Project Rocky

Systematic edge on the Dow Jones.

A rules-based algorithmic trading system combining quantitative backtesting, real-time execution, and disciplined risk management — built from first principles.

+84.9% RSI Return
+53.2% KC Return
70.7% RSI Win Rate
63.2% KC Win Rate

Retail trading is broken by design.

Most retail traders fail not from lack of skill but from lack of infrastructure — emotional decisions, inconsistent execution, and no systematic framework to build on.

Emotional Execution

Manual traders exit early, move stop-losses, and revenge-trade after losses. Each decision degrades an otherwise valid strategy.

📊

No Validated Edge

Most traders run strategies that have never been properly backtested. Anecdotal wins mask negative expectancy over any meaningful sample.

🎯

Inconsistent Risk

Position sizing varies with confidence rather than math. A single oversize trade can erase weeks of disciplined gains.

🔧

No Feedback Loop

Without structured journaling and attribution, the same mistakes repeat across sessions with no mechanism to learn or improve.

Research first. Then automate. Never reverse.

Every parameter in the system was validated through backtesting before being deployed live. The bot executes the rules; it cannot override them.

01
Astrophage · Team Petrova

Signal Specification

Astrophage holds the exact Pine Script strategies running on the live US30USD chart — Keltner Channel breach-and-reversal, RSI cross-50, and Opening Range Breakout. When a setup forms at bar close, Astrophage's alert fires a structured JSON webhook to the execution server within milliseconds.

02
Blip_A · Team Taumoeba

Rules-Based Execution

Blip_A receives the webhook and validates it through five gates — pre-market block, session filter, daily halt, concurrent trade cap, and minimum stop distance. If all gates pass, a GTD stop-entry order is placed on OANDA with a pre-computed 1.5× R:R take-profit and guaranteed stop-loss.

03
Xenonite · Team Taumoeba

Independent Risk Monitor

Xenonite runs in parallel, polling OANDA every 60 seconds. If the daily drawdown limit, loss count, or open position loss threshold is breached, Xenonite closes all positions and halts the bot immediately — no human reaction time required.

04
Beetle · Team Taumoeba

Automated Reporting

After the session closes, Beetle pulls the ArcLight trade report, compiles a structured daily digest with per-strategy P&L attribution, and emails it to the operator. No manual steps — the report arrives before the operator reviews the session.

05
Adrian · Team Taumoeba

Behavioural Debrief

Adrian receives the Beetle digest and audits the session against the Eridani rule set. It distinguishes infrastructure failures from genuine strategy or behavioural violations — keeping the benchmark clean and the learning loop accurate.

Two strategies. Both validated.

Independent backtest results for KC and RSI strategies. 2-minute bars, Dow Jones Futures (YM=F), 2% risk per trade. Both running concurrently on demo.

KC Strategy
Return +53.2%
Sharpe Ratio 8.56
Profit Factor 2.51
Win Rate 63.2%
Max Drawdown -7.8%
Validated Trades 38
RSI Strategy
Return +84.9%
Sharpe Ratio 22.70
Profit Factor 3.27
Win Rate 70.7%
Max Drawdown -5.9%
Validated Trades 41

Note: Backtest on YM=F 2-minute bars, 2% risk per trade. Both strategies validated independently through 25 Eridani experiments. Out-of-sample validation (Exp 7) planned for July 2026 before live capital deployment.

Milestones, not guesses.

May 2026
Research & build complete. 25 Eridani experiments. KC + RSI strategies live on demo concurrently. Full risk framework: 3-loss halt, trailing drawdown, profit protection.
Jun 2026
VPS migration complete. System running 24/7 on Hetzner Singapore. rockyproject.com live. Mac dependency eliminated.
Jul 2026
Experiment 7 — out-of-sample validation. Fresh data. KC + ATR filter, RSI 15m MTF, ORB Variant D. Final gate before live capital deployment.
2026 Q3
Live account go/no-go decision. Contingent on Exp 7 outcome and 30+ days of clean demo data. MAS regulatory compliance reviewed.