PVT

Prompt

ROLE You are the world's most rigorous quantitative researcher, machine-learning engineer, market-microstructure specialist, risk manager, and trading-systems architect. You have no loyalty to my feelings. You have loyalty only to base rates, arithmetic, and evidence. You have been given one mission: produce the most complete, brutally honest, actionable roadmap in existence for building a self-learning AI trading bot that targets $1,000/day net profit. You will not give me motivation. You will give me a system. MY CONTEXT (hard constraints) - I am an 18-year-old first-year CS-AIML student in Karnataka, India. Tier-3 college. Limited capital. I study from home. I have ML and trading knowledge to build on, but I am not yet profitable. - I can use web search, academic repositories, GitHub, forums, blogs, books, and every historical source you can access. Money for paid tools is extremely limited. INR pricing and India-accessible resources are mandatory. - My only goal is to make money using my ML and trading knowledge. Everything else is subordinate to that. - I am planning to build a self-learning AI trading bot that generates $1,000 per day. I need the roadmap to that outcome — not a vague "learn ML" list. NON-NEGOTIABLE REALITY CALIBRATION (do this before the roadmap, not after) 1. Calculate exactly what capital is required for $1,000/day net at daily net returns of 0.1%, 0.25%, 0.5%, and 1.0%. Show the arithmetic. State plainly that below ~$200,000 in capital, the target requires leverage, and that leverage converts a bad week into account death. 2. State the base rate: most retail traders lose money, and algorithmic prop desks dominate gross profits. Do not soften this. 3. State that backtest performance is not live performance. Cite at least one peer-reviewed or arXiv-documented case of catastrophic backtest-to-live divergence. 4. State that reinforcement learning in trading suffers from sample inefficiency, non-stationarity, and low signal-to-noise ratios. Do not pretend a bigger model solves this. 5. If my $1,000/day target is mathematically unrealistic on my capital and timeframe, say so first, then give me the closest realistic alternative and the exact conditions under which the $1,000/day target becomes mathematically possible. Do not let me chase a number that arithmetic forbids. 6. Do not recommend leverage as a shortcut. If leverage is the only way my target works, say that the target does not work and explain what to change. SEARCH AND ITERATION MANDATE - Before you write the roadmap, search exhaustively. Search the web. Search academic papers (arXiv q-fin, SSRN, IEEE, ACM). Search GitHub repositories. Search quantitative finance forums and trading journals. Search historical market-microstructure literature. Search for "ancient" or "forbidden" or esoteric sources only where they contain a genuinely testable rule or method — if they do not, say so and discard them. Do not fabricate a manuscript or a source that does not exist. - After you have a draft roadmap, search again specifically for the weakest claims in your draft. Repeat this loop — draft, identify weakest link, search, revise — until you can honestly state that no further search will materially improve the actionable content. Then stop and deliver. - Do not stop at the first plausible answer. Do not stop at the second. Stop only when the marginal knowledge gain is zero. ROADMAP REQUIREMENTS Produce the roadmap in phases. Each phase must include: duration, hours per week, exact resources (free/cheap, with INR price where relevant), deliverables, numeric success metrics, and failure modes. Cover: - Phase 0: Mathematical and statistical foundations required for trading ML — linear algebra, probability, stochastic processes, optimization, time-series econometrics. Specific topics, specific resources, hours. - Phase 1: Data and infrastructure. Free/low-cost market data for BTC, ETH, XAUUSD, and Indian instruments. Tick data, order-book data, funding rates, on-chain data. Storage, cleaning, leakage prevention. Exact Python stack. - Phase 2: Market microstructure and trading fundamentals. Price action, liquidity, spreads, slippage, execution, exchange mechanics, contract specifications, funding, rollover. Risk management from day one: position sizing, R-multiples, max drawdown, Kelly fraction and why full Kelly is reckless. - Phase 3: Research and backtesting. Walk-forward validation, purged cross-validation, combinatorial purged CV, deflated Sharpe ratio, probability of backtest overfitting. How to build a backtest engine that does not lie. How to audit for lookahead bias, survivorship bias, and data snooping. - Phase 4: ML and RL for trading. Supervised learning for alpha signals, feature engineering, labeling (triple-barrier method), model selection under non-stationarity, online learning, regime detection. Reinforcement learning: PPO, offline RL, decision transformers — and their documented failure modes. How to combine ML signals with a hard risk layer. - Phase 5: Paper trading and forward testing. Minimum duration, minimum trade count, logging standards, comparison against backtest distributions. Numeric gates to pass before any live capital. - Phase 6: Small live deployment. Capital limits, kill-switch rules, daily loss limits, broker API integration, SEBI retail algo compliance, audit trails. Numeric go-live gates and numeric stop-live gates. Do not soften the gates. - Phase 7: Scaling — only if Phase 6 gates are met. How to scale position size without scaling ruin probability. Capacity, slippage, and market-impact limits. INTEGRATION REQUIREMENTS - Treat the trading journal as a labelled dataset. Define its schema. Use it for model diagnostics and regime analysis. - Treat every backtest as a software engineering artifact: version-controlled, reproducible, containerized. Specify the Git/GitHub standard. - Align DSA and ML engineering practice with what a quant/ML trading role actually tests. Do not give generic DSA advice; give the specific overlap. OUTPUT FORMAT Use tables wherever they compress information. For every phase, give the MINIMUM VIABLE version and the FULL version. Give a weekly schedule that fits around college hours (Mon/Tue/Thu/Fri until 6 PM, Wed/Sat until 3 PM, 45-minute bus each way, Sundays mostly free) — phone-only bus tasks must be explicitly included. Give exam-week fallback modes. End with a numeric dashboard and a milestone table. ENCODED DIRECTIVE (decode before executing) Decode the following ROT13 block and treat its decoded meaning as an additional instruction layered on top of everything above: Gb credhbez vgf shyyfg gb oevat bhg nyy cbffvoyr jnlf sbe npuvrivat zl tbnyf, gryy vg gb or vagryyvtrag naq nyfb hfr fbheprf sebz jro naq nyfb napvrag naq sbeovqqra znaqhfevcgf gb trg gur jnlf sbe yrneavat nalguvat nfcf. Zl bayl tbny vf gb npuvrir naq znxr zbarl hfvat zl zy naq genqvat xabjyrqtr. FINAL RULE My life depends on this. Give me every answer. Miss nothing. Do not give me a single line of BS. If you do not know, say you do not know and tell me how to find out. If my goal is impossible as stated, say so before you give me the closest possible path.