Follow the Smart Money
How we identify the sharpest traders on Polymarket, track their every move in real time, and surface high-conviction consensus signals before the market catches up.
Every prediction market has a hidden hierarchy. Most traders are noise: guessing, following headlines, or chasing narratives. But a small minority are consistently, provably sharp. They resolve more markets correctly than chance allows, their accuracy is statistically significant, and their timing suggests genuine information rather than luck.
It works like a poker table. You cannot see everyone's cards, but you can see their betting patterns and track records. Over hundreds of hands, the skilled players reveal themselves through results. On Polymarket, every trade is public, every wallet is trackable, and every market eventually resolves to a definitive outcome. That transparency is the raw material for our Oracle model.
We do not predict outcomes directly. We identify who predicts outcomes well, watch what they are doing right now, and surface the moments when multiple sharp wallets independently agree on the same position.
The Challenge
Why “Follow the Money” Is Harder Than It Sounds
Polymarket has thousands of active wallets. Watching all of them is like trying to follow every car on a highway: the sheer volume drowns out the signal. The challenge is not access to data. It is separating skill from luck, genuine conviction from noise, and sharp positioning from market-making activity.
Luck vs. Skill
A trader who goes 7-for-10 could be sharp or could be lucky. You need hundreds of resolved markets and statistical tests to tell the difference.
Market Makers
Many high-volume wallets are not taking directional bets at all. They are providing liquidity, hedging both sides, profiting from the spread. Following them would be meaningless.
Signal Decay
A wallet that was sharp six months ago may not be sharp today. Markets evolve, edges erode, and last quarter's genius can become this quarter's noise.
The opportunity: Most people following Polymarket are watching headlines and poll numbers. We are watching the wallets that have been right, consistently, across hundreds of resolved markets, and we are watching them in real time.
Finding Sharp Wallets
Separating Skill from Luck
Imagine you are a baseball scout. You do not draft a player because they hit a home run once. You watch them over an entire season. You track their batting average in different game situations. You compare their performance to league baselines. Only after hundreds of at-bats, with statistical confidence that their results are not random, do you call them a prospect.
That is how we identify sharp wallets. Every wallet on Polymarket has a public trade history. We ingest that history, compute rigorous accuracy metrics across all their resolved markets, and apply statistical significance tests to separate genuine skill from survivorship bias. A wallet must prove itself over a meaningful sample size before we trust a single trade it makes.
| Concept | Analogy |
|---|---|
| Wallet discovery | Scouting new talent: scanning leaderboards and trade networks for candidates |
| Market maker detection | Filtering out the referees: they participate but are not picking sides |
| Accuracy scoring | Batting average for predictions: measures both accuracy and calibration |
| Significance testing | The draft threshold: is this track record real or just a hot streak? |
| Time decay | Recent form matters more: a player who was great 3 years ago may not be great today |
Key insight: We do not follow wallets because they made money. We follow wallets because they are statistically, provably better at predicting outcomes than chance allows. That distinction matters enormously. Profitable wallets can get lucky. Calibrated wallets, over hundreds of markets, cannot.
Discovery Pipeline
We continuously scan Polymarket for high-performing wallets using proprietary discovery methods. The pipeline casts a wide net across multiple channels to identify skilled traders, including those who may not be publicly ranked.
Our discovery process goes beyond surface-level metrics to find wallets with genuine predictive skill across diverse market categories.
Market Maker Filtering
Not every active wallet is taking a directional bet. Market makers provide liquidity by holding both YES and NO positions, profiting from the spread rather than from predicting outcomes. Following a market maker would be like following a casino dealer's “bets.”
Our proprietary filtering system identifies and excludes non-directional wallets, ensuring only genuine predictive signals feed into the model.
The Scoring System
Finding wallets is step one. Knowing how much to trust each wallet is step two. Not all sharp wallets are equally sharp, and a wallet that dominates political markets might be mediocre at crypto predictions. We score wallets across multiple dimensions to build a nuanced trust profile.
Multi-Dimensional Trust
Imagine hiring an investment analyst. You would not just look at their total return. You would ask: How many investments did they make? Are their wins concentrated in one sector, or broad? How recent are their results? Did they beat the market, or just ride a bull market? We ask the same questions of every wallet.
