Evaluating Probability Models — I can justify whether a...

SECTION 1

Historical Context & Motivation

People have been gambling and making predictions for thousands of years, but the formal study of probability only began a few centuries ago. Early mathematicians noticed a critical problem: sometimes people assumed all outcomes were equally likely when they clearly were not. This led to wildly inaccurate predictions and, in gambling, significant financial losses. The question of which outcomes are equally likely became one of the foundational challenges of probability theory.

1564

Cardano's Book on Games of Chance

Italian mathematician Gerolamo Cardano wrote one of the first systematic analyses of dice games, recognizing that each face of a fair die has an equal chance of appearing.

1654

Pascal & Fermat Correspondence

Blaise Pascal and Pierre de Fermat exchanged letters about the "problem of points," establishing the foundations of probability theory by carefully analyzing which outcomes were equally likely in games of chance.

1814

Laplace's Classical Definition

Pierre-Simon Laplace formalized the classical probability model: the probability of an event equals the number of favorable outcomes divided by the total number of equally likely outcomes.

1933

Kolmogorov's Axioms

Andrey Kolmogorov published a rigorous mathematical framework for probability that works for both equally likely and non-equally likely models, unifying the field.

The central question this lesson addresses is: How do we decide whether a probability model that treats outcomes as equally likely is actually reasonable for a given situation? If we assume equal likelihood when it doesn't exist — or ignore it when it does — our predictions and decisions will be flawed.

SECTION 2

Core Principles & Definitions

Before evaluating any probability model, you need to understand the key vocabulary. A probability model is a mathematical description of a random process that lists every possible outcome and assigns a probability to each one. The set of all possible outcomes is called the sample space. Every valid probability model must satisfy two rules: each probability is between 0 and 1, and all the probabilities in the sample space add up to exactly 1.

Probability Model

A complete list of all possible outcomes for a random process, along with the probability assigned to each outcome. The probabilities must be between 0 and 1 and must sum to 1.

Equally Likely Model

A probability model in which every outcome in the sample space has the same probability. For n outcomes, each probability is 1/n. Examples: fair coins, fair dice, well-shuffled cards.

Non-Equally Likely Model

A probability model in which some outcomes are more probable than others. The probabilities are different values but still sum to 1. Examples: loaded dice, weather forecasts, free-throw percentages.

Justification

The reasoning you give for why a particular model is appropriate. You must consider the physical setup, symmetry, and any available data to determine if equal likelihood is reasonable.

✦ KEY TAKEAWAY

Think of a probability model like a weather forecast app. A good app doesn't just say "it might rain or it might not" and call it 50/50. It uses real data — humidity, wind patterns, temperature — to assign a specific chance to each outcome. Similarly, a good probability model assigns probabilities based on the actual structure of the situation, not just a guess. If the situation has physical symmetry (like a fair die), equal probabilities make sense. If it doesn't (like predicting whether a basketball player makes a shot), you need a non-equally-likely model.

SECTION 3

Visual Explanation: Equally Likely vs. Non-Equally Likely

On the left, a fair die produces six outcomes each with probability 1/6 — all bars are the same height. On the right, a spinner with unequal sections assigns different probabilities to each color. Both are valid probability models because each sums to 1, but only the left one uses an equally likely model.

The diagram above shows the essential visual difference between the two types of probability models. In the equally likely model on the left, every bar reaches the same height, meaning each outcome has the same chance of occurring. In the non-equally likely model on the right, the bars have different heights because some outcomes are more probable than others. The key point is that both models are valid — both have probabilities between 0 and 1, and both sum to 1. The question is which model fits the real-world situation you're analyzing.

SECTION 4

Mathematical Framework

Every probability model must satisfy two fundamental rules. These rules are non-negotiable — if a model violates either one, it is not a valid probability model regardless of whether outcomes are equally likely.

RULE 1 — VALID RANGE

0 ≤ P(outcome) ≤ 1 for every outcome

P(outcome) represents the probability assigned to any individual outcome. No probability can be negative, and none can exceed 1.

RULE 2 — SUM EQUALS ONE

P(o₁) + P(o₂) + P(o₃) + … + P(oₙ) = 1

The sum of the probabilities of all outcomes o₁ through oₙ in the sample space must equal exactly 1. This guarantees that something in the sample space will happen.

