Glossary — TradeBytes Ideas

Pipeline stages

Each investigation runs through four agent stages. An investigation can stop early at any stage if there's no edge to publish.

Scout: Deep multi-hour research on one symbol. Reads SEC filings, earnings transcripts, options chains, insider activity, news. Builds a "dossier" of evidence. Decides if there's something worth pursuing or not. Model: MiniMax 2.7 on local hardware.
Analyst: Reads Scout's dossier, computes a 0-100 composite score against the methodology rubric (smart money, options flow, catalyst, mispricing, quality, technical), and decides: draft an idea or skip. Model: Qwen 3.6 35B.
Devil's Advocate: Adversarial review. Re-fetches sources to look for ways to break the Analyst's thesis. Issues PASS or KILL with rationale. Different model than Analyst (MiniMax 2.7 fresh context) so it doesn't just rubber-stamp.
Compliance: Final gate before publish. Verifies disclosure language, factual claims trace to sources, position sizing matches confidence. Hard-coded gates in code that can't be talked into PASS by an LLM hallucination.
Reviewer: Runs after an idea closes. Compares thesis vs actual outcome, extracts lessons. Future Scout/Analyst stages read those lessons before forming new theses, so the system gets sharper over time.

Outcomes (what shows in Recent Activity)

Each completed investigation gets one of these labels. Most are NOT publishes — by design. The pipeline is built to publish rarely and only when something clears a high bar.

Scout verdicts

promising: Scout found a long-side asymmetric setup worth advancing to Analyst. Mispricing + catalyst + edge in data. Often becomes a published idea.
range_bound_or_income: Stock is fundamentally healthy but at/near 52-week high, fully valued, IV elevated. Not an asymmetric long, but a candidate for income strategies — covered calls, strangles, or cash-secured puts on pullback. Analyst will pick the right structure.
bearish_setup: Deteriorating fundamentals + bearish catalyst + IV elevated. Open to short or naked-call thesis (carefully — uncapped risk). Rare verdict.
no_anomaly: Scout did the research and found genuinely nothing to act on — fully priced with low IV (no income premium), or no thesis of any direction works. Pipeline halts here. This is the most common outcome by design. The system is supposed to say "no" most of the time.
needs_more_data: Data gaps in filings, illiquid options, news blackout. Re-investigate later when more is available.

Analyst decisions

skip: Analyst scored the dossier against the methodology rubric and the score didn't clear the bar — or an anti-signal gate fired (customer concentration > 30%, accounting irregularities, etc.). Most Scout-promising names still skip here. The Analyst's bar is high.
draft: Score cleared the bar. Analyst wrote up a thesis with structure (long stock / CSP / CC / etc.), entry/exit/risk, sources. Hands off to Devil's Advocate.

Devil's Advocate verdicts

PASS: Adversarial review couldn't break the thesis. Forwards to Compliance for final review.
KILL: Devil's Advocate found a flaw — stale data, cherry-picked evidence, base-rate failure, hidden downside. Idea dies here. Most Analyst drafts die at this stage; that's the whole point.

Compliance verdicts

pass: Disclosure, sourcing, structure, and sizing rules all satisfied. Idea publishes to the live site immediately — no manual gate.
hold: Idea cleared Devil's Advocate but a gate fired late: missing/insufficient sources, wrong structure for the thesis, position size out of bounds, or the daily publish cap was hit. Moved to content/deferred/; promoted automatically the next day if budget allows.
compliance_kill: Hard fail: anti-signal gate (customer concentration, accounting irregularity, dilution overhang, etc.) detected only at the Compliance step. Idea is killed permanently — same fate as a Devil's Advocate KILL.

Final outcomes

published: Idea cleared all four stages and is live on the site. Joins the open ideas in Track Record. Position monitoring (every 15 min) tracks performance until close.
failed: Pipeline stage crashed unexpectedly (LM Studio error, network failure, etc). Queue item is recovered automatically; symbol gets re-tried later.
unparseable: Model emitted output that couldn't be parsed as valid JSON even after fallback repair. Rare. Symbol re-queues for retry.

Chart signal

A 1-year technical read on the stock — separate from the methodology's full thesis. Useful as a quick directional gut-check.

BUY: Composite chart score ≥ +5. Trend is up (above MAs), momentum positive, no extreme overbought.
HOLD: Score between -2 and +4. Mixed signals, sideways, or constructive but not screaming.
SELL: Score ≤ -3. Trend is down (below MAs), broken below 52-week high, momentum negative or RSI extreme overbought.
confidence: 1-5 rating of how aligned the underlying signals are. 5/5 = all factors point the same direction. 2/5 = mixed.
score: Signed integer summing the chart factors: trend (above MAs, golden/death cross), momentum (12-1 month return), RSI (overbought/oversold), MACD (cross), distance from 52-week high.

Price targets — bear / fair / bull

Three reference prices for every tracked stock, blended from Wall Street analyst consensus + chart-derived floors and ceilings.