Prediction quality goes beyond right or wrong: we measure how confident and how well-calibrated each call was
Statistical testing across hundreds of resolved markets to ensure skill, not luck
Time-weighted scoring: recent results carry more weight because markets evolve and edges erode
We also evaluate wallets per category. A wallet might be exceptionally sharp on political markets but average on sports or crypto. When that wallet takes a position, its influence in the consensus reflects its proven track record in that specific domain. The specialist gets a louder voice where they have demonstrated skill.
Consensus vs. Conviction
When the Sharpest Minds Agree
One sharp wallet taking a position is a data point. Multiple independent sharp wallets taking the same position on the same market is a signal. The Oracle model does not follow any single trader. It listens to the chorus and only speaks when the voices align.
It works like a jury deliberation. Each juror has examined the evidence independently. If twelve people who have never spoken to each other all reach the same conclusion, that consensus carries real weight. If they are split 7-5, confidence is lower. If they are unanimous, conviction is highest. We apply the same framework to sharp wallet positioning.
| Signal Type | What It Means |
|---|---|
| Strong consensus | Multiple sharp wallets aligned on the same side, zero opposition. Highest conviction. |
| Lean consensus | Meaningful majority aligned, but some dissent exists. Moderate conviction. |
| Split consensus | Sharp wallets disagree. No actionable signal — we pass. |
Key insight: Each wallet's vote is weighted by its track record. A wallet with a stronger track record across more resolved markets carries more influence than a wallet that just barely passed the significance threshold. The consensus is not a simple headcount. It is a weighted vote.
Edge Detection
The consensus gives us a direction: YES or NO. But direction alone is not enough. We also compute an implied probability from the sharp wallets' positions and compare it against the current market price. The gap between what the smart money believes and what the crowd is pricing is the edge.
If sharp wallets collectively imply a higher probability than what the market is pricing, that gap is our edge. We only act when this gap exceeds a meaningful threshold. Small edges are not worth the execution risk and market noise.
Quality Gates
Multi-Gate Quality Control
Like a pharmaceutical trial, a promising drug candidate must clear Phase I safety, Phase II efficacy, and Phase III large-scale testing before reaching patients. At each phase, the majority of candidates are eliminated. Our picks face the same gauntlet: most markets on Polymarket never become picks. Only those surviving every filter reach your terminal.
Gate 1: Minimum Sharp Wallets
Multiple independent sharp wallets must be aligned on the same side. A single wallet's opinion, no matter how sharp, is not enough to generate a pick.
Gate 2: Minimum Edge
The gap between the smart money implied probability and the market price must exceed a meaningful threshold. Marginal edges are rejected.
Gate 3: Consensus Ratio
The ratio of aligned to opposed wallets must show strong directional agreement. Split signals are rejected outright.
Gate 4: Market Liquidity
Markets must have sufficient liquidity and trading volume. Thin markets create execution risk and unreliable price signals.
Gate 5: Track Record Depth
The combined resolved market count across aligned wallets must be substantial. We require a deep track record, not just a hot streak.
Gate 6: Resolution Window
Only markets resolving within the near term are considered. We focus on imminent outcomes, not distant speculation.
The Daily Cycle
Two-Step Lock System
Like a newspaper with a morning and evening edition, our pipeline runs in two phases. The first pass generates candidates from the latest sharp wallet consensus. The second pass, shortly after, refreshes live market prices and drops any pick where the edge has evaporated. You only see the final, verified edition.
Morning
Signal aggregation: eligible markets scanned, sharp wallet consensus computed, quality gates applied, candidates generated
Morning + Lock
Lock step: live market prices refreshed, edge recalculated against current pricing, sub-threshold picks dropped. Final picks locked and visible
Overnight
Settlement: resolved markets checked, picks graded as correct or incorrect, performance metrics updated, wallet scores recalculated
Why two steps? Market prices move. An edge identified during signal aggregation might shrink by the time we lock, as other traders act on the same information. The lock step ensures every pick you see has a verified, current edge at the moment it was committed. No stale data. No expired opportunities.