EQUALLY LIKELY MODEL

P(each outcome) = 1/n, where n = number of outcomes

When all outcomes are equally likely, you simply divide 1 by the total number of outcomes n. For a fair coin, n = 2, so P(heads) = P(tails) = 1/2. For a fair die, n = 6, so each face has probability 1/6.

When you encounter a situation and want to build a probability model, you should ask yourself: Is there a reason to believe every outcome has the same chance? Look for physical symmetry — a fair die is a symmetric cube, a fair coin is balanced, a well-shuffled deck treats every card the same. If symmetry exists, the equally likely model is justified. If there is no symmetry, or if data suggests some outcomes happen more often, you need a non-equally likely model where you assign specific probabilities based on evidence.

SECTION 5

Classifying Situations: Equally Likely or Not?

The most important skill in evaluating probability models is the ability to look at a real-world situation and determine whether the outcomes are equally likely. Below is a decision-making framework you can use, followed by a visual flowchart.

This flowchart guides you through evaluating a probability model. Start by listing all outcomes, then check for physical symmetry. If symmetry exists, an equally likely model is justified. If not, look for data (like past frequencies) to assign specific probabilities.

Common probability scenarios and their classification

Situation	Equally Likely?	Justification
Rolling a standard fair die	Yes	The die is a symmetric cube; each face has the same size, shape, and weight distribution.
Drawing a card from a shuffled deck	Yes	Shuffling randomizes the order; each of the 52 cards is equally accessible.
Predicting if a student passes an exam	No	Passing and failing are not symmetric; they depend on preparation, difficulty, and prior knowledge.
Selecting a random day and checking if it rains	No	Rain depends on geography, season, and climate. In many locations, it rains on far fewer than 50% of days.
Spinning a spinner with four equal sections	Yes	The sections are the same size, so the spinner arrow is equally likely to land on each one.
Choosing the winner of a basketball game	No	Teams have different skill levels, records, and matchup advantages. The outcomes are not symmetric.

SECTION 6

Worked Example

Let's walk through a complete example where you must evaluate two different probability models for the same situation and determine which one is reasonable.

🎯 THE SCENARIO

A bag contains 10 marbles: 5 red, 3 blue, and 2 green. You draw one marble at random. Two students propose different probability models. Model A says P(red) = P(blue) = P(green) = 1/3. Model B says P(red) = 5/10, P(blue) = 3/10, P(green) = 2/10. Which model is reasonable?

Evaluating the Two Models

Step 1 — Check Validity of Model AModel A assigns P(red) = P(blue) = P(green) = 1/3. First, check the two rules. Each probability is between 0 and 1 — that's satisfied. The sum is 1/3 + 1/3 + 1/3 = 1 — also satisfied. So Model A is a valid probability model. But is it reasonable?

Model A is valid but needs justification check.

Step 2 — Check Validity of Model BModel B assigns P(red) = 5/10 = 0.5, P(blue) = 3/10 = 0.3, P(green) = 2/10 = 0.2. Each probability is between 0 and 1. The sum is 0.5 + 0.3 + 0.2 = 1.0. Model B is also a valid probability model.

Model B is valid. Now we compare reasonableness.

Step 3 — Evaluate SymmetryAre the three color outcomes equally likely? The bag contains 5 red, 3 blue, and 2 green marbles — these are unequal quantities. There is no physical symmetry among the color categories because there are more red marbles than blue, and more blue than green. Each individual marble is equally likely to be drawn, but the color groups are not equally likely.

No symmetry among color groups → equally likely model is NOT justified.

Step 4 — Match Model to SituationSince each of the 10 individual marbles is equally likely to be drawn (random selection from a well-mixed bag), the probability of drawing a specific color equals the number of marbles of that color divided by 10. This matches Model B exactly: P(red) = 5/10, P(blue) = 3/10, P(green) = 2/10.

Model B is reasonable. Model A is not, because the colors are not equally likely.

Step 5 — Write the JustificationA complete justification should state: "Model A is not reasonable because it assumes all three colors are equally likely, but the bag contains unequal numbers of each color. Model B is reasonable because each marble is equally likely to be drawn, and the probability of each color is the fraction of marbles that are that color: P(red) = 5/10, P(blue) = 3/10, P(green) = 2/10."

Always explain WHY, not just which model you chose.

SECTION 7

Strengths & Limitations of Each Model Type

Neither type of probability model is inherently better — each has strengths and weaknesses depending on the situation. Understanding these trade-offs helps you choose the right tool for the job.

