Testing Strategies Across Bull, Bear, and Sideways Markets

A strategy can look impressive across three years of historical data while quietly depending on one six-month bull run for most of its gains.

That is one of the problems a full-history backtest can hide.

A single aggregate result answers:

How did this strategy behave across the entire historical period?

Regime testing asks a more specific question:

Where did that performance actually come from?

To check whether a strategy behaves differently across market conditions, divide historical data using a consistent regime definition, evaluate the strategy across those conditions, and compare not only return but also trade frequency, drawdown, win/loss behavior, and dependence on individual periods.

The goal is not to prove that a strategy will work in the next bull, bear, or sideways market.

It is to understand whether the historical result is broad or heavily dependent on one type of environment.

Why One Full-History Backtest Can Be Misleading

Suppose a strategy produced a strong positive result across a three-year backtest.

That number alone does not tell you whether the strategy:

• performed reasonably throughout the entire period,
• made nearly all of its gains during one strong trend,
• lost money during sideways markets,
• stopped generating trades during bearish periods,
• or recovered one large drawdown because of a short burst of exceptional performance.

Those are very different historical stories.

An aggregate backtest compresses them into one result.

A simple hypothetical example

Consider this hypothetical strategy:

Looking only at the full-history return, you might conclude that the strategy was broadly successful.

The regime breakdown tells a more specific story:

historically, the strategy's positive result depended heavily on bullish periods.

That does not automatically make the strategy bad.

It changes the conclusion you should draw from the evidence.

Instead of:

"This strategy works."

a more accurate statement would be:

"Historically, this strategy performed substantially better during bullish conditions than during bearish or sideways periods."

That distinction is the main reason to test regimes separately.

What Is a Market Regime?

A market regime is a period characterized by a broadly similar type of market behavior.

Common labels include:

• bullish — sustained upward directional behavior,
• bearish — sustained downward directional behavior,
• sideways or ranging — limited directional progress,
• high volatility — relatively large price movement,
• low volatility — relatively compressed movement.

These categories can overlap.

For example, markets can be:

• bullish and highly volatile,
• bearish and low volatility,
• sideways and volatile.

That is why "bull, bear, sideways" should be treated as a research framework rather than a perfect description of market reality.

There Is No Universal Bull/Bear/Sideways Definition

One of the first mistakes in regime analysis is assuming there is one objectively correct classification method.

There is not.

Different methodologies can reasonably produce different regime boundaries.

Trend-based classification

A trend-based rule may classify markets using:

• price relative to a moving average,
• moving-average direction,
• moving-average slope,
• or combinations of short- and long-term trend measures.

For example, a researcher might define a bullish period as one where price remains above a rising moving average.

A bearish period could use the opposite condition.

Periods that fail either directional test could be classified as sideways.

The exact rule is less important than applying the rule consistently.

Fixed lookback return

Another approach classifies regimes according to price movement over a predefined historical window.

For example, the researcher could ask:

How much has price changed over the previous N candles?

Large positive changes could represent bullish conditions, large negative changes bearish conditions, and smaller moves sideways conditions.

Again, the thresholds are research choices.

Trend strength and volatility

A richer classification can combine direction with measures of trend strength or volatility.

This can distinguish between conditions such as:

• strong directional trend,
• weak directional trend,
• quiet range,
• volatile range.

This may be more informative than forcing every historical period into one of only three labels.

But complexity also introduces more choices.

Those choices should be justified rather than tuned until the strategy looks good.

Define the Regime Rule Before Looking at Strategy Performance

This is one of the most important disciplines in regime testing.

Imagine you first inspect the backtest and notice that the strategy performed poorly from April through June.

You then modify the regime definition until April through June becomes classified as a special "transition" regime.

The result now looks cleaner.

But the classification was influenced by the strategy result you were trying to explain.

That introduces researcher bias.

A better sequence is:

1. Decide how regimes will be classified.
2. Record the rule.
3. Apply it consistently.
4. Only then compare strategy performance.

This does not remove every source of bias.

But it makes the analysis much harder to manipulate unintentionally.

Avoid Look-Ahead Bias When Classifying Regimes

There is an even more serious problem than changing definitions after seeing results.

Your regime classifier must not use future information to label the present.

Suppose you label a historical candle as "bear market" because price eventually fell 25% over the following three months.

That classification may be useful for describing history after the fact.

But it would not have been knowable at the time.

If you then evaluate a strategy as though the bear regime was already identified at the beginning of the decline, you have introduced look-ahead bias.

The regime label contains future information.

A safer way to think about regime classification

At historical time T, ask:

What information would have been available at T?

A trailing moving average, previous-period return, or historical volatility calculation may be available.

A future drawdown is not.

This distinction matters especially if the regime label influences strategy decisions rather than being used only for retrospective analysis.

A Practical Regime-Testing Workflow

A disciplined workflow can look like this.

