Theoretical Probability: The expected probability based on mathematical analysis of the situation. Experimental Probability: The observed probability based on actual trials.
Theoretical: P(A) = Number of favorable outcomes / Total possible outcomes
Experimental: P(A) = Number of times A occurs / Total number of trials
For a fair coin: P(heads) = 1/2 = 0.5 or 50%
Based on 100 tosses: P(heads) = 58/100 = 0.58 or 58%
Theoretical: 0.50, Experimental: 0.58, Difference: 0.08 or 8%
As number of trials increases, experimental probability approaches theoretical probability
P(Heads) = 1/2 = 0.50
Ideal fair coin
No actual trials
P(Heads) = 58/100 = 0.58
Actual 100 tosses
Based on data
Experimental: 0.58 (58%)
Difference: 0.08 (8%)
Theoretical probability of heads = 0.50 (50%). Experimental probability = 0.58 (58%). The experimental probability differs slightly due to random variation. As the number of tosses increases, experimental probability converges to theoretical probability according to the Law of Large Numbers.
As the number of trials increases, the experimental probability approaches the theoretical probability.
For a fair die: P(rolling a 4) = 1/6 β 0.167 or 16.7%
Based on 300 rolls: P(rolling a 4) = 55/300 = 0.183 or 18.3%
Difference = 0.183 - 0.167 = 0.016 or 1.6%
With 300 trials, the experimental probability is quite close to theoretical
Experimental: 55/300 = 0.183 (18.3%)
Difference: 0.016 (1.6%) - Reasonably accurate
Theoretical probability of rolling a 4 = 1/6 β 0.167 (16.7%). Experimental probability = 55/300 = 0.183 (18.3%). The experimental probability is quite accurate with only a 1.6% difference, demonstrating good convergence with 300 trials.
β’ Sample size: Larger samples provide more accurate estimates
β’ Randomness: Fair die ensures theoretical probability is valid
β’ Convergence: Experimental probability approaches theoretical as trials increase
Convergence: The tendency of experimental probability to approach theoretical probability as the number of trials increases.
For equal sections: P(red) = 1/4 = 0.25 or 25%
After 10 spins: 4/10 = 0.40 (40%), After 100 spins: 28/100 = 0.28 (28%), After 1000 spins: 251/1000 = 0.251 (25.1%)
As sample size increases, experimental probability gets closer to theoretical
Pattern shows experimental probability approaching 0.25 as trials increase
100 spins: 0.28 (28%)
1000 spins: 0.251 (25.1%)
Pattern: Convergence toward 0.25
Theoretical probability = 0.25 (25%). Experimental probabilities: 0.40 (10 spins), 0.28 (100 spins), 0.251 (1000 spins). The pattern shows convergence to theoretical probability as sample size increases, demonstrating the Law of Large Numbers.
β’ Small samples: Greater variability from theoretical
β’ Larger samples: Closer approximation to theoretical
β’ Asymptotic: Approaches but never perfectly equals theoretical
Formula: P(A) = Favorable outcomes / Total possible outcomes
Characteristics: Predictive, consistent, based on model
Formula: P(A) = Times A occurred / Total trials
Characteristics: Observable, variable, based on data
Formula: lim(nββ) [Experimental P(A)] = Theoretical P(A)
Implication: More trials β More accurate estimates
Fairness Testing: Using experimental probability to assess whether a device behaves as expected theoretically, considering reasonable variation.
P(heads) = 532/1000 = 0.532 or 53.2%
Theoretical for fair coin: 0.50 or 50%, Difference: 0.032 or 3.2%
3.2% difference is relatively small for 1000 trials, suggesting coin may be fair
With more trials, probability should converge closer to 0.50 if coin is fair
P(Heads) = 0.50
Long-run average
Mathematical model
P(Heads) = 0.532
Actual 1000 flips
Reasonable variation
Theoretical: 0.50 (50%)
Difference: 0.032 (3.2%)
Conclusion: Likely fair with variation
The experimental probability is 0.532, which is reasonably close to the theoretical 0.50. The 3.2% difference is within expected variation for 1000 trials. The coin is likely fair. With more flips, the experimental probability should converge closer to 0.50.
