I Filtered 512 “Profitable” Polymarket Wallets Down to 314. Here’s What the Data Actually Showed

Polymarket smart money is easy to misread if you only look at total profit. I set out to answer a simpler question: who’s actually making money on Polymarket, and can you copy them?

Starting With 512 Wallets

The first pass was mechanical. Pull historical trades, calculate real profit and loss per wallet, keep everyone who came out ahead. That gave me 512 profitable addresses — a reasonable starting pool for a copy-trading system.

But profitable isn’t the same as copyable. A wallet that made $50,000 on one lucky trade and lost on everything else looks identical to a consistently skilled trader if you only check the final number. So the next filter wasn’t about size — it was about repeatability. I scored every wallet on how consistent its edge was over time, not how big any single win was.

314 wallets survived that cut. That’s the pool the alert system actually watches — it flags movement across this filtered pool as a capital-flow indicator, not a standalone trading signal.

The API Was Lying to Me (Sort Of)

Here’s where it got interesting. While digging into individual wallets to understand how each one makes money — not just that they do — I noticed the numbers didn’t add up. A wallet showing consistent profits in the platform’s own trade history had a P&L that didn’t reconcile with what the UI displayed.

The issue: Polymarket’s public API, pulled the straightforward way, skews toward showing filled, resolved, profitable-looking trades. Losses and messier positions don’t always surface the same way. If you build your profitability ranking off that data without correcting for it, you’re measuring a partial picture — closer to a highlight reel than a full track record.

Fixing it meant rebuilding the parser to pull full position history per wallet, request-by-request, across every market that address had touched — thousands of individual permission checks in some cases. Slower, but the resulting numbers matched the real P&L shown in the UI. Survivorship bias is easy to build into a system by accident, and it’s the kind of error that quietly ruins a copy-trading strategy: you end up copying wallets that look better than they are.

The Whale Who Can’t Be Copied

The clearest example of why “profitable” and “copyable” are different things came from one wallet in the top tier: real profit around $2.86 million, tiny drawdown, thousands of trades.

Looking at the actual trade log explained the number. This wallet’s edge is buying near-certain outcomes — positions priced at $0.99 or higher, essentially betting “this will not happen” on longshots — in enormous size. One entry: roughly $500,000 into a single position priced at 0.995, for a return of about half a percent. Sports markets made up 71% of the activity, and the pattern was almost entirely buy-and-hold to resolution — barely any selling before the outcome settled.

The math behind that strategy is brutal for anyone without deep pockets. At 0.995, a single loss wipes out roughly 200 winning trades of the same size. This wallet’s edge isn’t picking better outcomes than anyone else — it’s having enough capital to make a razor-thin, high-probability spread worth doing at scale, and enough discipline to almost never be wrong on the calls it makes.

Copy this wallet with a normal account size and the math breaks immediately. The position sizes that generate meaningful returns for a multi-million-dollar bankroll round down to nothing — or below Polymarket’s order minimums — when scaled to a retail balance. It’s a genuinely skilled trader. It’s also, for most people, a wallet to watch rather than copy.

What Actually Separates Copyable From Not

Running this filter across hundreds of wallets surfaced a pattern that isn’t obvious from a leaderboard: the traders worth copying usually aren’t the ones with the biggest total P&L. They’re the ones whose edge doesn’t depend on capital you don’t have.

A few markers that showed up consistently in the copyable group:

• Position sizes that scale down without breaking the strategy
• Entries across a range of prices, not exclusively near-certain outcomes
• Enough trade volume to judge consistency, not one or two lucky calls
• A win rate that holds up when you weight it by capital at risk, not just by trade count

None of this is visible from a simple profit ranking. It only shows up once you pull the real trade history and look at how the money was made, not just that it was.

Why This Matters If You’re Trying to Copy Trade

The takeaway isn’t “don’t copy trade” — it’s that the naive version of copy trading, where you just mirror whoever’s at the top of a leaderboard, is a good way to copy a strategy you can’t actually run. The wallet with the eye-watering P&L might be running a strategy that requires seven figures and near-zero error tolerance. Copying it with a few hundred dollars doesn’t scale the strategy down — it breaks it.

That’s the filtering problem I built this system to solve, and it’s also the reason the alert system I run now watches a curated pool of 314 wallets instead of chasing whoever’s on top of a leaderboard this week. Overdog tracks this pool directly. The top 117 of those wallets — sparkline P&L, archetypes, current positions, and a flag on which ones need more capital than a retail account can supply — are broken out on our Smart Money Dossier.

If you’re going to copy trade on Polymarket, the question isn’t “who made the most money.” It’s “whose edge can I actually run with the capital I have.” Those are frequently not the same wallet.