I Saw a Twitter Thread About an Options Trader Who Cracked Polymarket — So I Built the Tool
I was scrolling Twitter a few weeks ago and came across a thread that stopped me cold.
Someone posted a story about an options trader — desk closed after 8 years pricing binary options. HR calls on a Tuesday. Eight years of knowing exactly what binary contracts are worth, and now he's got no place to put that knowledge to work. A friend sends him a Polymarket link. He opens the order book and sees something that I imagine felt like finding a room full of people playing poker and not realizing there's a marked deck on the table.
Nobody in this market prices time.
A 65¢ contract expiring in 2 hours. A 65¢ contract expiring in 2 weeks. Same price. The crowd treats them like they're the same product.
On a real desk, these aren't even close to the same thing.
What He Saw
Binary options have a concept called theta — time decay. As a contract approaches expiry, its price should behave differently depending on where it sits relative to 50. A contract above 50¢ that's running out of time should be rising toward certainty. A contract below 50¢ running out of time should be collapsing toward zero. Time is running out for the probability to change.
The crowd doesn't think about this. They price probability. Not time.
Polymarket regularly had contracts at 80–85¢ that an options model would price at 72–74¢. That's a 10-cent gap. On $50K of weekly volume, that's real money, every week. The trader ran 708 trades in three months. He made $131,094.
"On the desk we called these counterparties uninformed flow. Here the whole market is uninformed flow."
I'm not claiming I know this trader personally or that this is some exclusive insight someone whispered to me. I read the thread, same as you could. But I came from the bond desk. I understand exactly what he saw. And I thought: I can build this.
The Math Behind the Edge
Here is the model, clean and precise.
A prediction market contract is a cash-or-nothing binary option. It pays $1 if the event resolves YES, $0 if NO. The current price is the market's probability estimate.
On a proper options exchange, the price of a binary at time τ before expiry is not just the raw probability. It's the probability adjusted for how much the underlying can still move before the clock runs out. The formula:
FV = Φ( logit(p) / (σ × √τ) )
Where:
- p = current market price as a decimal (0.82 for 82¢)
- logit(p) = ln(p / (1 − p)) — converts probability to log-odds scale
- σ = daily log-odds volatility of the market (how much the probability moves per day)
- τ = days until resolution
- Φ = the standard normal CDF (the bell curve)
The formula has clean properties. At τ = 0 (expiry), the fair value collapses to 0 or 1 — it resolves binary. At τ = ∞ (infinite time remaining), fair value converges to 0.5 — pure uncertainty. At p = 0.5, fair value stays at 0.5 for any time horizon — the model is symmetric. At p > 0.5 with decreasing τ, fair value rises toward 1. At p < 0.5 with decreasing τ, fair value falls toward 0.
That is the theta of a binary option. The crowd doesn't price it.
Calibrating Sigma
The σ parameter — daily log-odds volatility — is where the precision lives. I calibrated it from the thread's example.
The trader described a contract at 82¢ with 14 days remaining that his model priced at 72¢. Working backwards from that:
logit(0.82) = ln(0.82/0.18) = 1.5163
Φ⁻¹(0.72) = 0.5828
Solving: 1.5163 / (σ × √14) = 0.5828 → σ ≈ 0.70
At p = 0.5, a daily log-odds volatility of 0.70 equals roughly 17 probability points per day of movement. That's consistent with active political prediction markets during a live event cycle. Lower-volatility markets — stable economic forecasts, slow-moving policy decisions — run closer to σ = 0.20 to 0.40.
The tool lets you set this. If you are trading a slowly evolving market, dial it down. If you're in something fast-moving, dial it up. The 82¢ → 72¢ example is the "Active" preset.
What the Signal Means
SELL is when the market price is higher than the binary fair value. The crowd has priced in certainty the contract hasn't earned yet. This is most common with high-probability contracts that still have days or weeks left. Someone sees 82¢ and thinks "this is basically done." The binary model says there are still 14 days of σ=0.70 volatility in play. The fair value is 72¢. The 10¢ difference is the edge.
BUY is when the market price is lower than the binary fair value. The contract hasn't decayed enough. You're getting a discount on the certainty that's already embedded in the time decay. This happens more with high-probability contracts that the crowd hasn't pushed up far enough relative to how little time is left.
The decay curve in the tool shows you both. For an 82¢ contract with σ=0.70, the binary fair value at 30 days is 59¢, at 14 days is 72¢, at 7 days is 79¢, at 1 day is 98¢. The market holding the contract at 82¢ across all those time horizons is leaving money on the table — or giving it to you, depending on which side you're on.
The Second Layer: Gamma
The trader also described a second signal. When someone found out a result before the market, liquidity in the order book would start drying up before the price moved. The bid-ask spread widens. Depth thins. This is the gamma signal — informed flow entering the gap.
He would enter that gap. The book was showing him what the price hadn't moved to show yet.
This is harder to model without live order book data, but the pattern is identifiable: spread widening + decreasing depth + short time to resolution = potential informed trade incoming. I'm working on integrating this into the Theta Edge scanner as a secondary signal alongside the time-value calculation.
The Tool
I built it. It's called Theta Edge and it lives in the Pro suite.
You enter the market price, days to resolution, and volatility category. It outputs the binary fair value, the edge in cents, and the signal — BUY, SELL, or FAIR — with a confidence level. The decay curve shows you how the fair value evolves over the remaining time horizon so you can see where you're standing on the curve, not just what the point estimate says.
The formula is exact. The math is the same model a desk would use. The calibration is built from the example that came from someone who actually traded it for real money.
The crowd prices probability. Not time.
Find the gap. Trade the gap.