Step 1: Define the strategy hypothesis

Before looking at regime results, write down what the strategy is supposed to do.

Examples:

• follow sustained momentum,
• capture breakouts,
• fade short-term extremes,
• trade only when broader trend conditions are favorable.

That gives you something meaningful to compare against the later regime results.

Step 2: Define the regime rule

Choose the historical information used to classify markets.

Document:

• the indicator or calculation,
• the lookback window,
• thresholds,
• treatment of ambiguous periods,
• and whether transition periods receive their own category.

Do this before inspecting which definition makes the strategy look best.

Step 3: Classify the historical data

Apply the same rule across the entire dataset.

Avoid manually re-labeling inconvenient sections after inspecting results.

If the rule creates awkward classifications, document that limitation rather than silently correcting them with hindsight.

Step 4: Evaluate strategy behavior by regime

Compare more than total return.

Useful metrics include:

Do not expect every metric to improve simultaneously.

The goal is to understand behavior.

Step 5: Compare multiple occurrences of the same regime

One bull market is not enough to establish how a strategy behaves in "bull markets."

Likewise, one bear market does not define all bearish conditions.

This is a major limitation in crypto research because the historical record is relatively short compared with many traditional asset classes.

If possible, ask whether the same general behavior appears across:

• multiple bullish periods,
• multiple bearish periods,
• multiple sideways periods.

A strategy that worked in one particular bull cycle may have captured characteristics unique to that cycle rather than a broader regime effect.

Step 6: Compare the aggregate result with the regime breakdown

Now ask:

Would I have interpreted the full backtest differently if I had seen the regime breakdown first?

That question often exposes hidden dependency.

Compare Behavior, Not Just Return

Suppose two strategies produce the following hypothetical results during sideways markets:

Strategy A

• return: -3%
• 12 trades
• low activity

Strategy B

• return: -3%
• 160 trades
• persistent small losses

The headline return is identical.

The behavior is not.

Strategy A may simply have few opportunities in sideways conditions.

Strategy B appears to repeatedly engage with an environment that historically did not suit it.

That distinction could matter when deciding how to monitor or further research the strategy.

Regime testing should therefore describe what the strategy does, not only whether a column is positive or negative.

Be Careful With Regime Boundaries

Segmenting historical data introduces another subtle problem.

A strategy does not necessarily reset itself when a market regime changes.

Imagine a position opens near the end of a bullish period and exits after your classification has switched to sideways.

Which regime owns that trade?

There is no universally correct answer.

Possible approaches include:

• classify according to the regime at entry,
• classify according to the regime at exit,
• classify according to where most of the holding period occurred,
• or analyze transition trades separately.

The important point is consistency.

If regime-specific analysis depends heavily on how boundary trades are handled, that itself is useful information about the fragility of the conclusion.

This is also why simply cutting historical data into disconnected segments and rerunning the strategy can produce different behavior from one continuous backtest.

Strategy state, open positions, indicators, and prior candles may cross those boundaries.

Transition Periods Deserve Special Attention

Markets rarely switch cleanly from:

bull → sideways → bear

at one obvious candle.

Transitions can be messy.

A trend-following strategy may perform poorly during the early stages of reversal because the previous trend still influences its signals.

A mean-reversion strategy may behave differently as a quiet range begins expanding into a trend.

If a large share of losses occurs near regime transitions, forcing those candles into one of three clean categories can hide useful information.

You do not necessarily need a dedicated transition regime.

But you should at least ask:

Are the worst results occurring inside established regimes, or during the movement between them?

That question can reveal more than simply labeling the strategy "bad in bear markets."

Testing Across Symbols Can Strengthen or Weaken the Conclusion

Suppose a strategy appears strongly bull-dependent on Bitcoin.

What happens on other relevant crypto assets?

If the same general pattern appears repeatedly, the regime dependency becomes more interesting.

If it disappears completely on every other dataset, the conclusion may be more specific to Bitcoin than to bullish regimes generally.

Cross-symbol testing should therefore ask:

Does the underlying relationship appear elsewhere?

Not:

Does the exact same strategy produce identical returns everywhere?

Those are different standards.

Timeframes Can Change What a "Regime" Means

A market can be bullish on one timeframe and sideways on another.

For example:

• the daily chart may show a sustained uptrend,
• while the hourly chart contains multiple ranges and short bearish moves.

So "bull market" is incomplete without a timeframe.

When testing across timeframes, keep the regime definition tied to the resolution being analyzed.

Do not assume that a label derived from daily data necessarily describes every lower-timeframe trade in the same way.

Common Regime-Testing Mistakes

Choosing the regime rule after seeing the strategy results

This makes the classification part of the optimization process.

Define it first.

Using future information

A regime label that depends on what happened afterward may introduce look-ahead bias.

Use information that would have been available at the point being classified when the analysis requires real-time regime identification.

Treating one bull market as representative of all bull markets

One historical episode is still one sample.

