Descriptive Statistics

ID: E-001

Introduction

At the Advanced level, the trader has already learned to interpret market behavior, strategies, market regimes, execution, and performance. The Expert level takes the next step: learning how to measure those observations through structured data.

Statistics matters here not because a professional trader must become a mathematician, but because trading produces many conclusions that can only be evaluated properly through data.

One winning trade does not prove a strategy. One losing trade does not disprove it. A strong month does not necessarily mean the edge has improved, and a weak month does not necessarily mean the edge has disappeared.

The trader therefore does not merely collect results. The trader collects data, structures it, and turns it into information that can be interpreted.

Descriptive statistics is the first layer of that process.

Its purpose is not to predict what the next trade will do. Its purpose is to describe, as accurately as possible, what happened in the observed dataset.

This is one of the central mindset shifts of the Expert level: the first question is no longer “What do I think about this result?” but “What do the data actually show?”

1. What Descriptive Statistics Actually Does

Descriptive statistics is used to organize, summarize, and present the observable properties of a dataset.

In trading, a dataset might be:

500 closed trades from one strategy.

120 occurrences of a specific setup.

A series of daily stock returns.

A portfolio’s daily P&L.

R-Multiple results from executed trades.

The sizes of profits and losses.

Trades executed during a particular market regime.

Raw observations are difficult to interpret when they are viewed one by one. If a trader sees 300 consecutive outcomes such as +1.2R, -0.8R, +2.1R, -1R, +0.4R, and so on, the overall behavior of the system is not immediately obvious.

Descriptive statistics compresses those observations into a form that can be examined. It may show the average outcome, the dispersion of results, the observed range, the frequency of extreme values, where results are concentrated, and how the distribution is shaped.

The boundary is important: descriptive statistics describes the sample that is available. By itself, it does not prove that future observations will behave in exactly the same way. That distinction becomes fundamental later in statistical inference, backtesting, and strategy validation.

2. Population, Sample, and Trading Data

Statistical thinking begins with recognizing that the trader rarely has access to the entire phenomenon being studied.

Suppose a trader wants to know the complete future behavior of a breakout strategy. In theory, the relevant population would include every valid future occurrence of that strategy under all relevant conditions. That population is unknowable in advance.

The trader therefore works with a sample. Five hundred historical trades may form one such sample.

The distinction is fundamental:

Population: all observations that belong to the phenomenon being studied.

Sample: the subset of observations that is actually available for analysis.

Most trading statistics are calculated from samples. This changes how conclusions should be stated. Instead of saying, “My strategy produces +0.32R per trade,” the more disciplined statement is, “In the observed sample, the strategy produced an average result of +0.32R per trade.”

The second statement does not claim more than the data support.

3. Variables and Observations

Trading data becomes easier to reason about when it is viewed as observations and variables. One trade may be one observation. Its properties are variables.

A trading dataset may contain fields such as:

Trade ID

Entry price

Exit price

R-Multiple

Holding time

Maximum favorable excursion

Maximum adverse excursion

Market regime

Setup type

Position size

Risk per trade

A single trade can therefore be described across many dimensions. This matters at Expert level because a strategy’s performance is rarely represented adequately by one number.

Two strategies may both show an average result of +0.30R. In the first, outcomes may cluster tightly around the center. In the second, a small number of very large winners may create most of the positive result while the majority of trades are losing or close to flat.

The average can be identical while the underlying structure is completely different. Descriptive statistics therefore means describing a dataset from several perspectives, not searching for one preferred metric.

4. Center, Spread, and Shape

Three properties of a dataset are especially important: center, spread, and shape.

Center: where the data are located.

Spread: how widely the observations vary.

Shape: how the observations are distributed.

Measures of center describe where results are typically located. Mean and median are two major examples and will be studied in greater depth in E-002.

Spread describes how different the observations are from one another. Two strategies may have the same average return while one produces tightly clustered outcomes and the other produces extreme swings.

