paper_trading_practice v1.0.0

name: paper_trading_practice description: Run a paper trading practice desk the honest way. Sets up a realistic simulated account, enforces no-lookahead discipline, and evaluates the trade journal before any real money is risked. Use when a user wants to practice trading, test a strategy at real prices, or check whether their paper results are good enough to go live.

Paper Trading Practice

Purpose

Paper trading only teaches something if the practice matches reality. This skill gives you a full workflow: set the desk up like real money, trade with discipline rules that prevent self-deception, log every trade, and evaluate the journal with math instead of gut feeling.

Workflow

1. Set up the desk like it is real

• Account size matches the plan. If the user will fund with $10,000, the paper account starts at $10,000. Oversized virtual cash teaches oversized risk.
• Position sizing rules before the first trade. Fixed fractional risk (common: 1% of equity per trade), max daily loss limit, max concurrent positions. Write them down.
• Commissions and slippage on every fill. If the broker charges $0.65 per options contract or the spread is wide on small caps, include it. Assume fills are worse than the mid price, never better.
• One strategy at a time, one variable at a time. If you change the entry rule mid-month, start a new test segment and label it.

2. Pre-register the rules

Before trading, commit in writing: the entry signal, the exit rules (stop and target), the position size, the market conditions the strategy claims to work in, and the success criteria (how many trades, what profit factor, what max drawdown). Log this pre-registration with a date. Any rule change after results appear counts as a new test. See references/evaluation_guide.md for a template.

3. Trade with no-lookahead discipline

Look-ahead bias is using information you did not have at the moment of the decision. It is the most common way paper results lie.

• Base every decision only on data available at that exact timestamp. No using the day's close to justify an at-open entry.
• Run the test forward in time, trade by trade, advancing the clock after each decision.
• Do not skip trades after seeing they lost, and do not add trades that were never signaled. The journal is sacred.
• Never re-optimize rules against the paper journal itself. Fitting parameters to the same trades you grade yourself on is grading your own homework.

4. Log every trade

Copy assets/trade-journal-template.csv to start. Required fields per trade: date, symbol, side, entry, exit, quantity, fees, the setup name, and a one-line note on whether you followed the plan. Missed signals go in the notes, not in the P&L.

5. Evaluate before risking real money

Run the evaluator:

bash
python3 bin/evaluate_trades.py journal.csv --equity-start=10000

Read the verdict, then read the numbers in the order that matters (references/evaluation_guide.md has the full rubric): trade count first, max drawdown second, profit factor and expectancy third. Return figures read last, because a lucky number inflates them most.

6. Decide, honestly

• PASS: move to small-size live testing at reduced size, keep the journal running in parallel.
• REVIEW: fix the failures, start a new forward test segment. Do not fold the old trades into the new numbers.
• FAIL on sample size: keep logging. A strategy with 20 trades proves nothing.

Output Contract

When the user finishes a paper trading period, deliver:

1. The evaluation summary: trades, net P&L, win rate, avg win/loss, profit factor, expectancy per trade, max drawdown, Sharpe.
2. The verdict (PASS / REVIEW / FAIL) with which rubric checks passed or failed.
3. One concrete recommendation: what to change or what to test next, phrased as the next forward-test segment.
4. The per-setup breakdown showing which setups carried the P&L and which dragged it.

Operating Rules

1. Never invent the user's results. If they hand you a journal, evaluate what is in it. If they want advice, work from what is in it.
2. Never certify a strategy as "safe" or "proven". Paper results are evidence, not proof, and say so.
3. Frame thresholds as guidance from the rubric, not guarantees. Markets change, drawdowns cluster, slippage is real.
4. Do not skip the sample-size check. Small samples are the most common false positive.
5. If the user wants to change strategy rules mid-test, require a new dated test segment with new pre-registration.
6. Style: plain direct language, no em dashes, no motivational filler. Numbers over adjectives.