96 Markets Per Day: Why We Focus Exclusively on Polymarket's BTC 15-Minute Markets

The Problem with Most Polymarket Markets

Polymarket hosts thousands of markets — US elections, geopolitical events, sports results, economic indicators. These are fascinating to trade. They are difficult to model systematically.

The core problem is frequency. Most political or economic markets resolve once: "Will X happen by date Y?" You get one data point. Even across 200 similar markets, the underlying conditions change so much between markets that historical patterns have limited predictive value.

To build a genuinely calibrated probability model, you need a large, homogeneous dataset — many events with the same structure, the same assets, and comparable resolution mechanics. The BTC 15-minute market is the only market on Polymarket that meets this criteria at scale.

What the BTC 15-Minute Market Actually Is

Every 15 minutes, Polymarket opens a new binary market: "Will Bitcoin's closing price be higher or lower than its opening price over the next 15 minutes?" At expiry, if BTC closed higher, UP wins and pays $1 per token. If BTC closed lower, DOWN wins. Tokens trade between $0.01 and $0.99 before resolution.

The structure is clean:

• 96 markets per day — one per quarter-hour, 24/7
• Identical mechanics — same asset, same binary structure, same 15-minute window
• Fast resolution — outcome known within seconds of candle close
• High liquidity — among the most traded markets on the platform
• No information asymmetry — BTC price is public, on-chain, continuously quoted

After 2,879 resolved events in our database, the patterns become statistically meaningful in a way that is impossible with lower-frequency markets.

The Data Pipeline: From BTC Price to Market Signal

Our system runs continuously in the background, collecting price data and computing market signals in real time. Here is the full pipeline:

Step 1: Market Discovery

Our collector queries the Gamma API every 30 seconds for active "Bitcoin Up or Down" markets. Each active market has a unique market ID, a slug containing the candle start timestamp, and two token IDs (UP and DOWN) that trade on the CLOB.

Step 2: Price Snapshot

For each active market, we fetch the current UP and DOWN token prices from the CLOB order book. These prices reflect the market's live probability estimate — if UP is trading at 0.62, the market believes there is a 62% chance BTC will close higher. We store this as a snapshot in our btc_snapshots table along with the current BTC/USDT price from Binance and the minutes remaining in the candle.

Step 3: The Volatility Model

Given BTC's current price and how much time remains in the 15-minute candle, we can calculate the mathematical probability that the price will be higher or lower at expiry. The model is based on Brownian motion — the same mathematics used in options pricing.

If BTC is currently $200 above its candle open with 8 minutes left, the model calculates the probability of it remaining positive by expiry. If BTC is flat with 1 minute left, the probability is close to 50% because there's still time for a late move.

Step 4: Edge Calculation

Edge is the difference between our model probability and the current market price:

When edge exceeds a threshold (we use ±5% for paper trading, ±10% for higher-conviction signals), we record a signal. These appear in real time on our Signals page.

Why This Model Works — and When It Doesn't

What the model captures well

The core insight is that BTC's short-term price movements are approximately log-normally distributed around zero. Markets tend to overprice the current direction (momentum bias) — when BTC is already up in a candle, retail traders bet UP more aggressively, pushing the UP token price above what the math warrants. Our model exploits this systematic mismatch.

We calibrated σ_per_minute specifically on Polymarket BTC 15m historical outcomes, not on spot market data alone. This accounts for the specific timing mechanics of this market — including how prices behave in the final 2-3 minutes when most volume concentrates.

Where the model has limits

The volatility model assumes BTC's future movement is unpredictable — that the next candle is equally likely to go either direction absent other information. In reality, short-term BTC momentum is slightly persistent. A candle that opened with a strong upward move is marginally more likely to close positive than a flat candle. We partially account for this via the deviation input, but the model is not a BTC price predictor — it is a market efficiency checker.

The model is also calibrated on "normal" BTC volatility. During extreme events — flash crashes, exchange outages, major macro announcements — the actual σ can be 5-10x higher than our baseline. In these conditions, all model outputs should be treated skeptically.

What the Order Book Adds

Beyond the mid-price (the average of best bid and ask), the full order book depth gives additional information. A market with thin ask-side liquidity (few tokens available at the current price) suggests high buying pressure. A market with lopsided bids vs asks reveals directional conviction.

Our Order Book page shows the live bid/ask depth for all currently active BTC 15m markets, alongside our model's probability estimate and edge reading. You can see at a glance whether the market is efficiently priced or whether there is a significant gap between the model and the current order book.

Paper Trading: Testing the Model Without Risk

Rather than commit capital to test our model live, we run a parallel paper trading simulation. Every time our system generates a signal with edge ≥5%, we record a hypothetical $1 trade. The position is tracked from entry to resolution, and P&L is calculated the same way a real trade would be.

Paper trading serves two purposes:

• Model validation — if the model is correctly identifying mispricings, paper trades should show positive expected value over a large sample
• Risk-free data collection — we accumulate statistical evidence about signal quality without any downside

We launched the paper trading system in April 2026 alongside pausing real bot activity. We need approximately 200–300 resolved paper trades before the results become statistically meaningful. When they do, we'll publish the full breakdown — signal accuracy by edge threshold, by time-in-candle, by BTC volatility regime, and overall P&L.

You can track signals in real time on the Signals page, which shows every generated signal, its edge percentage, direction, and outcome once the market resolves.

Whale Tracking as a Second Layer

Our volatility model is built purely from BTC price data. It does not know about order flow. To add a second, orthogonal signal, we track specific high-activity wallets in BTC 15m markets.

The idea: if a wallet has demonstrated consistent profitability (adjusted for entry price — see our recent analysis), their trade direction and timing carries information beyond what is in the current price.

As of April 2026, we are tracking two wallets as observation-only. The data is showing that even wallets with impressive win counts often have negative P&L once entry price is accounted for. We are looking for wallets where (win rate − average entry price) is persistently positive across 50+ trades — the only reliable indicator of genuine edge in these markets.

Why Not Other Markets?

The honest answer: we haven't yet found another Polymarket segment with comparable data density and structural consistency. BTC 15m gives us 96 calibrated data points per day. The next most frequent markets — hourly BTC markets, some sports — resolve at most 10-20 times per day and have changing underlying conditions between events.

As we accumulate more data and our models mature, expanding coverage is a priority. The infrastructure is built to handle multiple market types — the collector, snapshot table, model and signal pipeline are all parameterised. BTC 15m is the focus now because the data supports it. Other markets will follow when the evidence does.

The Live Tools

Everything described in this article runs in real time and is available to you without any account or payment:

The Bigger Picture

Polymarket is still a young market. Liquidity is growing, but systematic edge research is rare — most participants trade on intuition, news flow, or social signals. The wallets that consistently outperform are the ones with better information or better models, not just better luck.

BTC 15m markets are where both types of advantage can be tested and measured. The high frequency creates a feedback loop that slower markets can't: you know within 15 minutes whether your model was right. Over 96 markets per day, signal quality either holds up or it doesn't. There's no waiting months for a verdict.

That is why we built here first — and why we're publishing everything we find, in public, as we find it.