt the complete strategy.

They don't make money by knowing who wins. They make money by knowing where Polymarket's pricing is wrong.

The pricing error is structural. It's still there today.

Image

Who Is RN1?

The account (

address prefix 0x2005

). Key stats:

  • $56.6M in total volume over 6 months
  • 17,809 unique markets
  • All-time PnL: +$5.27M
  • 91.4% maker ratio

That last number changes everything.

91.4% maker means this account isn't "betting" — it's posting limit orders and waiting for someone to fill them.

On-chain evidence is definitive: 530,000 maker buys totaling $56.5M. Maker sells? 64 trades. $70K total.

Nearly 100% buy-side limit orders. 88.5% of positions held all the way to settlement. This isn't prediction. It's systematic underwriting.

Operating Profile

Peak trading hours concentrate at UTC 17:00-23:00 — the exact window when European soccer matches kick off. Average daily volume around $300K; peak days exceed $1.4M.

Category breakdown: 72% soccer, 18% other sports, 7% tennis, 3% esports.

Median trade size: just $17. Not whale-sized bets. High-frequency, small-amount, continuous market-making. The top counterparty accounts for only 6.1% of flow.

The picture is clear: an automated market-making system that runs during European match hours, buys both sides, and holds to settlement.

What the Chain Data Actually Shows

I started with Polymarket's Activity API and initially concluded this was "directional betting with hedges."

Completely wrong.

The Activity API doesn't distinguish maker from taker — the single most important variable for understanding any trading strategy. I had to pull on-chain trades data — 348 million rows of full trading history — and rebuilt PnL for 3,188 resolved markets using DuckDB:

  • Total invested: $3.98M
  • PnL: +$291K
  • ROI: 7.3%
  • Win rate: 60.9%

7.3% might not sound impressive. But this is systematic edge generated across $56.6M in volume over six straight months. Not a lucky Super Bowl bet. Not a single presidential election call.

And when you break it down by subcategory, some lanes are significantly above average.

The Core Finding: Polymarket Systematically Misprices Favorites

I bucketed all 3,188 resolved markets by entry price. A textbook pattern emerged.

Polymarket systematically underprices favorites and overprices longshots.

In academia, this is called the Favorite-Longshot Bias — well-documented in traditional betting markets, now clearly present on Polymarket. Here are the numbers:

Entry price 0.60–0.70 (implied win rate ~65%) Actual win rate: 79.8% Edge: +14.8%
Entry price 0.70–0.80 (implied win rate ~75%) Actual win rate: 89.1% Edge: +14.1%

Longshots (0.10–0.30) go the other direction — actual win rates consistently fall below their implied probability.

Image

RN1 doesn't need to know who wins any specific match. They just need to know that "an option priced at 0.65 actually wins about 80% of the time."

Broken down by league, every major European soccer league is profitable:

  • EPL: 13.8% ROI (+$54K)
  • Serie A: 10.7% ROI (+$18K)
  • Ligue 1: 10.1% ROI (+$10K)
  • La Liga: 9.4% ROI (+$24K)
  • Bundesliga: 5.4% ROI (+$7K)

Not one lucky league. All five. Structural.

Image

Zooming Into Soccer Markets

Soccer alone accounts for 1,660 resolved markets and $150K in profit (10.2% ROI). The Favorite-Longshot Bias is even more pronounced here than across all sports.

In soccer specifically, favorites priced at 0.60–0.70 actually win 82.9% of the time — nearly 18 percentage points above the implied probability. At 0.70–0.80, the actual win rate is 91.8%.

Meanwhile, longshots priced at 0.10–0.20 win only 6.6% of the time — about half their implied probability.

Image

The gap between the blue line (actual outcomes) and the red line (implied probability) is where the money lives.

Why Does This Bias Persist?

Two reasons, both structural.

First, Polymarket's market design creates it.

Traditional sportsbooks use three-way pricing (Win / Draw / Lose) — one market, one vig layer. Probabilities sum to 100% plus a single margin.

Polymarket does something different. Each outcome gets its own independent Yes/No market. A single soccer match spawns three separate markets ("Will Team A win?", "Will Team B win?", "Will it draw?"), each carrying its own vig layer. Stacked together, this triple vig systematically compresses favorite prices.

Second, retail chases lottery tickets.

Spending $5 to chase a 10x return on a longshot feels exciting. Spending $70 for a 1.4x return on a favorite doesn't. Retail gravitates toward longshots, pushing their prices above fair value, while favorites sit quietly underpriced.

These aren't temporary sentiment shifts. They're built into how the market works. As long as Polymarket uses independent two-way pricing for sports events, this bias is likely to persist.

What I Extracted

From RN1's 17,809 markets, I filtered for the optimal subset:

European soccer only — EPL, Serie A, Ligue 1, La Liga, Bundesliga. All data-verified positive ROI.

Favorites only — Polymarket price between 0.50 and 0.85. This range has the largest, most stable edge.

Pinnacle as the sharp reference — Pinnacle is one of the sharpest sportsbooks globally. The filter: only enter when Polymarket's price is significantly below Pinnacle's implied probability.

First day of scanning produced 32 qualifying signals.

This isn't a strategy I invented. It's reverse-engineered from a $5.27M account's behavior — verified trade by trade with on-chain data.

The most counterintuitive thing about prediction markets: the biggest winners often aren't the best predictors. They're the ones who find where the pricing is wrong.

RN1 made $5.27M in six months. Not by knowing soccer better than everyone. By understanding Polymarket's pricing mechanics better than everyone.

The data is public. The bias is structural. The method is verifiable.

I research and study prediction market strategies on

Polymarket

.

This is a personal research and learning log, not investment advice. If you spot data errors or analytical flaws, I'd appreciate the feedback.