Dissecting 112,000 Polymarket Addresses: The Top 1% Who Actually Profit All Do These Five Things
TechFlow Selected TechFlow Selected
Dissecting 112,000 Polymarket Addresses: The Top 1% Who Actually Profit All Do These Five Things
Those loss-making addresses are not foolish—they simply lack discipline.
Navigating Web3 tides with focused insightsAuthor: darkzodchi
Translated by: Asher, Odaily Planet Daily
After systematically organizing and analyzing on-chain data from over 112,000 Polymarket wallets over a six-month period, a striking—and somewhat counterintuitive—result emerged: approximately 87.3% of users ultimately lost money on the platform.
This analysis covered multiple key dimensions, including every on-chain transaction record, trading volume, win rate, profit-and-loss (PnL) outcomes, market categories participated in, entry timing, and position sizes. The entire data-curation process took three weeks, and the final conclusion diverged significantly from many people’s intuition.
Many assume that top-tier players in prediction markets possess some obvious edge—such as access to insider information or use of obscure, sophisticated computational models. Yet the data tells a different story. The top 1% consistently do just a few things correctly—and repeat them relentlessly. Meanwhile, the other 99% tend to do exactly the opposite, then wonder why their capital keeps draining away.
Polymarket’s Leaderboard Is Highly Misleading
If you open Polymarket’s leaderboard today and sort by PnL, you’ll notice several anomalies. For example, the top-ranked wallet holds only 22 positions; the fourth-ranked wallet has executed just eight trades; and the eighth-ranked wallet placed only a single bet—yet still ranks among the all-time top ten.
These addresses can hardly be called “traders” in any meaningful sense. In many cases, they represent whales placing one-off, multi-million-dollar bets—e.g., $5M+ on a single event—and happening to get it right. Or they may reflect participants with informational advantages—or both. Regardless, conclusions drawn from just a handful of trades offer virtually no actionable insights for replicable trading behavior. Such outcomes resemble high-stakes “coin flips,” not reproducible strategies.
Therefore, the first step in our analysis was to filter out this noise and retain only statistically meaningful samples. Our filtering criteria included:
- At least 100 settled positions, ensuring sufficient sample size for statistical significance;
- Active trading duration of at least four months, excluding accounts that won big purely on luck;
- Participation in at least two distinct market categories, avoiding overreliance on a single event;
- Total trading volume exceeding $10,000, confirming genuine capital commitment.
Applying these filters reduced the original pool of 112,000 wallets down to roughly 8,400 addresses with sufficient data value. These 8,400 addresses constitute the truly research-worthy dataset—not the so-called “hero accounts” on the leaderboard who earned millions from just a few trades. Their shared traits are sustained trading activity and stable data profiles, making behavioral patterns far easier to observe.
Interestingly, once filtered, the most consistently profitable traders bear little resemblance to the leaderboard’s image. They’re unremarkable—most people have never heard of them. Their profits typically range between $50,000 and $500,000—not the multimillion-dollar windfalls seen elsewhere.
But what truly matters isn’t how much they earn—it’s their underlying process and methodology. Because what’s truly replicable is never the outcome, but the process.
Three Common Misconceptions to Discard
Misconception #1: Top Traders Win 80–90% of the Time
That’s false. Based on our filtered dataset—not the outlier whale accounts dominating the leaderboard—the win rates of genuinely long-term profitable wallets cluster between 55% and 67%. Even elite traders are wrong on a substantial share of trades. For instance, one wallet has settled over 900 positions and netted $2.6M—but its win rate stands at just 63%. In other words, more than one-third of its bets were incorrect, yet it still generated massive returns in prediction markets.
An obsession with win rate is often the rookie’s biggest trap. Many beginners buy contracts priced at $0.90 because they “feel safe”—the YES probability appears to be 90%, seemingly near-certain. So they pay $0.90, earning just $0.10 if correct. But one misjudgment costs the full $0.90—a risk-reward ratio of 9:1. Repeat this pattern enough times, and account equity evaporates rapidly. This exact dynamic has repeated across hundreds of addresses in our dataset.