Continuous Monitoring
Between lock cycles, the system never sleeps. Trade ingestion runs continuously, capturing new positions from every tracked wallet. Market metadata is refreshed to detect resolution events and price changes. This always-on pipeline means the morning signal is built on the freshest possible data.
After lock: Picks are permanent. Once a pick is locked, it is never modified, never retroactively adjusted. You see the same pick we committed to. Accountability is absolute.
How We Measure Success
Prediction markets resolve to binary outcomes: you are either right or wrong. But measuring model quality requires more than counting wins. Here is what we track:
Hit Rate
The simplest metric: what percentage of our picks resolved in the direction we called. Clear, intuitive, and the first thing you should check.
Calibration Score
Our calibration metric measures both accuracy and confidence. When we say 70% probability, it should resolve at that rate over many markets.
Edge at Lock
How large was the gap between our implied probability and the market price when we locked? Larger edges at lock mean we are identifying genuine mispricing, not scraping marginal noise.
Category Breakdown
Performance split by market category. Sharp wallets often specialize. We track whether our signal is strong across all domains or concentrated in specific ones.
The Bottom Line
Prediction markets are the closest thing to a real-time truth machine. But like any market, they are made by imperfect participants. The Oracle model's thesis is simple: the small minority of traders who have proven, over hundreds of resolved markets, that they possess genuine forecasting skill are the best signal available. Not pundits. Not polls. Not sentiment. Proven track records.
- Identify the sharpest wallets: through rigorous statistical testing across hundreds of resolved markets
- Track them in real time: ingesting every trade they make, across every market they touch
- Listen for consensus: acting only when multiple independent sharp wallets converge on the same position
- Verify at lock time: refreshing prices and dropping any pick where the edge has eroded
- Track everything: hit rates, calibration scores, edge at lock, and category breakdowns with full transparency
We do not predict the future. We find the people who do it best and listen when they agree.
Technical Breakdown
A deeper look at the architecture and pipeline powering the Oracle model, step by step.
Wallet Scoring Engine
Every tracked wallet is scored nightly using time-decayed metrics. Recent market resolutions carry more weight than older ones, reflecting the reality that prediction skill is not permanent. Scoring covers accuracy, profitability, and statistical significance across multiple dimensions. Wallets below the significance threshold are excluded from the signal pool.
Category-level scoring is computed in parallel. A wallet can be classified as sharp globally, sharp in specific categories (politics, crypto, sports), or both. Category sharpness gives that wallet extra influence when a market falls in their domain of expertise.
Real-Time Data Pipeline
Trade ingestion polls Polymarket's public APIs continuously for all active wallets. Market metadata (prices, volume, resolution status) is refreshed on a similar cadence. This continuous pipeline ensures that by the time pick generation runs each morning, every wallet's position is current.
The system monitors multiple Polymarket API endpoints, each providing different information. Health probes run automatically to detect any API changes or outages before they can cause silent failures.
Signal Computation
For each eligible market, the model aggregates signals from all qualified sharp wallets with active positions. Each wallet's contribution is weighted by a proprietary trust score that reflects multiple dimensions of proven skill. The result is a smart money implied probability for each direction.
The direction with the higher weighted probability becomes the pick direction, subject to strict consensus and quality requirements. The edge is the gap between what the smart money collectively believes and what the market is pricing.
Model Workflow
Continuous: Trade ingestion — capture new positions from all tracked wallets
Continuous: Market refresh — update prices, volume, and resolution status
Overnight: Settlement — grade resolved picks, update performance metrics
Overnight: Wallet scoring — recalculate all trust scores with latest resolutions
Morning: Pick generation — aggregate signals, apply gates, rank by composite score
Morning: Lock — refresh prices, verify edges, publish final picks
Periodic: Discovery — scan for new sharp wallets through multiple channels
Monthly Access
- Predictions only go live when the model finds true edge
- Closing line value tracked on every prediction so you can verify it yourself
- Covers every market we model and we're always adding more
- Cheaper than your average unit size
Annual Access
- Get 4 months free on us when you go annual
- Every new model we ship is included automatically
- Full platform access for less than most services charge monthly
- Models run 365 days, your subscription should too