Evidence, Confluence and Indicator Discipline

First Officer · C03: Indicators · About 25–30 minutes

Safety rule: Indicators describe calculated past observations. Work out risk before treating an indicator reading as evidence.

Your mission today

By the end you can judge whether adding an indicator improves a stated rule rather than merely making the chart look complex.

Before you start

Complete C02 and open a demo chart with a visible, fixed indicator setting. Use a calculator for the imaginary examples.

1. The idea in everyday words

Four friends repeating one rumour do not make four independent sources. Confluence means several conditions point to a related interpretation, but RSI, MACD and moving averages can recycle the same underlying prices. More matching colours do not automatically add information.

A useful filter should reduce a specific failure mode, not simply make the chart appear more technical. Confluence is valuable when different information sources change the probability or payoff distribution; four derivatives of the same closing prices may add virtually no information.

2. Worked example: what could be lost?

Imagine a simulated EURUSD demo method has 200 trades with average net result −$2 per trade. Adding a new filter keeps only 12 trades and produces +$5 average in the same historical sample. That result may be selection noise rather than a new edge.

For a separate arithmetic check, suppose each of those twelve hypothetical trades risks $10 on a loss but targets $3 on a win. Ten wins and two losses would yield 10 × $3 − 2 × $10 = +$10 before any unmodelled costs. The small sample and this one trade sequence establish neither future profitability nor a reliable EURUSD signal.

Original Black Box Journey schematic. Synthetic teaching example, not a historical strategy result.

3. How the calculation or observation works

Write one hypothesis before testing: for example, 'a volatility threshold may exclude low-movement breakout failures'. Specify the indicator, timeframe, threshold, price inputs and entry/skip rule. Compare results with and without the filter on identical chronological data and costs.

Technical detail: Write the hypothesis before testing an indicator: for example, a volatility filter might exclude dead ranges that previously stalled breakout trades. Compare trades with and without the filter on identical data and realistic costs. Reserve an untouched period for evaluation, and reject filters that leave too few independent observations.

4. Know the limitation

Keep an out-of-sample period untouched when choosing the filter. Test sensitivity: if changing RSI threshold from 59 to 60 destroys the apparent benefit, the rule may be fragile. Do not add twenty clauses after seeing twenty losing trades and report the fitted result as if it were expected in live markets.

Adding rules after every losing trade progressively shrinks frequency and overfits history. A high in-sample win rate without adequate sample and payoff distribution proves very little.

5. Practise independently

Create one strategy card with no more than two indicator conditions. Run a paper comparison on a clearly labelled fabricated dataset, counting all eligible and rejected cases. Write an acceptance rule before looking at the outcome.

This exercise demonstrates the concept, not a verified earnings edge. Save all qualifying examples and failures.

6. Keep indicators from counting the same evidence twice

Synthetic EURUSD example: An H1 chart is trending upward, and the M15 RSI is above 60. A second indicator, MACD, also rises because both react to many of the same recent closes. Treating this as two independent confirmations exaggerates the amount of new information. First write down exactly what each indicator measures, its lookback setting and the decision it changes.

Imagine your plan needs one trend description and one volatility measure. You could compare an EMA slope against ATR instead of adding four momentum indicators that all follow recent closes. This still does not validate a strategy. The useful question is whether the added rule improves a properly sampled decision process after costs, not whether the chart looks more convincing.

Suppose an imaginary long at Ask 1.08520 with a Bid stop at 1.08420 would lose 10 pips, or −$10 at $1 per pip before fees. A later Bid at 1.08670 would create +15 pips in one hypothetical path. Neither indicator removes the possibility of the planned loss.

Practice record: Run ten closed-bar replay cases with one indicator, then the same ten with a second. Keep entry and skip rules fixed. Record any changed decisions, duplicated information and the first timestamp each reading was available.

Common mistake

Treating an indicator label as an entry permission without knowing its calculation, decision timestamp, costs or failure condition. Fix it by stating what the indicator measures, what it cannot establish, and what your written risk rule requires.

Plain-English glossary

• Confluence: Several qualifying observations considered together.
• Overfitting: Choosing rules that mainly match past noise.
• Out-of-sample: Data held back from rule design.
• Expectancy: Average result per trade under stated outcome and cost assumptions.
• Filter: A rule that accepts or rejects candidate entries.

Try it yourself

Create one strategy card with no more than two indicator conditions. Run a paper comparison on a clearly labelled fabricated dataset, counting all eligible and rejected cases. Write an acceptance rule before looking at the outcome. Save the parameters, chart timestamp and worked arithmetic in your journal.

What you can now do

• Judge whether adding an indicator improves a stated rule rather than merely making the chart look complex.
• Explain one important indicator limitation without predicting returns.
• Make a risk-aware and reproducible decision to proceed, wait or skip.

Next lesson

C04.1 · Supply and Demand Zones.