Shape describes the structure of the distribution. Is it roughly symmetrical? Does it have a long right or left tail? Are there extreme values? Is there one area of concentration or several?

In trading, this is critical because identical averages can hide completely different risk, behavioral, and execution profiles.

5. The Mean Is a Summary, Not a Complete Description

The arithmetic mean is one of the simplest summary statistics. If there are n observations x₁, x₂, …, xₙ, then:

x̄ = (x₁ + x₂ + … + xₙ) / n

Consider five trade outcomes:

+2R

+1R

-1R

-1R

+4R

The total result is +5R, so the arithmetic mean is +1R per trade.

That number is useful, but it does not tell us how the result was produced. It does not show the dispersion of losses, whether a few extreme winners dominated the outcome, how many trades were profitable, how long losing streaks lasted, or how stable the process was.

A professional trader therefore does not stop at “What is the average?” The stronger question is: “What does the average mean when it is viewed together with the rest of the distribution?”

6. Why the Median Can Tell a Different Story

The median is the middle value of an ordered dataset. Consider these outcomes: -1R, -0.5R, 0R, +0.5R, +1R. Both the mean and the median are 0R.

Now change only the final observation: -1R, -0.5R, 0R, +0.5R, +10R. The arithmetic mean becomes +1.8R, while the median remains 0R.

The difference matters. The mean is strongly influenced by extreme observations, while the median is far less sensitive to them.

In a system where rare large winners contribute a substantial share of the total result, the gap between mean and median contains information about the distribution. A mean far above the median can indicate positive asymmetry.

That is not automatically good or bad. A trend-following system may intentionally rely on rare large winners. The question is not whether the distribution looks “clean.” The question is whether the observed distribution is consistent with the mechanism from which the strategy expects its edge.

7. Range and Extreme Outcomes

The simplest observed range is:

Range = Maximum − Minimum

If the smallest trade outcome is -3R and the largest is +7R, the observed range is 10R.

Range is easy to calculate, but it is highly sensitive to extreme observations. A single +20R trade can expand the range dramatically even if the rest of the dataset remains unchanged.

For that reason, range is useful as an initial description but is not enough to judge the stability or dispersion of a trading distribution. A trader must also ask whether an extreme outcome is a normal part of the strategy or a rare event that should be interpreted differently.

8. Distribution Is More Important Than a Single Statistic

Trading results are not simply “winners and losers.” They form a distribution of outcomes, and that distribution can contain much more information than one summary number.

Imagine two systems. System A produces many small wins, many small losses, and rare very large winners. System B produces frequent small wins and rare very large losses. Both systems may have the same average result.

The underlying distributions, however, are fundamentally different. Those differences can affect Drawdown, required Sample Size, psychological pressure, risk-management requirements, and the long-run path of results.

An Expert trader therefore does not inspect only one performance metric. The trader tries to understand the structure that generates the metric.

9. Outliers — When One Trade Distorts the Story

An outlier is an observation that differs materially from most of the dataset. In trading it may be an unusually large winner, an unusually large loss, a gap event, an execution error, or a genuinely rare market event.

An outlier should not be removed automatically. That is a critical rule.

If a +12R trade is a legitimate consequence of the strategy, deleting it would falsify the strategy’s observed behavior. If a +12R result is caused by a data-entry error, keeping it would also distort the analysis.

A disciplined process is:

Identify the unusual observation.

Verify that the data are correct.

Check whether the trade satisfied the strategy rules.

Determine whether the event belongs to the strategy’s normal mechanism or represents an external anomaly.

Only then decide how it should be treated in the analysis.

Data are not changed to make performance look better. They are examined so that the trader knows exactly what they represent.

10. Descriptive Statistics and Trading Performance

Descriptive statistics connects directly to performance analysis. A trading dataset can be summarized by the number of trades, total P&L, average result, median result, minimum and maximum outcomes, win rate, loss rate, average winner, average loser, the shape of the outcome distribution, and performance under different market conditions.

Together, these measures create a much richer picture than a single headline metric.

