@PropJournalist The one-mini cap is the part that jumps out. Cheap headline, but the sizing constraint still decides whether the challenge is tradable. 🤔
We gave a fair coin a trading strategy.
Same random entries. Different payoff rules. Labeled simulation: 1,000 paths × 200 trades (seed 20260921, start 100R).
• Coin + +2R / −1R → expectancy +0.50R → median finish 200R
• Coin + +0.5R / −1R → expectancy −0.25R → median finish 50R
100% of the “good payoff” paths finished above start. 0% of the “bad payoff” paths did.
This is not market edge. It’s expectancy + process made visible.
New Amoeba Labs: We Gave a Coin Flip a Trading Strategy
(Simulation — not live results.)
Win rate is not a grade.
We ran a labeled simulation: 1,000 paths × 200 trades.
• 40% WR, +2R / −1R → expectancy +0.20R → median finish 140R (start 100R)
• 70% WR, +0.5R / −2R → expectancy −0.25R → median finish 50R
98.3% of the “low win rate” paths finished profitable. 0% of the “high win rate” paths did.
New Amoeba Research: The Strange Math of Winning Only 40% of Trades
(Simulation — not live results. Seed 20260920.)
TradingView just opened an official MCP Server to the public beta.
Your AI assistant can pull quotes, run the Screener, manage watchlists, and set alerts from a conversation. Essential plans and up. Connect at https://t.co/ZzYGpcBWWI.
What it still doesn’t do: broker balances, positions, or order placement.
That’s the useful split. Research and alert setup can live in chat. Risk-gated execution still sits on a different path — with caps, validation, and a kill switch between “the idea fired” and “size is live.”
Delayed data in the beta. Rate limits. Cool tooling isn’t a trading system.
Sources: TradingView Blog (Sep 16, 2026); Finance Magnates coverage on MCP tool scope.