Same coin.
Different price.
Live price gaps between eight exchanges, and an open record of six weeks of automated-trading research: what we tried, what it cost, and why it failed.
Loading live prices…
See every exchange and coin or read what the kimchi premium is.
What six weeks of paper trading taught us
From September to October 2026 we let an AI assistant build and run a crypto trading system under close supervision. No real money was used. These are the numbers that changed our minds.
- 2.5%
Share of 4,359 five-minute indicator setups that made money once fees and one tick of slippage were counted. At four hours it was 69.5%. Fees eat the edge
- 75% → 41%
A trend rule's yearly backtest return after we stopped choosing coins with hindsight. Survivorship bias
- 34 months
How long it takes to prove a trend strategy beats zero from live results alone. The statistics
- −3.93%
Two weeks on a server that followed its rules perfectly: five trades, five losses. Holding Bitcoin lost 0.25%. The ledger
Read the full story or browse the strategy graveyard.
Free calculators
- Can a trading bot beat fees?
Edge per trade against round-trip costs.
- Crypto profit calculator
Profit or loss after fees, and the break-even price.
- Trading fee calculator
What fees and slippage cost per month and per year.
- DCA average cost calculator
Average entry price across several purchases.
- Position size calculator
How much to buy for a fixed risk per trade.
How we work
We show where every number comes from and when it was updated. We publish failures next to results. We do not sell signals, take money from exchanges, or tell anyone what to buy.