DAY 33: How to revise rules based on verification results

DAY 32, we learned the importance of assessing not only win rate, but also multiple elements such as drawdown and the profit factor (PF) in a comprehensive manner.Indicators to Look at in Validationand more: the importance of judging not just win rate, but also drawdown and PF (profit factor) as part of a holistic assessment.
Today, as DAY 33, we will explain the theme of how to specifically modify the rules based on validation results.
When trying to improve a method, it is not uncommon to add indicators one after another or fall into the “optimization trap.” Therefore,we will organize the steps and cautions for gradually updating the logic..

1. Why do rule modifications become necessary?

1. Market environment changes

   • Laws that worked in the past can become difficult to profit from due to changes in monetary policy, global conditions, and volatility.
   • Movements in stock prices, commodities, and interest rates affect the FX market, making the same method less profitable.

2. Insights from validation results

   • In backtests or forward tests, if there are unexpectedly large drawdowns or a string of losses, it signals that some conditions should be reviewed.
   • Win rate might be decent, but the risk-reward ratio is poor and profitability is hard to sustain—leaving room for improvement.

3. Mental state and lifestyle changes

   • Working hours or mental conditions to trade can change due to job or family circumstances.
   • As a result, adjustments such as a more swing-oriented approach or an automated exit system before indicators may be necessary.

2. Points to be careful about in rule modifications

(1) Don’t change too much at once

• If you add three or four indicators at once and head in a direction believing you won’t lose, the risk of over-optimization increases dramatically.over-optimization
• The basic approach is to make small changes one at a timeand revalidate before moving to the next step.

(2) It’s important not only to “make it easier to win” but also to “limit losses”

• Include measures to reduce drawdown, such as adjusting stop loss levels and lot sizes.
• Often, adjustments that reduce large losses are more effective than simply increasing win rate.

(3) Be cautious when adding new indicators

• Increasing signals might reduce false alarms, but it can also reduce the number of trading opportunities, risking missed chances.
• Be clear about why you are using the indicator (e.g., using Bollinger Bands plus RSI to refine timing for contrarian entries).

(4) Don’t judge by appearance; always revalidate

• Even if you intuitively feel that adding MACD during a certain period would have created entries, do not implement based on sensation alone.
• Partial optimizations are likely to be counterproductive in other periods or market conditions.
• Introduce carefully through out-of-sample validationand forward testing while testing in practice.

3. Example steps for rule modification

1. Identify the causes of underperformance

   • From validation data or real results, identify concrete causes such as “drawdowns increase when range persists” or “big losses cluster around indicator moments.”
   • For example, “the stop loss width is too wide/narrow,” or “take profits are too early during trends.”

2. Form improvement hypotheses

   • “Do not enter new trades in the 30 minutes before indicators,’”
   • “If the market is range-bound, stop EA or reduce lot size manually,”
   • “Extend take profit from Fibonacci 61.8% to 161.8% expansion,”

3. Small-scale revalidation

   • BacktestApply the new rule to historical data and see if win rate and drawdown improve.
   • Forward testRun on a demo or small lots for 1–3 months to check real-market benefits.

4. Evaluate results and fine-tune further

5. Regular review

   • Even after finalizing a rule, major market changes may require adjustments.
   • Make a habit of reviewing performance metrics (profit/loss, win rate, drawdown) periodically (weekly, monthly) to identify improvement points.

4. Concrete modification examples

(1) Adjust the stop loss width

• ChallengeDrawdowns are large due to frequent losses.
• HypothesisStops are too wide causing a single loss to weigh heavily, or too tight causing noise to whipsaw.
• Practice:
  • Example: widen stop loss from recent high/low by 10 pips to 5 pips beyond.
  • Recalibrate using 2x ATR (Average True Range).
• Revalidate:
  • Backtest and forward test with small lots to see changes in win rate and average loss.

(2) Add an indicator to filter entries

• ChallengeMany false signals in range markets.
• HypothesisIf we gauge breakout reliability with volatility or oscillators, tightening entries might improve win rate.
• Practice:
  • For example, “only breakout after a Bollinger Band squeeze” or similar conditions.
• Caution:
  • Be wary of drastically reducing trade frequency or falling into over-optimization traps.

(3) Introduce entry restrictions by indicators or time zones

• ChallengeBig losses around major announcements and strange losses during late-night low-liquidity periods.
• Hypothesis“Avoid high-risk times to reduce drawdown.”
• Practice:
  • Examples: “Do not enter new trades 30 minutes before/after major indicators,” “Pause auto-trading after 23:00 Japan time,” etc.
• Revalidate:
  • Compare results when excluding entries around indicators in backtests, then test further in forward testing.

5. How to avoid falling into the “optimization trap”

1. Use out-of-sample validation

   • Lock in conditions with in-sample (validation period), then try the exact same conditions on a different period (out-of-sample).
   • If you don’t win at all, you’re likely too closely aligned to past data.

2. Don’t overcomplicate conditions

   • For example, aligning five or six signals from MA, RSI, MACD, Stochastic, Bollinger Bands, etc. may create scenarios where almost nothing meets the conditions.In some cases, simple rules with minimal essential adjustments are more stable in the long run.
   • Keep expectations realistic about losses

3. Even after many fixes, aiming for zero drawdown or over 90% win rate leads to large losses when sudden market moves occur in real conditions.

   • The orthodox path is a steady, consistently profitable logic through the year even if modest.

4. Forward testing before going live

   • Instead of pouring a large amount of capital immediately after modifications, recheck with small lots or a demo to ensure it fits real markets.
   • Don’t judge from a few profit/loss moments; gather a reasonable sample size.

6. Summary & next forecast

Summary

1. Rule modification based on validation results is a fundamental process to brush up a trading method..
2. Things to be careful about:
   • Make small changes gradually (avoid over-optimization).
   • Adjust comprehensively: not only win rate, but drawdown reduction and avoidance of unfavorable market conditions.
   • Add indicators cautiously, and always go through revalidation and forward testing.
3. To prevent optimization trapsUse out-of-sample data, avoid overly complex conditions, and validate with forward testing against real market data.

Next time (DAY 34) theme: How to run a validation cycle that connects to practice

• We will further develop the rule-modification mindset learned here and introduce concrete examples of the PDCA cycle—how to cycle validation, implementation, and re-validation.
• Trading isn’t a one-off rule setting; it’s about continuously updating to match market conditions and your situation. Stay tuned for next time!

If you’re interested in automated trading, please also check this out below.
https://www.gogojungle.co.jp/users/147322/products

If this was helpful, I would be grateful if you click "Read more."
Thank you.