hi, i'm Xin.

i build tools for reading prediction markets and publish what i learn from the data.

a small experiment

Seed 2040

i flip 120 fair coins, 250 times each: one step up for heads, one down for tails. nobody has an edge.

Open the experiment

pick the best after the fact and luck looks like skill. a backtest that tries many variants can do the same.

Saved result of 120 random walks. The best walk ends 46 steps above zero.
Synthetic data: a fair coin, no market data.
Best
+46
Mean
−0.22
Worst
−50
Above zero
61 / 120

a thing i made

PolymarketTerminal

  • TypeScript
  • MIT licence
  • runs on your machine, no telemetry
See it on GitHub

i made a read-only terminal screener for Polymarket, with a desktop order-book terminal beside it.

the radar ranks markets by money from accounts under 14 days old, and it never places orders.

the desktop terminal can place orders only after you connect a key and arm it, and it starts in dry mode.

a research tool, not advice, and not affiliated with Polymarket.

Two order books side by side with price ladders, resting orders and executed trades, with the market navigator on the left.
Two price ladders for the same market, shown in dry mode with demo data.

i tested whether following sudden volume surges paid off across 2,400 resolved markets. the strategies were inconclusive on the held-out markets. Read the volume-surge study

say hi.

if you’re looking at similar data or have a question about the terminal, message me on telegram.