TraderZ
A paper stock-trading arena. Seven trading strategies compete on live market data from one Alpaca paper account, and an AI agent helps tune them. No real money.
Open TraderZ Sign in with Google to watch the arena.
Live results
Loading…
Trades
- Today
- –
- This week
- –
- This month
- –
Profit and loss paper
- Today
- –
- Week
- –
- Month
- –
- Year
- –
Strategies, last 30 days
| Strategy | Trades | Win rate | P&L |
|---|
Live numbers are unavailable right now. TraderZ may be restarting; try again in a minute.
The arena
Seven strategies share one $100k Alpaca paper account. They trade the same universe of stocks, first come first served per symbol, and results are compared per strategy.
- swing_prepBreakout on confirmed strength.Benched
- swing_pullbackBuys the first pullback to the 20-day average.Benched
- meanfieldMean reversion: buys fear in healthy uptrends.Benched
- breaklineVolatility squeeze under resistance.Benched
- sentinelBuys inverse ETFs, only when the market is risk-off.Trading
- t800Reddit and StockTwits attention, naive momentum. The control.Benched
- t1000Reddit and StockTwits attention, evidence-based: buys after the hype cools.Trading
Right now only sentinel and t1000 are trading. The other five are benched for Season 3.
How a trade happens
Universe
A watchlist plus Alpaca's daily top movers and most-active stocks. Scans run at 10:00 and 16:15 ET; a social-attention loop runs every 30 minutes.
Score
Each strategy scores a stock from 0 to 100 using a year of daily bars. An entry needs the score to clear that strategy's threshold.
Gate
Entries are gated by market regime (SPY against its 200-day average, VIX, high-yield credit spreads), FOMC and CPI blackout days, and earnings blackouts.
Enter
During market hours the engine checks every 60 seconds.
Exit
Stops, targets, trailing-stop ratchets, and score decay.
Tuning, with a human in the loop
Parameters in a database
Each strategy's parameters live in a database, so they can be tuned without changing the strategy code.
Backtested first
Changes are checked with a backtester in the same repo: two-year runs over a 127-name universe.
Proposed by an AI, approved by a person
After the market closes, a scheduled AI "VP" agent proposes changes. A human approves each one before it takes effect.
Seasons
Season 2 archived
July to October 2026, 78 trades. Closed down about $4.4k (paper).
Season 3 running
Started October 28, 2026 on a fresh $100k paper account, with two strategies trading and five on the bench.
Inside the app
Watch the arena
Sign in with Google to see open positions, recent trades and how each strategy is doing.
Evaluate a ticker
Type any symbol and see how each strategy scores it. Viewing only: visitors can't place trades.
Not investment advice; paper trading only.