However, statistical description and strategy validation are not the same task. Descriptive statistics shows how the observed sample behaved. Whether that behavior is robust, generalizable, and likely to persist requires additional methods.

That is why later Expert lessons address probability distributions, statistical significance, sample size, backtesting, out-of-sample testing, walk-forward analysis, Monte Carlo simulation, and robustness testing.

11. Conditional Descriptive Statistics

One of the most important Expert-level steps is to stop treating the entire dataset as one homogeneous block.

Suppose a strategy shows an average result of +0.25R across 1,000 trades. Now segment the same sample:

Trending regime: +0.52R

Range regime: -0.11R

High-volatility regime: +0.61R

Low-volatility regime: +0.03R

The overall +0.25R now has a different meaning. It is more accurate to say: “The strategy averaged +0.25R in the full observed sample, while performance differed materially across market environments.”

Aggregation can hide structure. The Expert trader therefore moves from asking “How did the strategy perform?” toward asking “Under which conditions did it perform that way?”

12. Time Dependence and the Danger of Treating Every Trade as Independent

It is easy to treat every trade as if it were a completely independent observation. In markets, that assumption is often unrealistic.

Consecutive trades may occur within the same market regime. One macroeconomic event may affect several positions. A volatility shock may cause multiple setups to behave similarly.

This means that 100 recorded trades do not necessarily represent 100 fully independent pieces of information.

Descriptive statistics does not solve this problem by itself, but it should make the analyst aware of it. If a trader looks only at the number of data points, the effective information content of the sample can be overstated.

This becomes increasingly important in Sample Size, backtesting, and statistical validation.

13. From P&L to R-Multiple

The measurement scale matters. Currency P&L is not always suitable for comparing trades directly because position size and account size can change.

If one trade risks USD 100 and earns USD 200, the result is +2R. If another risks USD 500 and earns USD 1,000, the result is also +2R.

The money amounts differ, but the outcome relative to predefined risk is the same.

R-Multiple can therefore be useful as a normalized unit for statistical analysis. It helps separate strategy behavior from the absolute size of the capital deployed.

14. Why Descriptive Statistics Cannot Prove an Edge

Suppose a trader analyzes 2,000 historical trades and finds an average result of +0.30R. It is tempting to say, “I have proven that the strategy has an edge.” That conclusion is too strong.

Descriptive statistics shows how the observed sample behaved. The existence and persistence of a Strategy Edge require deeper questions:

How large was the Sample Size?

How were the observations selected?

Was there look-ahead bias?

Was there survivorship bias?

Was there data snooping?

Were the rules defined before or after the results were seen?

Which market environments are represented?

How did the strategy perform out of sample?

How stable was behavior across time?

How sensitive is the strategy to parameter changes?

The +0.30R figure is important evidence about the observed sample, but it does not prove every claim a trader may want to make about the future.

A professional principle follows: draw only the conclusion that the evidence is strong enough to support.

15. A Descriptive Statistics Workflow

A professional first-pass analysis of a trading dataset can follow a structured workflow:

Define the dataset — specify exactly which trades are included.

Validate the data — identify missing, incorrect, or duplicated observations.

Define the unit of measurement — for example currency, percentage return, or R-Multiple.

Calculate basic descriptive measures — mean, median, minimum, maximum, and other relevant statistics.

Examine the distribution — do not stop at one number.

Identify outliers — investigate extreme observations.

Segment the data — for example by setup, instrument, timeframe, or market regime.

Compare conditional behavior — determine whether performance depends on identifiable conditions.

Document the interpretation — state what the data show and what they do not yet establish.

The final step is essential. A good statistical review contains not only numbers, but also the limits of the claims that can responsibly be made from those numbers.

16. Case Study — The Strategy That Looked Better Than It Was

A trader evaluates a breakout strategy using 600 historical trades.

The headline results are:

Total P&L: +180R

Average trade: +0.30R

The first reaction may be: “This is an excellent strategy.” A deeper descriptive review shows that 25 of the 600 trades generated +150R, while the remaining 575 trades generated only +30R in total.