Comparison of equally likely and non-equally likely probability models

Feature	Equally Likely Model	Non-Equally Likely Model
Simplicity	Very simple — just divide 1 by the number of outcomes. No data collection needed.	More complex — requires data, research, or analysis to assign each probability.
Accuracy	Only accurate when physical symmetry truly exists. Can be very wrong otherwise.	Can be highly accurate when based on good data, but only as good as the data it uses.
When to use	Fair coins, fair dice, well-shuffled cards, random number generators.	Sports outcomes, weather, medical test results, surveys, any situation without symmetry.
Common mistake	Assuming equal likelihood just because there are two outcomes (e.g., "it rains or it doesn't" ≠ 50/50).	Using outdated or biased data, or applying data from one situation to a very different one.

⚠️ WATCH OUT FOR THIS TRAP

The most common error students make is assuming that because there are only two outcomes, each must have a 50% chance. Think of it this way: when you walk outside, either a meteor hits you or it doesn't — but that doesn't mean there's a 50% chance of getting hit by a meteor! The number of outcomes does NOT determine whether they are equally likely. You must look at the physical situation and the evidence.

SECTION 8

Connection to Advanced Probability

The skill of evaluating probability models is foundational for everything that comes next in statistics and probability. As you move into more advanced topics, the same question — "Is this model reasonable?" — keeps coming back in increasingly sophisticated forms.

How evaluating probability models connects to advanced topics

This Lesson	Where It Leads
Deciding if outcomes are equally likely	Hypothesis testing — using data to decide if a proposed model fits the evidence (chi-square tests)
Using counts to determine probabilities	Relative frequency and the Law of Large Numbers — running many trials to estimate probabilities experimentally
Assigning probabilities that sum to 1	Probability distributions — continuous and discrete distributions like the normal distribution and binomial distribution
Justifying model choice with reasoning	Statistical modeling — building regression models and evaluating their assumptions

In future courses, you'll encounter situations where you must not only choose a probability model but also test whether data supports your model using formal statistical procedures. For now, the key insight is that every statistical test begins with someone proposing a model and asking, "Does this model actually fit the data?" — which is exactly what you're learning to do in this lesson.

SECTION 9

Practice Problems

PROBLEM 1 — CONCEPTUAL

A student says, "There are two outcomes when you flip a thumbtack — it lands point-up or point-down — so each has a probability of 1/2." Is this reasoning correct? Explain why or why not.

PROBLEM 2 — BASIC CALCULATION

A spinner is divided into 5 equal sections colored red, blue, green, yellow, and orange. (a) Is an equally likely model reasonable? Justify your answer. (b) What is P(blue)?

PROBLEM 3 — INTERMEDIATE

A bag contains 8 marbles: 4 white, 3 black, and 1 red. Marcus proposes the model P(white) = 0.50, P(black) = 0.30, P(red) = 0.20. Elena proposes the model P(white) = 0.50, P(black) = 0.375, P(red) = 0.125. (a) Are both models valid? (b) Which model is more reasonable? Justify your answer.

PROBLEM 4 — APPLIED

A school collects data on how students get to school: 55% drive, 25% take the bus, 15% walk, and 5% bike. A new student proposes an equally likely model with P(drive) = P(bus) = P(walk) = P(bike) = 0.25. (a) Is the equally likely model valid? (b) Is it reasonable? (c) Construct a more reasonable model and justify it.

PROBLEM 5 — CRITICAL THINKING

Consider rolling two standard fair dice and recording the sum. The possible sums range from 2 to 12, giving 11 possible outcomes. A student argues that since each die is fair, the equally likely model applies, so P(sum = 7) = 1/11. Identify the flaw in this reasoning, explain what the correct probability of rolling a sum of 7 is, and describe what kind of model should be used for dice sums.

SUMMARY

Lesson Summary

A probability model assigns probabilities to every outcome in the sample space, and every valid model must have probabilities between 0 and 1 that sum to exactly 1. An equally likely model assigns P = 1/n to each of n outcomes and is justified only when physical symmetry makes every outcome genuinely the same — like fair dice, fair coins, or well-shuffled cards.

When symmetry is absent, you need a non-equally likely model where probabilities are determined by data or relative frequencies. The critical skill is justification: always explain why your chosen model fits the situation. Remember that having only two outcomes does not make them equally likely — the number of outcomes alone never tells you whether equal probability is appropriate. Look at the structure of the situation, check for symmetry, and use data when symmetry doesn't exist.