How to Convert Discretionary Observations Into Systematic Trading Rules

Why Discretionary Patterns Are Hard to Replicate

Experienced discretionary traders frequently identify patterns that have genuine edge before they can articulate what the pattern actually is. "It looks like it wants to go higher" is a real signal — experienced market readers integrate many simultaneous data points — but it is not a tradeable rule. You cannot backtest it, teach it, automate it, or reliably execute it during emotionally charged live trading conditions.

The process of converting a discretionary observation into a systematic rule is one of the highest-value activities a developing trader can do. It forces precision — you must define exactly what "looks like it wants to go higher" means in measurable terms. It enables testing — once defined, the rule can be evaluated against historical data. It enables scaling — systematic rules can be automated, removing execution emotion from the equation.

This is the core of what Cameron Bennion has built with YMI's KPL methodology: a system that started as experienced market observation ("price tends to react here") and was refined through rigorous definition work into the Key Price Level algorithm — systematized, backtested, and automatable.

Step 1: Document the Observation Precisely

Start by writing the observation in plain language with as much specificity as possible, including all the contextual factors that seem relevant. Avoid the temptation to make it sound sophisticated — clarity matters more.

Bad observation: "ES rallies after big down moves." This is too vague to test — how big is "big," how long is "after," how much is a "rally"?

Better observation: "On days when ES opens more than 15 points below the prior close (gap down open), price tends to recover at least 50% of the gap within the first 90 minutes of the regular session."

Best observation: "On days when ES opens 15+ points below the prior regular session close, and the prior day had a 2:1 or higher volume day (suggesting institutional distribution), and the overnight Globex session did not reach the opening gap level — price recovers 50%+ of the gap within the first 90 minutes, with 68% frequency based on informal observation of approximately 20 recent instances."

The "best" observation includes: the specific condition (15+ point gap), contextual filters (prior day volume, Globex behavior), the expected behavior (50%+ gap recovery), the timeframe (90 minutes), and the approximate frequency (68%). This is still informal, but it is specific enough to build a test from.

Step 2: Define Measurable Entry, Stop, and Target Criteria

For an observation to become a rule, every component of the trade must be precisely defined. For the gap recovery example:

Entry trigger: "Buy the first 5-minute bar that closes above the prior day's close minus 7.5 points (50% of the minimum 15-point gap), after the initial open." This is specific: a bar close above a mathematically defined level.

Stop loss: "Below the opening bar's low, with a minimum of 4 ES points." This protects against the gap not recovering while allowing normal post-open volatility.

Target: "At the prior regular session close price (100% gap fill), or at a 2:1 reward/risk minimum." This is specific and measurable.

Time filter: "Entry must occur within 60 minutes of market open. If the entry trigger is not triggered by 10:30 AM ET, skip the trade." This prevents chasing a gap fill attempt that materializes too late to meet the entry criteria.

Notice that every element is now a yes/no question during live trading: Did the gap exceed 15 points? Yes. Did the 5-minute bar close above the 50% level? Yes. Is it before 10:30 AM? Yes. Each filter either confirms or disqualifies the trade — there is no room for "looks like it might work."

Step 3: Define Filters That Disqualify the Setup

Equally important to the entry criteria are the conditions that disqualify the setup. Most patterns work in specific market contexts and fail in others. Identifying the disqualifying conditions is often the hardest part of the systematization process because it requires admitting the pattern doesn't work universally.

For the gap recovery setup, potential disqualifying conditions discovered through testing: (1) High-impact economic release scheduled within 90 minutes of market open (CPI, NFP, FOMC) — these frequently extend gaps rather than fill them due to fundamental news flow. (2) Gap size exceeding 40 ES points — extremely large gaps often reflect significant fundamental events where price discovery is ongoing, making gap recovery less reliable. (3) VIX above 30 — high volatility regimes show lower gap fill rates as institutional traders are less likely to fade the move in risk-off conditions. (4) The gap occurred on a daily chart structure break (e.g., below 200-day moving average for the first time in a trend) — structure breaks may not recover.

Each disqualifying condition reduces the total setups available but improves the win rate and expectancy of the setups that remain. The goal is not to maximize setups taken; it is to maximize expectancy on the setups taken.

Step 4: Test Against Historical Data

With entry criteria, stop, target, and filters defined, test the rule against historical data. In NinjaTrader 8, Market Replay is the most accessible validation tool for discretionary rules — replay 30–50 historical sessions looking specifically for instances that meet all entry criteria, and record the outcome of each trade.

For more systematic approaches, NinjaScript allows coding the rule as an automated strategy that can be backtested across years of data in minutes. If you cannot code the rule yourself, the definition work in Steps 1–3 provides a precise specification for a developer to implement.

What to record during testing: entry price, stop price, target price, whether the trade won or lost, and the maximum adverse excursion (lowest point before resolution for longs, highest for shorts). The maximum adverse excursion data tells you whether your stop is placed correctly — if losing trades consistently reach 80% of the stop distance before recovering, the stop is too tight. If winning trades have very small adverse excursion, the stop may be tighter than necessary, allowing for better reward/risk ratios.

Step 5: Evaluate and Iterate

After 30–50 historical test instances, calculate the expectancy ratio. If positive (above 0.20), the rule has preliminary evidence of edge. If negative or near zero, revisit the definition — are the filters correct, is the entry timing right, is the stop placement causing unnecessary losses?

Iteration is normal. Most rules require 3–5 cycles of definition → testing → revision before reaching a form with genuine measured edge. Common iteration patterns: (1) Tightening filters to improve win rate — removing marginal instances where the setup was present but contextual factors undermined it. (2) Adjusting stop placement based on maximum adverse excursion analysis. (3) Splitting the rule into two variants based on market regime — the gap recovery setup may work at 68% in low-VIX environments but only 45% in high-VIX, suggesting separate rules with separate parameters for each regime.

The iteration process is not overfitting if each change is driven by a logical market mechanism rather than by "what makes the backtest look better." If you add a VIX filter because you understand why institutions are less likely to fade gaps in high-VIX regimes, that's mechanistically justified. If you add a "don't trade on Tuesdays" filter because Tuesday instances happened to lose in your sample, that's overfitting.

Step 6: Document the Rule as a Written Protocol

The final step is writing the rule as a formal protocol — a document that specifies every decision point without ambiguity. A trader (or automated system) following the protocol should produce identical decisions to the original developer, in identical market conditions.

The protocol format used in YMI's strategy development: Rule Name → Setup Definition (the preconditions) → Entry Trigger (the specific action that initiates the trade) → Position Sizing Method → Stop Loss Rule → Target Rule → Time Filter → Disqualifying Conditions → Expected Performance (win rate, average R:R, expectancy from testing) → Known Failure Modes.

This document is the difference between "I have a setup I trade" and "I have a systematized trading rule." The former exists only in your head and degrades under emotional pressure. The latter is portable, testable, improvable, and eventually automatable. Every YMI strategy — the KPL levels, Marty, and the Opening Price strategy — began as a documented protocol of this type before being coded into an automated NinjaScript strategy.