More regime transitions provide stronger evidence.

Forcing every candle into three categories

Transition periods and mixed conditions may not fit cleanly into bull, bear, or sideways labels.

Do not create false precision.

Ignoring boundary effects

Trades and strategy state may cross regime boundaries.

Document how they are handled.

Optimizing the strategy separately inside every regime until each one looks good

This can simply create three overfit strategies instead of one.

Regime analysis should challenge the strategy, not automatically become another excuse to search for better historical parameters.

Treating regime testing as prediction

Historical regime analysis tells you how a strategy behaved under your classification of past conditions.

It does not tell you with certainty what regime exists now or what comes next.

What If the Strategy Only Works in One Regime?

That is not automatically a failure.

Some strategies should be regime-dependent.

A breakout or trend-following strategy may historically behave better during sustained directional movement.

A mean-reversion strategy may historically behave better during ranges.

The useful discovery is not:

My strategy failed the regime test.

It may instead be:

My strategy appears historically dependent on a particular environment.

That changes how the strategy should be described and investigated.

A regime-dependent strategy requires a stronger understanding of:

• when it historically struggled,
• how large those losses became,
• how often unfavorable regimes occurred,
• and whether the intended monitoring process can identify relevant conditions without hindsight.

A Practical Regime Review Checklist

Before drawing conclusions from a regime-segmented backtest, ask:

• Did I define the regime rule before seeing strategy performance?
• Does the classifier use only information that was available at the time?
• How many separate examples of each regime exist in the dataset?
• Are most of the strategy's gains concentrated in one regime?
• Does one regime produce most of the drawdown?
• Does trade frequency change materially across regimes?
• Are results consistent across multiple occurrences of the same regime?
• Do transition periods explain a meaningful share of poor performance?
• Could regime-boundary treatment materially change the conclusion?
• Does similar behavior appear on other relevant symbols or timeframes?
• Am I describing historical regime dependence rather than predicting future performance?

If the answer to several of these questions exposes fragility, the regime analysis has done its job.

It has shown you where the backtest deserves more investigation.

Applying This Research Method in PulseStats

PulseStats is designed to help users move from a trading idea to explicit rules, historical testing, parameter research, saved results, and ongoing monitoring without requiring strategy code.

Regime testing can fit into that workflow without requiring the product to issue a "bull," "bear," or "sideways" verdict automatically.

A practical research flow is:

strategy idea → explicit rules → historical backtest → regime hypothesis → comparison experiments → optimization where appropriate → saved research → monitoring

Start with explicit rules

Use the Strategy Builder to define the strategy you actually want to investigate.

That can include configurable factors such as:

• momentum or RSI,
• moving-average trend conditions,
• market regime factors,
• volume,
• breakout behavior,
• and execution settings.

Defining the logic first gives the later regime comparison a clear hypothesis.

Run the baseline historical backtest

Start with the strategy across the full historical period.

Understand the trade-level and strategy-level result before trying to explain it through regimes.

Ask:

• How many trades occurred?
• Where did drawdowns appear?
• Were results evenly distributed?
• Did one historical period dominate?

Structure deliberate comparison experiments

PulseStats supports backtesting across supported crypto symbols and timeframes.

That allows you to run additional experiments when those comparisons are relevant to the regime hypothesis.

For example, if a strategy appears dependent on strong directional conditions, you can compare its historical behavior across different datasets rather than relying on one asset and one timeframe.

The purpose is not to keep searching until another attractive result appears.

It is to challenge the original conclusion.

Use optimization carefully

If you later optimize strategy or execution parameters, do not automatically optimize each regime independently until every segment produces attractive results.

That can turn regime analysis into another source of overfitting.

Treat optimizer outputs as historical candidates and preserve the distinction between:

finding a parameter combination

and

validating whether its behavior is robust.

Save the research

Saved backtest and optimizer results make it easier to revisit previous experiments without relying on memory.

That becomes especially useful when comparing:

• different symbols,
• different timeframes,
• different strategy configurations,
• or research performed at different stages.

Monitor explicit conditions after research

Once you understand where a strategy historically behaved well or poorly, PulseStats' scanner and strategy-based alerts can help monitor explicit setup conditions.

That does not mean PulseStats knows which market regime will come next.

It means you can move from repeatedly checking charts manually toward monitoring the conditions you deliberately defined.

The Main Principle: Find Out Where the Strategy Works — and Where It Does Not

A full-history backtest can make a strategy look more universal than it really is.

Regime analysis forces a better question:

What market conditions actually produced this historical result?

Sometimes the answer will be encouraging.

Sometimes it will reveal that almost all of the performance came from one environment.

Both outcomes are useful.

The goal of regime testing is not to make every strategy perform well in every market.

It is to replace a vague conclusion like:

"The backtest looks good."

with a more precise one:

"Historically, this strategy behaved this way under these conditions, and differently under others."

That is a much stronger foundation for systematic strategy research.