Paper Trading Hardening Guide

1 — What Changed At System Level

Stale Price Protection

Cached prices are checked for age before return.
Stale prices log warnings but do not block trades.
MCX alias path uses the same protection.

Viability Engine

Binary cost filter evolved into graded scoring.
Uses per-share efficiency instead of total-size distortion.
Blocks low-edge and tiny-absolute-edge trades.

Manual Override Control

Manual detection is source/metadata-first.
STRICT / WARN / OFF modes supported.
Manual override affects viability only, not all risk checks.

2 — Stale Price Protection

How It Works

_get_current_price() reads cached price and timestamp.
If age is over 2 seconds, it logs [STALE PRICE].
Price is still returned. This is soft validation, not rejection.

if age > 2.0:
    logger.warning(f"[STALE PRICE] {symbol} → {age:.2f}s old")
return price

Why This Is Correct

Real brokers also operate through delayed or bursty ticks.
Blocking on every stale event causes missed trades and unstable behavior.
Logs give observability without disrupting execution flow.

Recommended: keep this warn-only

3 — New Viability Engine

The paper gate is now a graded decision engine. It combines cost realism, per-share efficiency, minimum meaningful edge, volatility-aware adaptive filtering, and safe manual override behavior.

Core Formula

per_share_profit = expected_move
per_share_cost = est_cost / qty
viability_score = per_share_profit / per_share_cost

dynamic_min_edge = volatility_unit * multiplier * qty
final_min_edge = max(min_expected_profit, dynamic_min_edge)

if expected_gross < final_min_edge:
    tier = "REJECT_ABSOLUTE"

Why Per-Share Is Better

Total-profit scoring is quantity-dependent and can mislead decisions.
Per-share scoring measures setup quality itself.
This makes comparison across cheap and expensive symbols more reliable.

Tier	Rule	Meaning	Default Behavior
STRONG	score ≥ 2.0	Clear edge after cost	Allow
ACCEPTABLE	score ≥ 1.2	Tradable but not elite	Allow
WEAK	score ≥ 0.8	Marginal edge	STRICT uses configurable REJECT/WARN
REJECT	score < 0.8	Poor cost efficiency	Reject unless override or mode allows
REJECT_ABSOLUTE	expected_gross < final_min_edge	Too small to matter	Hard reject

4 — Context-Aware Minimum Edge

Adaptive Edge Logic

Static floor alone is too rigid across quiet and volatile markets.
The engine computes a volatility-aware dynamic minimum edge.
Final filter uses the larger of static and dynamic minimum edge.

dynamic_min_edge = volatility_unit * 0.5 * qty
final_min_edge = max(min_expected_profit, dynamic_min_edge)

Volatility Proxy

If stop loss exists: volatility_unit = abs(entry - stop_loss)
Otherwise: fallback proxy from expected move and RR.
This gives ATR-like behavior without requiring new candle dependencies.

Minimum Volatility Floor

Ultra-tight stops can make volatility too small and weaken adaptive filtering.
The engine now applies min_volatility_unit = entry * min_volatility_floor_pct.
Default is 0.001 = 0.1%, configurable in the UI.
The setting is clamped to 0.0001 through 0.02 for safety.

min_volatility_floor_pct = max(0.0001, min(user_value, 0.02))
min_volatility_unit = entry * min_volatility_floor_pct
volatility_unit = max(volatility_unit, min_volatility_unit)

5 — Manual Override Modes

STRICT

Default production mode.
Rejects REJECT tier trades.
WEAK tier follows REJECT or WARN setting.
Manual override can allow manual-only non-viable trades.

WARN

Allows trades but logs viability tier and score.
Best for testing execution pipeline and monitoring behavior.
Do not use as default production mode.

OFF

Turns off viability scoring gates.
Useful only for troubleshooting or isolated experiments.
Risk logic outside viability still remains.

Important: manual override is not a trading edge. It is a testing convenience. Do not rely on override to justify poor signals.

