📘 Playbook — how to run the Trading Bot

• Alerts Only — bot signals but places no orders; you get Telegram pings instead. The safest mode.
• Disable Claude — turns off AI trade validation entirely; the bot runs purely on strategy rules + risk guards. This is the AI-free mode — everything works without it (signals, orders, stops); you lose only the AI sanity-check that skips low-confidence setups.
• Paper Trade — real signals, fake money. Leave ON until proven.
• Require Approval — every entry waits for your approve/reject (Inbox tab / Telegram) before placing.
• AI Provider — which brain validates trades when Claude validation is enabled (Claude CLI / OpenAI / Perplexity / Ollama).

1. Open the 🔬 Backtest tab. Test a strategy (or preset mix) on your watchlist over 1–3 years. Mind the net-of-cost numbers: win rate alone is meaningless without profit factor and max drawdown.
2. Only enable strategies whose backtest shows: profit factor > 1.3, max drawdown you could stomach in real money, and enough trades (n ≥ 30) to mean something.
3. Enable them in STRATEGIES, run in Paper for 2+ weeks, then compare live-paper results vs the backtest in the 🏆 Strategies tab. If live is far worse than the backtest, the strategy is overfit — turn it off.
4. Re-backtest quarterly and after regime changes. An edge that stops working should be retired without sentiment.

• P&L / Monthly — realized results by day/month. The equity/drawdown chips up top are the health summary.
• 🏆 Strategies — per-strategy live win rate and P&L: this is your "what's actually working" scoreboard.
• 🔭 Scanner — every symbol evaluated on the last cycle and why it did/didn't fire (RSI, RVOL, ADX, reason column).
• 📊 Positions / 📋 Trade Log — open exposure and the full audit trail.
• 📬 Inbox — pending approvals when Require Approval is on.

• MAX_POSITION_PCT — % of equity per position; MAX_OPEN_POSITIONS — concurrency cap; MAX_DAILY_LOSS_PCT — the daily circuit-breaker that halts trading. These guards run independent of any AI.

📘 Playbook — how to trade with this tool

1. Run Scan (top right) — refreshes the market regime, technical signals, and new candidates.
2. Refresh quotes — re-pulls live bids (incl. after-hours) and re-checks your stop/profit rules. Do this before acting; it's fast and uses no AI.
3. Analyze (in My Trades / My Stocks) — runs the pro-trader AI review on your held positions. Slower; do it once a day or after big moves.
4. Read Action Today top-to-bottom. If it says "nothing to do," that's a valid, calming answer — don't go looking for trades.

• Market Regime — GREEN = ok to add risk; YELLOW = be selective, halve size; RED = no new buys, protect the book. The scan auto-suppresses BUY signals when RED.
• Signal pills — counts of Buy/Wait/Avoid + how many positions "need a look today."
• Action Today — the only list you must act on. Order: YOUR RULE (your pre-set exits) → RULE (validated exit signals) → AI (judgment).
• Exit rule hits — when a position crosses your pre-committed level, act and tap "I followed it." Adherence is tracked.
• Book Health — effective leverage, premium-at-risk, total P&L. Enter your real equity once for accurate %.
• Concentration — red bars = a single name is over your cap across all accounts.

• Regime is GREEN or YELLOW (never add in RED).
• Conviction score ≥ 70 and the weekly trend agrees (✓ in the Wk column).
• Liquidity grade is Tradeable (not Marginal/Avoid) — wide spreads silently eat returns.
• Not within ~5 days of earnings (IV-crush risk) unless you intend to hold through.
• Adding it won't breach your 20% per-name concentration cap.
• Delta 0.70–0.85 (a real stock-replacement, not a lottery ticket).
• Vol isn't "expensive" (IV/RV) — you don't want to overpay for the option.

• Set a profit-take and a stop on every position (tap a trade → "Pre-committed exit rules"). Decide it calmly, once, in advance.
• When a rule fires, follow it — don't re-litigate emotionally. That's the entire point.
• Respect RULE exit signals: downside stop, theta roll (don't hold a LEAPS into its last ~12 months), delta roll (a 0.95-delta call is just expensive stock — roll up).
• Around earnings, decide in advance: trim the extrinsic gamble, or hold the multi-year thesis.

• Keep effective leverage sane — over ~2x means a 20% tech drop hurts disproportionately.
• Remember every LEAPS can go to zero: total premium-at-risk is your real downside, not market value.
• The whole book is one mega-cap-tech bet. In a selloff, all your names fall together — diversification by ticker is an illusion here.