Misconception #2: The Best Traders Participate in Every Market
Reality is precisely the opposite. Top-performing wallets typically engage in no more than three market categories—and most focus on just one or two. Some trade exclusively in crypto-related events; others specialize solely in weather markets; one wallet even trades almost exclusively on questions like “Will Bitcoin hit X price before Friday?”
In prediction markets, excessive diversification usually signals declining judgment quality. Generalists underperform; specialists consistently profit.
Misconception #3: Speed Is Everything
This holds true only in rare cases—e.g., certain 15-minute-settling crypto markets requiring rapid response. But in most markets, elite traders don’t win by speed. Instead, they commonly build positions gradually over days—or even weeks. They don’t race to click first; they patiently wait for clear pricing deviations. When prices drift far enough, even if correction takes two weeks, the overall mathematical expectation remains favorable.
Five Trade Patterns Worth Emulating
Pattern #1: Contrarian Trading During Extreme Sentiment
Across the entire dataset, this is the most visible and stable profitability signal. Among the 8,400 filtered wallets, this behavior is nearly the strongest predictor of long-term profitability.
When a contract surges to 88% due to market sentiment, many top wallets begin selling YES; when it plunges to ~12%, they start buying incrementally. This isn’t blind contrarianism—they aren’t opposing the market for opposition’s sake. They enter only when they judge sentiment to be clearly overextended.
This strategy works because of a well-documented phenomenon known as the “favorite–longshot bias,” first observed in 1940s horse-racing betting studies—and replicated across virtually all human-driven betting markets. Simply put, people systematically overestimate outcomes that “look almost certain,” while underestimating low-probability events.
Further analysis revealed that the top 50 most profitable wallets entered trades, on average, at prices deviating 6–11% from consensus probabilities. They avoid 50/50 situations entirely, waiting instead for odds clearly tilted in their favor. This approach may seem dull—but long-term data shows it’s remarkably stable and profitable.
Pattern #2: Position Sizing Closely Mirrors the Kelly Criterion
Comparing position sizes of the top 200 profitable wallets against their perceived “implied edge” reveals a strong correlation. In other words, they don’t bet randomly—their position size scales almost proportionally with their assessed edge: larger positions when the edge is large, smaller positions when it’s modest, and zero exposure when no clear edge exists.
Whether these traders have actually read the Kelly Criterion—or simply internalized it through years of trial, error, and loss—is unclear. But mathematically, their behavior closely approximates it.
The Kelly Criterion is typically expressed as: f* = (p × b − q) / b, where: p = trader’s estimated true probability of the event occurring; q = 1 − p; b = net odds received on the bet (potential gain ÷ risked amount).
Simple example: A trader estimates a 60% chance of an event occurring, while the market price is $0.45. Net odds: b = (1 / 0.45) − 1 ≈ 1.22. Plugging into the formula: f* = (0.60 × 1.22 − 0.40) / 1.22 ≈ 0.272. Full Kelly suggests allocating 27% of capital to this trade.
But full-Kelly sizing carries extreme risk—volatility would likely trigger massive drawdowns in short order. Real-world data shows profitable wallets adopt a far more conservative version: roughly one-quarter Kelly. So if full Kelly recommends 27%, they’d typically allocate ~7%.
For highest-conviction opportunities, positions may rise to 12–15%; medium-conviction trades usually see 2–5% allocation; and markets offering no clear edge are simply skipped. By contrast, losing accounts fall into two extremes: either risking 80% of capital on a single trade—purely gambling—or spreading $10 across 40–50 markets, mistaking dispersion for risk management. In reality, this just pays fees repeatedly—keeping the account busy, not profitable.
Pattern #3: Hyper-Focused Specialization
Categorizing the 112,000 wallets by market type reveals stark differences. Categories include crypto, politics, sports, weather, geopolitics, entertainment, and science. Key findings:
- Wallets participating in only 1–2 categories averaged +$4,200 PnL;
- Those in 3–4 categories averaged −$380 PnL;
- Those in 5+ categories averaged −$2,100 PnL.
This relationship follows an almost perfectly linear trend: the more categories traded, the higher the loss probability.
Different market types rely on entirely distinct information ecosystems. Crypto markets respond to exchange order-book flows, whale addresses, funding rates; political markets hinge on polling data, grassroots intel, congressional calendars; weather markets depend on NOAA models, atmospheric data, and satellite observations.
