How This Tool Works
Everything you need to know — explained in plain language, no jargon
What does this system do?
The Champion Trader System (CTS) is a fully automated swing trading intelligence platform. Every day, it scans ~464 NSE stocks, identifies setups matching the Champion Trader methodology (PPC, NPC, Contraction patterns), and manages a virtual paper-trading portfolio to test its decisions.
The system runs autonomously. It scans stocks at market close, generates buy alerts, executes virtual trades, monitors stop-losses, and books profits — all with virtual capital. You review results and can use its signals for your own real trading decisions.
It improves itself overnight. After market hours, the AutoOptimize engine runs experiments — systematically tweaking scanning parameters and backtesting each change against 90 days of data. Changes that improve the composite score are kept; the rest are reverted. One AI analysis call per session evaluates the batch and guides the next session.
Monthly cost: ~₹180 ($2.20). The system uses AI for one strategic analysis per overnight session. Everything else — scanning, regime detection, risk monitoring, position sizing, learning notes — runs on pure math and rules. No per-trade AI calls.
The Daily Pipeline
Everything runs automatically on the server — independent of your laptop. Here’s the sequence.
Exit Monitor watches open positions
9:00 AM onwardsEvery 2 minutes during market hours, the system checks all open trades against their stop-loss and profit targets (2R, NE, GE, EE). If a stop-loss is hit, it generates a SELL alert. If a target is reached, it books the partial exit per the framework.
Go to Actions →Entry Monitor scans for trigger breaks
3:00 - 3:30 PMIn the last 30 minutes of market (entry window), the system checks every READY watchlist stock every minute. If price breaks above the trigger level, it generates a BUY alert. Autopilot then auto-executes with virtual capital.
Go to Actions →Daily Scanner runs PPC + NPC + Contraction scans
4:00 PMAfter market close, all three scan types run across the full NSE universe. Qualifying stocks are auto-added to the watchlist. A parallel baseline scan with frozen default parameters runs for A/B comparison.
Go to Pipeline →Regime Classifier detects market mood
4:45 PMPure math — no AI. Analyses NIFTY 50 trend, ADX, India VIX, and breadth to classify the market as TRENDING, RANGING, VOLATILE, or BEARISH. This regime affects which parameter bank is active.
Go to Intelligence →CIO Brief summarizes the day
5:00 PMGathers regime, overnight results, open positions, risk status, and top setups into a structured daily brief. The recommendation is rule-based: it checks regime, risk limits, and setup quality to give a clear action call.
Go to Intelligence →AutoOptimize runs overnight experiments
6:00 PM - 8:00 AMRuns up to 10 experiments per session. Each experiment: pick a parameter via systematic sweep, modify strategy.py, run a full 90-day backtest, compare composite scores, keep or revert. After all experiments, ONE AI call analyses the batch and logs strategic insights for the next session.
Go to Optimize →What Each Page Does
Morning overview — system health, open positions, watchlist
Your home base. Shows at a glance:
- System Status — Is the server healthy? Are scheduled jobs running?
- Open Positions — Current virtual trades with P&L
- Watchlist — Stocks being monitored for entry triggers
- Quick Actions — Links to pipeline, actions, and intelligence
Scanner + Watchlist Kanban — the stock discovery engine
Run manual scans or view today’s automated scan results. Stocks are organized into a Kanban board:
- READY — Trigger level set, waiting for breakout
- NEAR — Close to forming a setup, needs another day or two
- AWAY — Detected pattern but not yet actionable
Daily scanner runs automatically at 4:00 PM IST and auto-populates the watchlist via Autopilot.
BUY and SELL alerts generated by the price monitor
Real-time alert feed. The system generates two types:
- BUY alerts — Trigger level broken during entry window (3:00-3:30 PM). Shows entry price, stop-loss, position size, and targets.
- SELL alerts — Stop-loss hit or profit target reached. Shows exit price, P&L, and R-multiple.
Autopilot auto-executes alerts with virtual capital. For real trades, you execute through your broker.
Full trade log with P&L, R-multiples, and performance stats
Two tabs: Trades (individual trade history) and Performance (aggregate stats).
Every trade shows entry/exit dates, quantities, partial exits at each target level (2R, NE, GE, EE), gross P&L, and R-multiple. This is your trade journal.
Backtest the strategy against historical data
Historical Backtest — “If I had run this system from date X to date Y with ₹1,00,000, what would have happened?” Fetches real OHLCV data for ~464 stocks, runs the full PPC scan + entry/exit framework, and produces an equity curve with drawdown stats.