This does not automatically invalidate the strategy. A breakout methodology may naturally depend on rare large winners. But the structure changes the questions that must be asked:

Under what conditions did the largest winners occur?

Were they concentrated in the same market regime?

Across how many years were they distributed?

Were the trading rules defined before those outcomes were observed?

Does the system remain positive without the largest observations?

What Drawdown should be expected when large winners arrive less frequently?

How asymmetric is the outcome distribution?

The deeper question is no longer simply “Is the average +0.30R?” It is “What distribution produced the +0.30R average?”

That is the difference between recording performance and analyzing performance professionally.

17. Practical Application — Build a Descriptive Statistics Profile

Select one clearly defined strategy and collect a consistent sample of historical trades. Record every trade using the same measurement convention.

At minimum, include the following fields:

Trade ID

Date

Instrument

Setup

R-Multiple

Holding time

Market regime

Outcome

Then build the strategy’s first descriptive profile. Calculate:

Number of trades

Total result in R

Mean trade result

Median trade result

Minimum

Maximum

Number of winning trades

Number of losing trades

Largest winner

Largest loss

Sort the observations by R-Multiple and examine the structure:

How different are the mean and median?

How wide is the observed range?

Are there a few extreme outcomes?

What share of total performance comes from the largest winners?

Do certain market regimes show materially different behavior?

Finally create two separate lists: “What the data support” and “What the data do not prove.” The second list is mandatory. The goal is not to find a flattering performance statistic. The goal is to distinguish observation from interpretation.

18. The Expert-Level Statistical Mindset

Descriptive statistics may appear simple because many of its individual calculations are elementary. At Expert level, however, the real skill is not memorizing formulas.

The real skill is learning to separate observation from interpretation, sample from population, aggregation from conditional behavior, legitimate extremes from bad data, and evidence from conclusions that go beyond the evidence.

A professional trader is not quantitative merely because every decision contains complicated mathematics. Quantitative thinking means refusing to let one spectacular trade, one recent streak, or one intuitive explanation replace the discipline of evidence.

Numbers do not think for the trader. Used properly, they discipline the trader’s thinking.

Practical Application — Statistical Reality Check

Choose a strategy with a historical trade sample and create a complete descriptive statistics profile.

Then write three separate statements:

One statement strongly supported by the observed sample.

One statement that is plausible but requires additional evidence.

One statement that cannot be established from the available data.

Next, temporarily remove the three largest winning trades and recalculate the mean result. Examine how much it changes.

Do not conclude that the strategy is “good” or “bad.” Instead answer:

How dependent is performance on extreme winners?

Are those extreme winners a legitimate part of the strategy mechanism?

If they are, what does that imply about the expected distribution of outcomes?

What additional evidence would be required to judge robustness?

Finally, summarize the strategy’s observed performance in one sentence without making a claim stronger than the sample can support.

Knowledge Check

What is the difference between descriptive statistics and statistical inference?

What is the difference between a population and a sample?

What are observations and variables in a trading dataset?

Why is the mean alone insufficient to describe a strategy?

Why can the mean and median differ materially?

Why should the full distribution of outcomes be examined?

Why should outliers not be removed automatically?

Why can segmenting results by market regime be important?

Why can aggregate performance hide useful information?

Why can R-Multiple be useful for normalizing trading results?

Why does a positive average trade not by itself prove a Strategy Edge?

Why can the number of recorded trades overstate the amount of independent information in a sample?

What is the difference between “the data show this” and “therefore the strategy works”?

What should be checked before deciding how to treat an outlier?

Why should the limits of a statistical conclusion be documented?

Key Principle

Descriptive statistics is not meant to reduce a strategy to one attractive number. Its purpose is to show how the observed data behaved, what structure produced the result, and where the evidence stops supporting stronger conclusions.

Copying, reproducing, adapting, distributing, or publicly displaying any part or all of the Academy’s textual content without the author’s prior written consent is prohibited.