Bot Architecture

The research drives a trading bot. Every strategy module traces back to a vetted trader. The 11 modules live in bot/lib/setups/; each uses a shared precompute() hook so scan() is O(1) per bar — the runner walks ~2,700 bars × 25 symbols × 11 strategies in seconds.

Platform stack

Layer	Choice	Why
Research / backtest / live	QuantConnect (LEAN, Python)	Same code runs backtest and live; 20+ years tick data included
Broker — equities / options	Interactive Brokers Pro	DMA, sub-penny commissions, full API
Broker — futures	Tradovate	Cheap micro-futures for Raschke / index-futures modules
Options / GEX data	SpotGamma or MenthorQ	Karsan's dealer-positioning feed
Second-source data	Polygon.io	Corroboration + news timestamping for Episodic Pivot
Dashboard	Vercel + static HTML	Human-facing cockpit — not the bot itself

Strategy → trader mapping (all 11 implemented)

Setup module	Trader	Side	Notes
`minervini_sepa`	Mark Minervini	Long	Trend Template + VCP pivot break on volume surge
`kullamaggie_breakout`	Kristjan Kullamägi	Long	30–100%+ leader run, 2–8 week tight base, rising 10/20-EMA, volume breakout
`kullamaggie_episodic_pivot`	Kristjan Kullamägi	Long	10%+ catalyst-gap proxy, exceptional volume, sideways/neglected prior base
`kullamaggie_parabolic_short`	Kristjan Kullamägi	Short	50–100%+ climax run, 3–5+ up days, topping failure into 10/20-MA target
`shannon_avwap`	Brian Shannon, CMT	Long	AVWAP reclaim from confirmed swing low + MTF alignment
`raschke_holy_grail`	Linda Raschke	Both	ADX(14) > 30 + 20-EMA pullback + resumption
`raschke_turtle_soup`	Linda Raschke	Both	Failed 20-day Donchian breakout — fade back into range
`raschke_80_20`	Linda Raschke	Both	Range-edge close + next-day probe-and-reverse
`brandt_classical`	Peter L. Brandt	Both	Double bottom / double top neckline break (semi-auto in prod)
`bruzzese_rs`	Vincent Bruzzese	Both	SPY-regime gate × cumulative RS vs SPY × prior-extreme break
`karsan_gex`	Cem Karsan	Both	Realized-vol regime proxy → fade or continue large daily moves

Meta-layers

Three traders don't generate signals but shape the infrastructure:

Dr. Brett Steenbarger → the journal / report-card framework in lib/journal/ — auto-logged process score, rule violations, and the daily narrative.
Lance Breitstein & Tom Hougaard → the risk manager's hard-coded discipline rules (per-trade cap, daily loss cap, drawdown kill-switch) so the bot enforces what a human trader can't reliably enforce mid-session.
Vincent Bruzzese's walk-away analysis — MFE / MAE vs actual exit, surfaced per trade.

Deliberately NOT automated

Breitstein's A+ capitulation fades. Pure tape-reading; attempting to code it produces the canonical "bot shorts the bottom tick" disaster.
Hougaard's discretionary size-ups. His edge is psychological — coding "when I feel conviction, 3x size" cannot work.

System flow

                        ┌─────────────────────────────┐
                        │  DATA LAYER                 │
   broker feeds ──────► │  - price (QC + Polygon)     │
   SpotGamma ─────────► │  - fundamentals             │
   news APIs ─────────► │  - options chain / GEX      │
                        └──────────────┬──────────────┘
                                       ▼
                        ┌─────────────────────────────┐
                        │  UNIVERSE + REGIME          │
                        │  - Minervini Trend Template │
                        │  - Bruzzese RS table        │
                        │  - SPY regime + GEX regime  │
                        └──────────────┬──────────────┘
                                       ▼
             ┌───────────────── SETUP MODULES ────────────────┐
             │  Minervini · Kullamägi · Shannon · Raschke    │
             │  Brandt · Bruzzese · Karsan                   │
             └──────────────┬──────────────────┬──────────────┘
                            ▼                  ▼
                 ┌──────────────────┐  ┌────────────────────┐
                 │ RISK MANAGER     │  │ STRATEGY ALLOCATOR │
                 │ (Hougaard/       │◄►│ (regime-weighted)  │
                 │  Breitstein)     │  └────────────────────┘
                 └─────────┬────────┘
                           ▼
                 ┌──────────────────┐
                 │ EXECUTION ENGINE │   →   IBKR / Tradovate
                 └─────────┬────────┘
                           ▼
                 ┌──────────────────┐  ┌────────────────────┐
                 │ JOURNAL          │─►│ DASHBOARD          │
                 │ (Steenbarger)    │  │ (live cockpit)     │
                 └──────────────────┘  └────────────────────┘

Promotion pipeline

A strategy doesn't reach live capital by decree. Three gates: