The $1.7M Google Search Arbitrage Nobody's Talking About
We built a probabilistic engine on two decades of Google "Year in Search". 10,000 Monte Carlo runs. 47 input features. Several outcomes look mispriced by 2–3×. This is math vs. mob psychology — and it creates tradable spread.
Monte Carlo: 10,000 sims
47 input signals
200+ event-decay fits
STATIC VIEW
Market vs Model — Probability Comparison
Arbitrage — Undervalued in Green
Market vs Model — Scatter (bubble size = volume, $K)
$ cat opportunities.txt
Methodology
- Data: Google Year-in-Search 2004–2024.
- 47 inputs incl. event timing, decay, cross-media lift.
- Event half-life curves fit on 200+ major global events.
- 10,000 Monte Carlo runs per candidate with rank aggregation.
- Output: fair probability vs market, conviction score.
NFA. DYOR.
Real Examples (Actionable)
Taylor Swift — Long
Super Bowl visibility + Q4 tour finale + high odds of new music/relationship catalyst → sustained Q4 searches.
Market 15%Model 48%
Donald Trump — Long
Inauguration in Jan + quarterly controversy cadence. Miss requires near-zero noise (unlikely).
Market 44%Model 70%
Pope Leo XIV — Long
Election timing (May) → 8-month attention arc of tours, speeches, retrospectives.
Market 82%Model 92%
Bianca Censori — Short
Tabloid ceiling. Even Kim K never hit Top 5. Model fair far below hype.
Market 65%Model 18%
What is PolySearch.fun?
A quant-style layer for prediction markets. We ingest search and event data, build probability models, and surface mispricings you can trade.
Where does the data come from?
Google Year-in-Search archives, event calendars, historical decay patterns, and public market prices.
Is this financial advice?
No. It’s research. Use your judgment. Markets are risky.
Can I plug in my own priors?
Planned. We’re building sliders for event timing/volatility to sandbox custom scenarios with live spreads.