Bear: Pessimistic case. The lower of: lowest analyst target, the 200-day moving average, or 30% below the 52-week high. Where the price could go if things break the wrong way.
Fair: Consensus blend. Average of: mean analyst target, current price, and the 50-day moving average. Where the price probably is if everyone's right about today.
Bull: Optimistic case. The higher of: highest analyst target, or 15% above the 52-week high. Where the price could go if the bull thesis plays out.
Wall Street consensus: Aggregate of analyst targets reported via Yahoo Finance. The system shows the analyst count so you know how diluted the consensus is (3 analysts vs 57 is a different beast).

Trade structures

Different shapes of trade for different thesis shapes. The Analyst picks the structure based on what the dossier supports.

long_stock: Buy the stock outright. Used for asymmetric long theses with multi-quarter time horizon.
csp (cash-secured put): Sell a put option, set aside the cash to buy 100 shares per contract if assigned. Bullish, income-generating, defined "buy zone" via the strike. You're essentially being paid to wait for a pullback.
cc (covered call): Sell a call option against shares you own. Income-generating on a stock you'd hold anyway. Caps upside at the strike but collects premium. Best when stock is range-bound or fully valued.
strangle (short strangle): Sell an out-of-the-money call AND an out-of-the-money put at the same expiry. Profit if stock stays between the two strikes. Open-ended risk beyond either breakeven.
naked_put: Sell a put without setting aside the cash. ~2× the yield of a CSP but uses margin. Margin-call risk if underlying drops sharply.
naked_call: Sell a call without owning the underlying. Theoretically unlimited downside if the stock rallies. Reserved for clear bearish-with-IV-elevated setups. Confidence ceiling 4 (never sized at full conviction).

Risk class

core: Standard sizing. Methodology caps: confidence 3 → up to 2%, conf 4 → 3%, conf 5 → 5%.
⚠ ASYMMETRIC: Small-cap moonshot. Sub-$5B market cap, "$1,000 → $100,000" payoff math. 90%+ of these bets lose money or go to zero; the few that work pay for the rest. Position size capped at 0.5% / 1.0% / 1.5% by methodology — much smaller than core. Always rendered with the orange ASYMMETRIC badge so the risk profile is unmissable.

Queue & sources

Every investigation starts as a queued request. Where it came from determines its priority.

P0: Manual injection — operator dropped this symbol in directly. Front of queue.
P1: Catalyst-driven. Came from Hunter (options flow / SEC 8-K / Form 4), Sector Sweep (top of daily ranking), or Synergy Scout (cross-thesis beneficiary). Investigated ahead of routine watchlist rotation.
P2: Watchlist rotation. Oldest-due symbol from one of the six sector watchlists. Fills idle gaps when no catalyst-driven items are pending.
Hunter: Periodic agent (every 2h) scanning each watchlist symbol for catalysts: unusual options flow, recent 8-K filings, Form 4 cluster activity. News headlines are demoted to confirmation only — institutional positioning shows up first in options/filings.
Sector Sweep: Daily top-down ranking. Scores every watchlist symbol on quick technical and signal data, queues the top N for full Scout investigation.
Synergy Scout: Daily cross-thesis discovery. Reads recent dossiers + open ideas, identifies companies that benefit from the SAME catalysts (primary supplier, secondary capex flow, tertiary macro). Surfaces non-obvious connections; can auto-add new tickers to the appropriate sector watchlist if validated.
Watchlist Seeder: Every 6h, picks the oldest-due symbol from any watchlist (rotation interval per watchlist, 14-21 days) and queues it. Ensures the universe gets fair coverage even when nothing catalyst-driven is firing.

Methodology terms

Score (0-100): Analyst's composite score across six categories: smart money (insider buys, 13F initiations) max 25 · options flow (whale blocks, V/OI) max 10 · catalyst (earnings, M&A, guidance) max 25 · mispricing (P/E, EV/EBITDA, FCF yield) max 15 · quality (ROIC, margins, balance sheet) max 15 · momentum/technical max 10.
Confidence (1-5): Derived from the score. 60-69 → conf 3 (cap 2%), 70-79 → conf 4 (cap 3%), 80+ → conf 5 (cap 5%) for core. Asymmetric is much tighter (0.5% / 1.0% / 1.5%).
Anti-signal gates: Things that force a skip regardless of score: going-concern audit, pending material litigation, customer concentration > 30%, dilution risk > 10%, accounting irregularities, low liquidity. Compliance enforces these in code, not just the prompt.
Non-obviousness (Synergy Scout): 1-5 rating of how cross-cutting a synergy is. 5 = an analyst desk would miss this. 1 = "NVDA → AMD is correlated" (not really a synergy, just a sector trade).
Directness (Synergy Scout): How direct the supply-chain link is. Primary: direct customer/supplier. Secondary: capex flow downstream. Tertiary: same macro tailwind, different exposure.
12-1 momentum: The 12-month return ending one month ago — skipping the most recent month to avoid mean-reversion noise. Standard momentum factor in academic finance.

For the full framework, see the methodology page.