Optimization Without Curve-Fitting + Writing Your Trading Plan

The last chapter — once the system has passed the backtest + forward test, you will want to make it "even better."

Improvement (optimization) is a good thing — but a dangerous trap is hidden inside it, named curve-fitting.

What is curve-fitting — the cleverest trap

Curve-fitting (or over-optimization) = adjusting the system to "fit exactly" the historical data — until it works only on that past, but is unusable on the future.

The clearest example:

You backtest a system, get Win 55% — you want it better, you start adjusting: - "If I change the EMA from 50 to 47 → Win becomes 58%" - "If I enter only on Tuesday-Thursday → Win becomes 64%" - "If I do not trade in August → Win becomes 71%" - "If the SL is exactly 23 pips → Win becomes 79%!"

You are thrilled — a 79% win-rate system!

But you have just built a system that "memorized the answers to an old exam" — not a system that "understands the subject."

EMA 47 (not 50), avoiding August, an SL of exactly 23 pips — these numbers have no logical reason, they just "happened to fit" that particular past.

When the future arrives — not exactly like the past — this 79% system collapses instantly.

Warning signs that you are curve-fitting: - The system has very many rules (10+ conditions) and several are "oddly specific" - Parameter values are strange numbers (EMA 47, RSI 31.5) instead of round numbers - There are many exceptions ("except Fridays," "except this month") - An abnormally high win rate (> 75%) in the backtest - A tiny adjustment changes the result a lot (a fragile system)

Correct optimization — 3 principles:

1. Adjust with "logic," not "the number that gives the best result" Ask "why" — if you widen the SL, there must be a reason (e.g. "because the old SL got hit by noise often"), not "because 23 pips gives the highest win rate."

2. Use an out-of-sample test Split the data into 2 parts: adjust the system on part 1 (in-sample) → test on part 2, which it has never seen (out-of-sample). If the system still works on the out-of-sample = genuinely robust · if it collapses on the out-of-sample = curve-fitted.

3. Prefer a system that is "simple and durable" over "complex with a high win rate" A Win 55% system with 4 plain rules and clear logic is better than a Win 79% system with 12 rules and exceptions everywhere. A simple system withstands a changing market better — because it captures "principles," not "the details of the past."

> The goal is not "the best system in the past" — but "a system still good enough in an unknown future"

---

Pulling it all together — write a complete Trading Plan

A Trading Plan = one document that gathers everything, the "constitution" of your trading, written with a cool head, enforced with a hot one.

The structure of a Trading Plan — 7 parts:

1. The system (the 5 components, from Chapter 2) — Market/TF, entry checklist, SL, TP, risk sizing

2. The edge numbers (from Chapters 1+3) — win rate, expectancy, profit factor, max DD from the backtest

3. Risk rules (from the Risk Management ebook) — risk/trade, daily stop, weekly stop, max correlation

4. Discipline rules (from the psychology ebook) — trading hours, the physical condition in which you will trade, a circuit breaker after a loss

5. Routine — before trading (what to check), during, after trading (journal)

6. Review rules — review the journal every week, revisit the system every 100 trades

7. System-change rules — never change the system mid-way · you may change it only when you have evidence over 100+ trades that the edge is genuinely gone

Part 7 is the most important — it is your armor against "setup hopping" (changing the system every time you lose a few in a row), which kills more traders than anything.

---

Summary of this ebook — building a system with an edge in 5 chapters:

Chapter 1 — an edge is a number · positive expectancy = an edge · a high win rate is not enough Chapter 2 — the 5 components · a system must be complete: Market/Entry/SL/TP/Risk · rules must be binary Chapter 3 — backtest 100-1,000 trades · prove it with numbers · honesty is the iron rule Chapter 4 — forward test · demo → small live → gradually increase the size Chapter 5 — optimize with logic, not curve-fitting · write a Trading Plan

Final words from me:

A beginner has a new "system" every month, because he never had a real system to hold on to — he has only a feeling, which changes every time he gets hurt.

A professional has one system, used for years, tuned slowly with data — because he has the "numbers" that tell him it has an edge, and he trusts the numbers more than he trusts his emotions.

You have read all 5 chapters — you know how to build a system, prove it, and protect it from emotional change.

Now there is only one step left — take action. Open TradingView, open Bar Replay, start backtesting your first system today.

A "good enough" system that you have backtested, believe in, and can follow — is better than a "perfect" system you merely read about on YouTube yesterday.

May you have a system that is truly your own