Paper Trading — Simulate the strategy day-by-day on live market data. Process one day at a time.
AutoOptimize uses backtests internally to score each parameter experiment. The composite score formula: expectancy × √trade_count × (1 - max_drawdown).
Regime, risk status, daily brief, and setup cards
Shows 4 things at a glance:
- Market Regime — TRENDING / RANGING / VOLATILE / BEARISH. Pure math, no AI.
- Risk Status — Open risk as % of capital. Entry freeze if limit breached.
- Daily Brief — Structured summary with rule-based recommendation.
- Top Setups — Today’s highest-scoring stock picks with entry/SL/targets.
AutoOptimize experiment history and parameter tuning
The overnight engine that makes the system smarter. Each session:
- Runs up to 10 experiments using a deterministic parameter sweep (no AI cost)
- Each experiment modifies ONE parameter, runs a 90-day backtest, compares the composite score
- Improvements are kept (git committed); no-improvement is reverted
- After all experiments, one AI call analyses the full batch and logs strategic insights
Cost: ~₹8/night ($0.10). Total ~₹180/month ($2.20).
Which signal types work best in which market regimes?
A scorecard that tracks win rate and average R-multiple for each signal type (PPC, NPC, Contraction) crossed with each market regime (Trending, Ranging, Volatile, Bearish).
After 20+ trades per combo, the system flags underperforming signal-regime pairs for priority re-optimisation.
Check monthly — needs enough data to show meaningful patterns.
Behind the Scenes
10 Scheduled Jobs
These run automatically on the server. Each is a Python function triggered by APScheduler at the specified time. They run independently of your laptop.
Only AutoOptimize uses AI (one call per session, ~$0.10/night). Everything else is pure math and rules.
Exit Monitor
Every 2 min, 9:00-15:30 ISTChecks all open positions against stop-loss and profit targets. Generates SELL alerts on hits.
Entry Monitor
Every 1 min, 15:00-15:30 ISTChecks READY watchlist stocks for trigger-level breakouts. Generates BUY alerts.
Risk Guardian
Every 10 min, 9:00-15:30 ISTMonitors total portfolio risk. Freezes new entries if open risk exceeds 10% of capital.
Daily Scanner
4:00 PM weekdaysRuns PPC + NPC + Contraction scans on ~464 stocks. Auto-populates watchlist. Runs A/B baseline comparison.
Regime Classifier
4:45 PM weekdaysClassifies market as TRENDING/RANGING/VOLATILE/BEARISH using NIFTY trend, ADX, VIX. Pure math.
CIO Daily Brief
5:00 PM weekdaysGenerates structured daily summary with regime, risk, positions, setups, and rule-based recommendation.
Learning Agent
Every 30 min, 9:00-16:00 ISTWrites post-mortem for each closed trade: regime at entry, exit quality, R-multiple, learning insight. Updates signal attribution table.
Shadow Portfolio
Every 30 min, 9:00-16:00 ISTTracks exits on shadow (paper) trades to compare machine suggestions vs human-approved picks.
Corpus Updater
5:30 PM weekdaysSaves today's market data (index levels, sector performance) to the RAG memory system.
AutoOptimize
6:00 PM weekdays (overnight)Runs 10 parameter experiments with backtests. One AI call analyses the batch. Keeps improvements, reverts failures.
Jargon Buster
Every technical term explained in plain language
PPC / NPC / Contraction
Three types of stock patterns the system scans for. PPC (Positive Price Candle) = a strong green day with high volume near the top of the range. NPC (Negative Price Candle) = the opposite pattern, used as a filter. Contraction = the stock's daily range is narrowing, often a sign it's about to make a big move.
Stop-Loss (SL)
A safety net price. If a stock falls to this level, the position is exited to limit loss. Calculated as Entry Price minus TRP value. The system never moves a stop-loss down.
TRP (True Range Percentage)
How volatile a stock is, expressed as a percentage of price. Higher TRP = more volatile = wider stop-loss = smaller position size. Minimum 2.0% to be tradeable.
RPT (Risk Per Trade)
Percentage of total capital risked on each trade. Default 0.5%. So with 1,00,000 capital, you risk 500 per trade. This determines position size.
R-Multiple
Profit measured in units of risk. If you risked 500 and made 1,000, that's 2R. If you lost 500, that's -1R. The exit framework targets: 2R (mathematical), 4R (normal extension), 8R (great extension), 12R (extreme extension).