• Analyze verdicts carry a source badge: RULES = deterministic, from the validated exit-signal/trend layers (works with zero AI, never "down"). AI = LLM judgment — richer narrative, but NOT backtested; weight it accordingly.
• If the AI layer is down, Analyze still gives real hold/trim/roll/exit calls from the rules — you lose only the news narrative.
• 📋 Copy AI prompt (in each position's detail panel) — a self-contained brief of the position, greeks, vol edge, rule signals and headlines. Paste into your own Claude/Perplexity Pro chat for a second opinion with a forced HOLD/TRIM/ROLL/EXIT verdict. No API key needed.

• Each signal shows its historical win-rate over 1/3/6 months with a 95% confidence interval and sample size n, tested on ~4 years of price data — plus a live record accruing from this app's own scans.
• Verdicts: edge = statistically better than buy-and-hold; not proven = indistinguishable (greyed out — don't lean on it); negative edge = worse than baseline; "too few" = not enough samples yet.
• Use it to calibrate trust: AVOID and WAIT have shown real value; BUY NOW hasn't beaten buy-and-hold — treat it as timing guidance, not alpha. Re-run monthly (button refreshes the cache) and after adding new tickers to the universe.

📘 Pro Picks — playbook

1. Pick a mode: Daily swing (hold days–2 weeks, faster momentum + pullback entries) or Weekly position (hold weeks–months, established trends + relative strength).
2. Hit Generate picks. It screens the universe, pulls news on the top candidates, checks the market regime, and returns 6 ranked picks with full plans (~1–2 min).
3. Read the market note first — it tells you how aggressive to be today. In a RED regime, take fewer/smaller trades or stand aside.
4. For each pick you get: entry (buy now / pullback / breakout with a price), stop, two targets, hold horizon, R:R, news sentiment, thesis, and what invalidates it.

• Only take trades where the plan's risk/reward is ≥ 1:2 and you're comfortable with the stop.
• Enter at the plan's zone — if it says "buy pullback to $X," wait for $X; don't chase.
• Set the stop the moment you enter — it's non-negotiable; that's the whole edge.
• Scale out at target 1, let a runner go to target 2 with the stop trailed up.
• Size so that hitting the stop loses only ~0.5–1% of your account (use the ATR% to size).

• Daily: run each morning; these are shorter-fuse swings — manage them actively, honor stops same-day.
• Weekly: run over the weekend; these are trend-follows — give them room, trail stops on the weekly.
• Don't act on stale picks — regenerate if the market has moved materially since.

• Every pick has a 📋 Copy AI prompt button — a self-contained brief with all the numbers, the plan, and recent headlines.
• Paste it into your Claude Pro or Perplexity Pro chat (your subscriptions — no API key). It's asked to open with a one-word CONFIRM or REJECT verdict, cite anything the data missed (earnings dates, news, macro), and refine the levels.
• Mark the result on the card: ✓ Confirmed (green border) or ✗ Rejected (card dims). Click again to clear. Only trade the confirmed ones.
• News sentiment on each pick is scored deterministically from live headlines (a financial lexicon) — it works even when the AI layer is down. NEGATIVE NEWS picks are demoted and docked a conviction star: read the headlines before entering.

• Setup score (0–100) — the overall rank. A weighted blend of every factor below, percentile-ranked against the other candidates in today's scan. Higher = the strongest-aligned setup relative to the rest. It's a ranking, not a probability — #1 ≠ "93% win chance."
• Confluence (0–100) — technical-setup quality: how many chart factors align — trend structure (price > 50-day > 200-day), weekly-timeframe agreement, a constructive (not overbought) RSI, nearness to support, and not being overextended. Higher = a cleaner chart.
• Relative strength (RS) vs SPY — the stock's 3-month return minus the S&P 500's. +75% means it beat the index by ~75 points over 3 months. Leaders outperform — "buy strength."
• 3-month % — the stock's raw price return over ~3 months (pure momentum).
• % of 52-wk high — where price sits vs its highest price in the past year. 90% = 10% below its 52-week high (near highs = strength). Closer to 100% is stronger; far above 100% would be extended.
• Vol (e.g. 2.06x) — volume surge: recent 5-day avg volume ÷ 60-day avg. 2x = trading at twice normal volume — institutional interest / a catalyst.
• RSI(14) — momentum oscillator. 40–60 = healthy/room to run; >70 = overbought (chasing risk); <40 in an uptrend = a dip-buy.
• ATR% — average daily range as a % of price = the name's volatility. It drives the stop distance and how small to size (a 12% ATR stop needs a smaller position than a 5% one for the same dollar risk).
• Signal — entry timing: BUY NOW (uptrend + constructive RSI + weekly agrees), WAIT FOR PULLBACK (extended — wait for a dip), AVOID (downtrend), NEUTRAL (unclear). Picks are filtered to uptrends only.