I spent eleven months trying to beat Polymarket. I read every market description. I built a forecast spreadsheet. I marked my calendar with Fed meetings, election dates, earnings windows. I finished those eleven months down roughly eighteen percent on a five-figure starting bankroll. The day I gave up was the day the company I now run became inevitable.
This is a short essay, not a study. The studies are elsewhere on the blog — we are obsessive about them and they hold up. But before we ran the numbers we made the mistake, and I think the mistake is the more useful thing to read about if you are new to this market.
Why I thought I could beat it
I came to Polymarket from a Chicago prop desk where I traded equity-index variance for four years. I had a quant background, a working trading instinct, and the conviction that a venue full of retail bettors had to be inefficient enough for someone like me to take real money out of it. That conviction was half right. The market was inefficient. I was not the one taking money out.
The flaw in my reasoning was a category error. I assumed that “market inefficiency” meant “easy edge for a smart trader.” What it actually means is “edge available to whoever has the right combination of information, infrastructure, and time on this specific venue.” I had the first one in patches. I had none of the second. And the third — the patience to do this consistently across categories I did not specialise in — turned out to be the binding constraint.
What I actually did for eleven months
I traded everything. Fed decisions because I had a macro view. NBA playoffs because I had watched basketball since I was nine. Elections because I read three newsletters. Earnings because I had spent a career in adjacent markets. Crypto-price binaries because of course.
Across that range I averaged a hit rate that beat implied probability by roughly two points. On any normal venue, two points of edge is a real number. On Polymarket, with the bid-ask I was paying, the slippage I did not measure, and the position-sizing discipline of a man who was certain he was right, two points evaporated. By month four I was flat. By month seven I was down twelve. By month eleven I had stopped opening the app on weekends because I did not want to know.
The temptation in a piece like this is to blame the venue. I have read those posts and I do not believe them. Polymarket is a fine market. The problem was that I was a generalist trader on a venue where edge lives in specialisation, and I was undercapitalised relative to my variance, and I was making decisions with my evenings instead of with infrastructure. None of these are Polymarket’s problem. All of them were mine.
The thing I noticed before I quit
Around month nine I started screening the leaderboard. Not to copy — copy-trading was not really a category yet and the existing tools were thin — but because I wanted to see what I was competing against. The patterns were immediate.
The top wallets were not generalists. The number-one ranked wallet in late summer 2025 had filled four hundred and forty-three trades, every single one of them in US state-level politics. The number-three wallet was a soccer specialist with eight hundred and sixty-two NBA-zero trades. The crypto-price book had its own small group of dominant addresses that never touched a market that did not have BTC, ETH, or SOL in its name.
These people were not smarter than me. They were narrower. They had decided in advance which set of questions they would answer and which set they would refuse to answer, and they had stuck to that decision through every interesting-looking distraction the venue threw at them.
The edge on Polymarket is not in being right more often. It is in being right about fewer things, more reliably, with capital sized small enough that you live to see the next opportunity.
What changed when I stopped
I closed every open position in October 2025 and stopped trading on my own account. I spent the next month writing the wallet-scoring system that became the first version of the leaderboard you see on this site now. The work was easier than the trading had been, partly because I had finally accepted that I did not need to be the source of edge — I only needed to find the source of edge and route capital to it.
The first version of the system was crude. It ranked wallets by raw 30-day return. It got fooled by lucky variance in the first two weeks and I had to add a Sharpe filter, then a drawdown filter, then an edge-adjusted hit rate, then a category-concentration check. Each filter was a thing I had personally done wrong over the previous year. The wallet-scoring methodology page is the formal version; the informal version is just “everything Eli got wrong, expressed as a constraint.”
When I deployed real money against the system in late November, it worked. Not spectacularly — the first three months were up seven percent, which sounds modest until you remember that I had been losing eighteen. The shape of the P&L changed too. My solo trading had been a saw-tooth of conviction bets and forced exits. The mirrored book was smoother, smaller wins, fewer days where I cared what the market was doing.
What I would tell myself a year ago
Not much, honestly. I would not have listened. The version of me that started in late 2024 was sure that the only worthwhile way to engage with this market was to develop my own view, run my own book, and prove I belonged on the leaderboard rather than under it. There was real ego in that and ego does not get talked out of itself by paragraphs on a blog.
So instead of advice, here is the one thing I can say honestly that took me eleven months to learn. The reason copy-trading on Polymarket makes mathematical sense is not that the leaders are super-human. It is that the venue is structurally specialised, the specialists are findable on-chain in real time, and the capital cost of routing to them is small relative to the cost of competing with them.
The first version of that sentence I could have read on day one and dismissed. The version I have written here is the one I needed eleven months to be able to hear. If you are earlier in your own version of those eleven months, I cannot make you skip them. But the door is open whenever you decide you are done with that particular education.
The leaderboard is where I would have started if someone had handed me one in October 2024. The methodology page is the system the leaderboard is built on. And the data posts about edge decay and liquidity and resolution time are the patterns I now wish I had understood before I started clicking buy on my own.
None of this means do not trade Polymarket. It means be honest about what kind of trader you are, and then build the smallest workflow that gives you exposure to the edge you cannot personally generate.
That is the whole essay.