6 — How To Configure It

Production / Auto Bot

Trade Quality Gate = ON
Non-Viable Mode = STRICT
Weak Tier Action = REJECT
Manual Override = OFF
Min Expected Profit = ₹100 or higher
Adaptive Min Edge = ON
Adaptive Multiplier = 0.5
Min Volatility Floor = 0.1%

Testing / Debug

Trade Quality Gate = ON
Non-Viable Mode = WARN
Weak Tier Action = WARN
Manual Override = ON only if needed
Keep minimum edge slightly lower to observe more cases

Asset Portability

Equities: start around 0.1% floor
Forex: can be lower, around 0.05%
Crypto: often needs 0.2% to 0.5%
Tune only after reviewing logs, not by intuition

7 — How This Helps Profitability

What Improves

Fewer trades where costs silently dominate edge.
Less churn in low-priced names with tiny targets.
More realistic paper results before live deployment.
Better net expectancy by removing weak or meaningless trades.

What It Does Not Do

It does not guarantee profitability.
It does not create edge if the strategy has none.
It does not replace good signal quality or market structure.

Real objective: fewer but better trades. A drop in trade count is usually healthy if net expectancy, realized Sharpe, and cost efficiency improve.

8 — What To Watch In Logs

New Debug Visibility

expected_gross=84.00
est_cost=62.30
per_share_profit=0.4200
per_share_cost=0.3115
min_expected_profit=100.00
dynamic_min_edge=73.50
final_min_edge=100.00
min_volatility_floor_pct=0.0010
effective_volatility_unit=0.3500
tier=REJECT_ABSOLUTE
score=1.348

score tells you per-unit edge quality.
tier tells you decision class.
final_min_edge shows whether static or adaptive rule dominated.
effective_volatility_unit shows whether the floor likely kicked in.

9 — Quick Reference

Question	Answer
Does stale price protection block trades?	No. It logs stale price age and still returns the price.
What is viability score based on?	Per-share profit divided by per-share cost.
What blocks tiny but efficient trades?	`REJECT_ABSOLUTE` via `final_min_edge`.
What is adaptive min edge?	`max(min_expected_profit, volatility_unit × multiplier × qty)`
How is volatility estimated?	Stop distance when available, otherwise expected move divided by RR.
What stops ultra-tight SL abuse?	A configurable minimum volatility floor percentage.
Is the floor safe against bad configs?	Yes. It is clamped between 0.01% and 2% in the engine.
Should manual override be used in production?	No. Keep it for testing, debugging, and isolated operator workflows.

10 — Recommended Presets For Indian Markets

Scope: these presets are for Indian-market workflows only: NSE cash, NFO index or stock derivatives, and MCX commodities. No crypto assumptions are used here.

NSE Cash Intraday

Trade Quality Gate = ON
Non-Viable Mode = STRICT
Weak Tier Action = REJECT
Min Expected Profit = ₹100 to ₹250
Adaptive Min Edge = ON
Adaptive Edge Multiplier = 0.5
Min Volatility Floor = 0.10%
Manual Override = OFF

Best for equities like SBIN, RELIANCE, INFY, ICICIBANK

NFO Options / Futures

Trade Quality Gate = ON
Non-Viable Mode = STRICT
Weak Tier Action = REJECT
Min Expected Profit = ₹250 to ₹750
Adaptive Min Edge = ON
Adaptive Edge Multiplier = 0.7
Min Volatility Floor = 0.15% to 0.25%
Manual Override = OFF

Use stricter edge because derivatives can look attractive but decay fast after costs

MCX Intraday Commodities

Trade Quality Gate = ON
Non-Viable Mode = STRICT
Weak Tier Action = REJECT
Min Expected Profit = ₹300 to ₹1000
Adaptive Min Edge = ON
Adaptive Edge Multiplier = 0.75
Min Volatility Floor = 0.15% to 0.30%
Manual Override = OFF

Useful for GOLD, SILVER, CRUDEOIL, NATURALGAS where volatility regimes shift sharply

Testing Preset For Indian Markets

Non-Viable Mode = WARN
Weak Tier Action = WARN
Manual Override = ON only for manual signal testing
Keep Min Expected Profit lower than production
Watch logs first, then tighten filters after 1 week of observation

What Not To Do

Do not copy low-volatility equity settings into MCX blindly.
Do not lower slippage or spread just to increase paper profitability.
Do not leave Manual Override enabled in production.
Do not set Min Volatility Floor too high or almost every trade will fail edge tests.