β’ Small differences: Acceptable variation even with fair devices
β’ Large differences: May indicate bias or unfairness
β’ Sample size matters: Larger samples provide better fairness assessment
Probability Estimation: Using experimental probability to estimate unknown theoretical probabilities when the underlying distribution is not known.
P(red) = 125/200 = 0.625 or 62.5%
About 62.5% red marbles, 37.5% blue marbles
With 200 trials, confidence is moderate; more trials would increase confidence
Future draws should yield approximately 62.5% red marbles
Bag Composition: ~62.5% red, ~37.5% blue
Confidence: Moderate with 200 trials
Prediction: 62.5% red in future draws
The estimated probability of drawing a red marble is 0.625 (62.5%). The bag likely contains approximately 62.5% red marbles and 37.5% blue marbles. With 200 trials, confidence is moderate; more trials would provide a more precise estimate.
β’ Sample size: Larger samples provide more reliable estimates
β’ Confidence: Related to sample size and variability
β’ Accuracy: Experimental probability approximates true probability
β’ Prediction: Future trials should reflect estimated probability
π Definitions of Key Concepts
βοΈ Core Rules and Principles
βοΈ Step-by-Step Methods for Analysis
- Identify the situation: Determine what is being tested/probabilized
- Calculate theoretical probability: Use mathematical model to find expected probability
- Conduct experiment: Perform trials and record results
- Calculate experimental probability: Use observed data to find actual probability
- Compare probabilities: Analyze difference between theoretical and experimental
- Assess sample size: Consider if more trials are needed
- Draw conclusions: Interpret results in context of Law of Large Numbers
π§ͺ Examples: From Simple to Advanced
π‘ Tips, Tricks, and Common Pitfalls
π Key Notes for Memorization
π Convergence Analysis Table
| Sample Size | Expected Behavior | Variation Level | Reliability |
|---|---|---|---|
| Small (β€50) | May deviate significantly | High | Low |
| Medium (50-500) | Reasonable approximation | Moderate | Moderate |
| Large (>500) | Closer to theoretical | Low | High |
| Very Large | Very close to theoretical | Very Low | Very High |
Questions & Answers
Question: Why don't experimental probabilities always match theoretical probabilities, even with large sample sizes?
Answer: There are several reasons for this:
- Random variation: Even fair coins can have slight deviations due to chance
- Physical imperfections: Real coins, dice, or spinners may have tiny biases
- Convergence is asymptotic: Experimental probability approaches theoretical but never perfectly equals it
- Sample size: "Large" sample sizes are still finite
Analogy: A fair coin tossed 1 million times might show 50.01% heads instead of exactly 50%.
Reality: Perfect convergence happens only at infinity, which is impossible to achieve.
Question: How many trials do I need to be confident that my experimental probability is accurate?
Answer: It depends on your desired accuracy and confidence level:
- General rule: At least 30 trials for basic reliability
- Reasonable accuracy: 100-500 trials
- High precision: 1000+ trials
- Diminishing returns: Each additional trial adds less precision
Guideline: For most classroom experiments, 100-200 trials provide good balance of accuracy and feasibility.
Statistical note: Precision improves with the square root of sample size, so quadrupling trials doubles precision.
Question: Can experimental probability ever be more accurate than theoretical probability?
Answer: In a specific sense, yes, when the theoretical model is incorrect:
- Biased device: If a coin is actually biased (60% heads), experimental probability with large sample reveals true probability better than theoretical assumption of 50%
- Model mismatch: When theoretical assumptions don't match reality
- Discovery: Experimental probability can reveal that theoretical model needs adjustment
Example: A coin that lands heads 55% of the time. After 10,000 flips, experimental probability (β0.55) is more accurate than theoretical (0.50).
Reality: Both have value - theoretical provides expectations, experimental validates or refutes them.