Two cases illustrate this vividly. Case 1: Wallet A trades exclusively in Polymarket’s 15-minute-settling Bitcoin markets—e.g., “Will BTC exceed X price within the next 15 minutes?”—and never touches any other category. It completed 502 predictions, achieving a 98% win rate and $54,000 in cumulative profit. Its edge is simple: continuously monitoring Binance’s order-book depth, then executing trades when Polymarket lags by 10–30 seconds. That tiny informational edge—just seconds—was exploited hundreds of times.
Case 2: Wallet B trades only weather markets. Its strategy is equally direct: reading NOAA’s daily publicly released temperature forecasts, comparing them to Polymarket’s pricing, and entering trades whenever market prices diverge meaningfully from these decades-optimized supercomputer projections. In New York temperature-prediction markets alone, its accuracy reached 94%.
Crucially, these individuals aren’t geniuses. The real key is that they identified a narrow domain where they understand more than the average Polymarket participant—and then leveraged that advantage repeatedly. No constant strategy shifts. No FOMO-driven chases after trending topics. Just consistent execution of the same logic, around the same edge.
Pattern #4: Trading Price Dislocations, Not Event Outcomes
Most Polymarket users trade simply: buy a contract and hold until settlement—then collect profit or absorb loss. A classic binary result. Top wallets operate differently. Often, they buy at $0.40, then sell at $0.65 when news or sentiment pushes price up—regardless of whether the event ultimately occurs. They exit as soon as new information is priced in.
In our dataset, some top-performing wallets have zero settled positions. They never hold contracts to expiration—instead, they continuously trade mispricings as short-term swings. Data shows top wallets’ average holding time is just 18–72 hours, whereas wallets in the bottom 50% by profitability typically hold until settlement—sometimes for weeks or months.
This doesn’t mean holding to settlement is inherently wrong. When conviction is extremely high, long-holding can indeed be optimal. But overall, top wallets deploy capital far more actively and flexibly than most imagine. They’re not passive bettors—they’re genuine traders.
Pattern #5: Always Avoid Breaking News
Intuition says the sharpest capital should rush in at the first sign of breaking news—military conflicts, election results, CEO resignations. Yet data shows top wallets deliberately avoid the immediate post-news window. They let emotional capital flood the market first, causing sharp short-term volatility, then begin trading only after sentiment stabilizes.
Across the full dataset, a clear pattern emerges: the best trading opportunities arise either *before* the market notices an event—or *after* market sentiment has overreacted. When everyone’s talking about the same thing, it’s often the worst possible entry point: prices are already highly efficient, leaving minimal exploitable edge.
Five Actionable Recommendations
Pick One Niche—and Stick With It Long-Term
Crypto, politics, weather, sports—it doesn’t matter which. Choose the domain you know best. Then trade *only* that category for at least three months. No exceptions. No impulsive side-bets on trending elections. Even “just a quick wager” risks shattering your carefully calibrated decision framework.
Log Every Prediction
Before each trade, record several key metrics: your estimated true probability, the current market price, your expected edge, and your planned position size. After accumulating 50+ trades, review the data. For instance, if you labeled certain predictions as “70% likely,” did they actually hit ~70%? Significant deviation means your probability calibration is off—and must be corrected *before* scaling position sizes.
Size Positions Close to One-Quarter Kelly
Calculate the theoretical Kelly position size, then divide by four for your actual allocation. This number will often seem small—but that’s precisely how risk is controlled. Over-leveraging leads inevitably to blowup.
Trade Only When Your Edge Is Clearly Meaningful
If your expected edge falls below 8–10%, skip it—no matter how tempting. The top-performing wallets in our data typically execute only 2–3 trades per week *per market category*. Quality always trumps quantity.
Maintain Rigorous Logging and Review
Build a complete trade log—recording every trade, outcome, and lessons learned. Wallets showing sustained performance improvement almost universally conduct systematic post-mortems on their mistakes. Those stuck in place—or perpetually losing—tend to repeat identical errors, blaming bad luck rather than flawed process.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@zodchiii