Composite Score
A single number measuring strategy quality. Formula: expectancy x sqrt(trade_count) x (1 - max_drawdown). The overnight optimizer tries to maximise this. Penalised if drawdown exceeds 15%.
Market Regime
The system classifies market conditions into 4 modes: TRENDING (strong direction), RANGING (sideways), VOLATILE (big swings), BEARISH (falling). Different parameter banks activate for each regime.
Expectancy
Expected R per trade, calculated as (Win Rate x Avg Win R) - (Loss Rate x Avg Loss R). Positive means the strategy has an edge over many trades.
Autopilot
The virtual paper trading engine. Uses 1,00,000 virtual capital, 0.5% RPT, max 5 positions, max 10% open risk. Automatically executes BUY/SELL alerts. No real money involved.
AutoOptimize
The overnight self-improvement engine. Runs 10 experiments per session, each testing a single parameter change via a full 90-day backtest. One AI call per session analyses results. Cost: ~180/month.
Parameter Banks
Different sets of scanning thresholds for different market regimes. Like driving settings: highway mode (trending), city mode (ranging), rain mode (volatile). The regime classifier decides which bank is active.
A/B Comparison
Every day, the scanner runs twice: once with optimised parameters and once with frozen default parameters. This measures whether AutoOptimize improvements are real or just noise.
Signal Attribution
A scorecard tracking win rate and average R for each signal type (PPC, NPC, Contraction) in each market regime. Flags underperforming combos after 20+ trades.
Frequently Asked Questions
Does this system trade real money?
No. The Autopilot runs on virtual capital (1,00,000). It generates alerts and executes paper trades to test the strategy. For real trading, you use the signals as input and execute through your own broker.
What runs automatically without my laptop?
Everything. The system runs on a cloud server (EC2) with 10 scheduled jobs via APScheduler. It scans stocks, monitors positions, classifies market regime, generates briefs, runs overnight optimisation, and manages the virtual portfolio. All independent of your laptop.
How does the system learn and improve?
Three feedback loops: (1) AutoOptimize runs 10 parameter experiments per night, backtesting each change and keeping improvements. (2) The Learning Agent writes a post-mortem for every closed trade, tracking which signal types work in which regimes. (3) Signal Attribution flags underperforming patterns after 20+ trades.
What does the AI actually do?
Very little. AI makes ONE call per overnight session (~$0.10) to analyse the batch of 10 experiment results and suggest strategic direction. Everything else — scanning, regime detection, risk monitoring, position sizing, exit framework, learning notes, daily brief — is pure math and rule-based logic.
How much does it cost to run?
About $2.20/month for AI calls (22 trading days x $0.10/session). The EC2 server cost is separate (infrastructure). No per-scan, per-trade, or per-alert AI charges.
What is the exit framework?
When a trade hits 2R profit: sell 20% (mathematical exit). At 4R: sell 20% (normal extension). At 8R: sell 40% (great extension). At 12R: sell 80% (extreme extension). Stop-loss trails up after each target. Remaining position exits when price closes below 50-day DMA.
What's the difference between Simulation and AutoOptimize?
Simulation is a tool YOU use to manually test the strategy over any date range. AutoOptimize uses backtests AUTOMATICALLY overnight to test parameter changes. Same engine underneath, different purpose.
How does the A/B comparison work?
Each day, the scanner runs twice: once with the current optimised parameters and once with frozen default parameters (never changed). The daily comparison shows whether the optimiser's changes are actually helping or hurting.
Getting Started
The system runs autonomously. Here’s how to monitor it.
Check the Dashboard
See system health, open positions, and watchlist at a glance. Verify all 10 scheduled jobs are running.
Review the Intelligence Hub
Check today's market regime, risk status, and the daily brief. See if there are any high-scoring setups.
Monitor the Pipeline
View scan results from 4:00 PM. READY stocks will be checked for trigger breaks in tomorrow's entry window (3:00-3:30 PM).
Watch Actions for BUY/SELL alerts
The price monitor generates alerts automatically. Autopilot executes them with virtual capital. You can use these signals for your real broker trades.
Review Trades for P&L tracking
All virtual trades with entry/exit, partial exits, R-multiples, and gross P&L. Use the Performance tab for aggregate stats.
Check Optimize results each morning
See how many overnight experiments were run, how many improved the composite score, and read the AI session analysis.