Combining Indicators

Combining Indicators

Every indicator produces false signals. An oscillator flags "oversold" in markets that keep falling; a crossover fires at the start of moves that never develop. Combining indicators means requiring more than one signal to agree before acting -- one indicator proposes the trade, another confirms the conditions around it. Done well, it filters out a disproportionate share of the bad signals. Done badly, it either filters nothing or filters everything.

This article is about the difference: what makes a combination genuine confirmation rather than an echo, how to express combinations in Simula, and what each added condition costs you.

Confirmation vs. Redundancy

The most common combining mistake is stacking indicators that measure the same thing. RSI, Stochastic, and CCI are all transformations of recent price movement -- when one reads oversold, the others usually do too. Requiring all three to agree feels rigorous, but their agreement is close to automatic. You have added parameters and complexity without adding information.

Real confirmation comes from combining indicators that answer different questions. Indicator families exist precisely because there are different questions to ask:

A useful rule of thumb: pick at most one indicator per row. A trend gauge plus a timing oscillator is a combination; two timing oscillators is a redundancy. The full category tour lives in the trend, momentum, oscillator, and volatility family articles.

How to Combine in Simula

Combinations are built with AND logic: add each requirement as its own entry condition and set the combinator to AND, so the entry only fires on a candle where every condition holds at once. The natural pattern pairs a continuous condition (the context: "price is above the SMA(200)" stays true for as long as it holds) with an event condition (the trigger: "Stochastic crossed under 20" fires on one candle). The event picks the moment; the continuous condition decides whether that moment is in the right kind of market.

Two extensions worth knowing. Condition groups let you build multi-setup strategies like "(pullback setup) OR (breakout setup)" where each setup is its own AND-block. And the two sides of a condition can live on different timeframes -- "act on the 1-hour chart only when the 4-hour trend agrees" -- which is multi-timeframe analysis, confirmation across time horizons rather than across indicator families.

Two Worked Patterns

Trend Filter + Oversold Timing

Entry: price greater than SMA(200), AND Stochastic %K crossed under 20. The oscillator alone would buy every dip -- including every stage of a downtrend, where dips are just the trend continuing. The moving-average condition restricts it to dips inside rising markets, which is where a dip has the larger structure working for its recovery. This is the regime filter from the mean reversion article, expressed as one extra AND condition.

Signal + Strength Filter

Entry: MACD crossed over its signal line, AND ADX greater than 25. MACD crossovers happen constantly in quiet markets, where the two averages hug each other and cross on noise. ADX measures whether a directional move of any real strength is underway -- gating the crossover on it discards the crossings that happen inside drift. Here the two indicators split the work: MACD picks the direction and the moment, ADX vouches that there is something worth catching.

Every Filter Has a Price

Each AND condition removes trades -- that is its job. But it removes good trades along with bad ones, and it shrinks the sample your conclusions rest on. A strategy that took 200 trades before a filter and 40 after it had better show a much higher quality per trade, because 40 trades is thin evidence for any statistical claim. Filters also add parameters -- the SMA period, the ADX threshold -- and every parameter is another knob that can be fitted to the past instead of to anything real.

In practice the returns diminish quickly. The first well-chosen filter tends to do real work, cutting signals in hostile conditions. By the fifth condition, the strategy is mostly memorizing the historical dataset: each extra requirement carves away specific past trades, and the more precisely the rule set describes what happened before, the less it describes what happens next. Two or three conditions that each answer a different question beat six that jointly describe one backtest.

Check What Each Condition Contributes

Simula's results include a condition performance breakdown showing how trades attributed to each condition (and each combination of conditions) performed -- see Reading Your Results. It turns pruning into an evidence-based exercise: a condition whose trades perform no better than the rest of the strategy is not earning its parameters, and removing it simplifies the strategy while growing the trade sample. If you want to test threshold variations systematically instead of one manual run at a time, parameter optimization sweeps the grid in one job.

Common Mistakes

Stacking one category. Three oscillators that agree are one opinion expressed three times. Combine across the table's rows, not within one.

Filtering after the fact. Filters designed to delete known losing trades describe the past, not the future.

Requiring unanimous agreement. Five simultaneous conditions can shrink a strategy to a handful of trades per decade -- unmeasurable and untradeable. If the trade count collapses, remove the weakest condition rather than extending the date range and hoping.

Confusing more complexity with more edge. The goal of combining is fewer, better trades justified by independent evidence -- not a strategy that is harder to explain.

Where to Go Next

Entry & Exit Conditions -- AND/OR logic, condition groups, and how multi-condition signals become trades.

Multi-Timeframe Analysis -- confirmation across timeframes instead of across indicators.

Trend Following and Mean Reversion -- the two base strategies these combinations are usually built on.

Overfitting & Curve Fitting -- why every added parameter makes the backtest less trustworthy.

Statistical Significance & Trade Count -- how many trades you need before a filtered strategy's results mean anything.