galaxis-po/docs/plans/2026-02-19-kjb-strategy-design.md

Kim Jong-bong (KJB) Strategy - Full System Design

Date: 2026-02-19 Status: Approved

Overview

Implement the Kim Jong-bong short-term trading strategy from quant.md as a full system: backtesting, daily signal generation, portfolio management, and frontend dashboard.

This strategy is fundamentally different from existing strategies (Quality, MultiFactor, ValueMomentum) which are periodic-rebalancing based. KJB is signal-based daily trading with individual position management, stop-losses, and trailing stops.

Trading Rules (from quant.md)

Universe

• Market cap rank: 1-30 (large caps)
• Daily trading value >= 200 billion KRW

Entry Signals (all must be true)

• Relative Strength vs KOSPI > 100 (stock outperforming market over 10 trading days)
• AND at least one of:
  • Box breakout (close > 20-day high)
  • Large bullish candle (daily return >= 5% AND volume >= 1.5x 20-day average)

Exit Rules

• Stop-loss: -3% from entry price
• Take-profit 1: +5% → sell 50% of position
• Take-profit 2: +10% → sell remaining
• Trailing stop: when +5%, move stop to entry price; when +10%, move stop to +5%

Portfolio Rules

• Cash reserve: 30% (investable = total capital x 0.7)
• Max positions: 10
• Position size: equal weight (investable / max_positions)

Architecture

1. Strategy Layer

KJBStrategy(BaseStrategy) - services/strategy/kjb.py

• Compatible with existing strategy pattern (returns StrategyResult)
• Used for stock ranking by KJB criteria
• Scores based on: relative strength, breakout proximity, volume trend

KJBSignalGenerator - services/strategy/kjb.py

• Separate class for daily signal generation
• Input: stock ticker + price data
• Output: buy/sell signals with entry, target, stop-loss prices
• Used by: DailyBacktestEngine, daily scheduler job

2. Daily Backtest Engine

DailyBacktestEngine - services/backtest/daily_engine.py

• New engine alongside existing BacktestEngine
• Daily simulation loop (not rebalance-based)
• Each day: check entry signals for universe, check exit conditions for positions
• Uses TradingPortfolio for position management

TradingPortfolio - services/backtest/trading_portfolio.py

• Separate class from existing VirtualPortfolio
• Supports: enter_position(), exit_position(), partial_exit()
• Per-position tracking: stop_loss, target_price, trailing_stop
• Cash reserve enforcement (30%)
• Transaction recording with reasons

3. Signal System

DB Model: Signal - models/signal.py

• Fields: date, ticker, signal_type (buy/sell/partial_sell), entry_price, target_price, stop_loss_price, reason, status (active/executed/expired)

Scheduler Job: kjb_daily_signal_job - jobs/kjb_signal_job.py

• Runs daily at 18:00 KST (after price data collection)
• Generates signals for all universe stocks
• Saves to DB

4. API Endpoints

Signal API - api/signal.py

• GET /api/signal/kjb/today - today's buy/sell signals
• GET /api/signal/kjb/history - historical signals
• GET /api/signal/kjb/positions - active positions with P&L

Strategy API additions:

• POST /api/strategy/kjb - run KJB ranking

Backtest additions:

• Existing backtest API works with strategy_type="kjb"

5. Frontend

• KJB signal dashboard page
• Today's signals cards (buy/sell recommendations)
• Active positions table (entry price, current price, P&L, stop-loss level)
• Portfolio value chart

Files to Create

File	Purpose
services/strategy/kjb.py	KJBStrategy + KJBSignalGenerator
services/backtest/daily_engine.py	DailyBacktestEngine
services/backtest/trading_portfolio.py	TradingPortfolio
models/signal.py	Signal DB model
schemas/signal.py	Signal Pydantic schemas
api/signal.py	Signal API endpoints
jobs/kjb_signal_job.py	Daily signal scheduler job
Alembic migration	signals table
Frontend pages	KJB dashboard

Files to Modify

File	Change
services/backtest/engine.py	Add "kjb" to _create_strategy()
models/__init__.py	Register Signal model
api/__init__.py	Register signal router
schemas/backtest.py	Add "kjb" to strategy_type enum
jobs/scheduler.py	Register KJB signal job
api/strategy.py	Add KJB strategy endpoint