A self-hosted Python application called TradeSight is drawing attention among algorithmic-trading enthusiasts for running AI-driven strategy tournaments overnight, backtesting technical indicators, and executing simulated trades through the Alpaca brokerage API — all from a locally hosted dashboard.
Self-Hosted AI Trading Lab · TradeSight
A local AI strategy tournament won the backtests — then lost real money in a 30-day paper run
TradeSight runs overnight strategy competitions on your own machine — no cloud, no subscription. But a hands-on paper-trading test shows a tournament winner is not a live winner.
15+
technical indicators — MACD, RSI, Bollinger, EMA, ATR, volume
282/282
passing tests under pytest
$0 / $49
free MIT source vs one-time packaged build
The tournament vs. the outcome
Backtest tournament scores looked healthy — but the live 30-day paper account slipped into the red.
TOURNAMENT SCORE
0.62 – 0.72
RSI Mean Reversion &
→
LIVE PAPER RESULT
−$54.55
−0.55% on a $10,000
How the strategy tournament works
Parameter variants
strategies spun up with many configs
→
Head-to-head
overnight tournaments after market hours
→
Winners evolve
weak configs eliminated, strong ones advance
What it does well
Runs 100% locally — no cloud, no monthly fees
Flask dashboard at localhost:5000, cron-scheduled runs
Tournament quickly weeds out clearly weak strategies
Demo mode works without Alpaca API keys
The caveats
A tournament winner is not a live winner
Needs stop-losses and circuit breakers
No live real-money trading; backtests aren't predictions
Early-stage, limited large-scale validation
What's next — v2
Planned additions include a crypto scanner and push alerts , with the packaged price rising from $49 to $79 . Requires Python 3.11+, runs on macOS or Linux (Windows via WSL).
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…