Drunk guy by a cliff
There's a drunk guy standing next to a cliff which is one step to his left. He steps to the left with probability $1-p$ and to the right with probability $p$. What is the probability that he dies?
Let $k$ be the position of the guy, where death is at $k=0$. Let $D_k$ be the event that he dies given that he is at position $k$. Therefore $P(D_0) = 1$ and
$$P(D_1) = P(D_1 \mid \text{Step left}) P(\text{Step left}) + P(D_1 \mid \text{Step right}) P(\text{Step right})$$and we note that $P(D_1 \mid \text{Step left}) = P(D_0) = 1$, so
$$P(D_1) = (1-p)P(D_0) + pP(D_2) = 1-p + pP(D_2) $$At this point we cleverly realize that $P(D_2) = P(D_1)^2$ and letting $x = P(D_1)$ we have
$$x = (1-p) + px^2$$ $$0 = px^2-x + 1 - p$$and using the quadratic formula,
$$x = \frac{1 \pm \sqrt{1-4p(1-p)}}{2p} = \frac{1 \pm \sqrt{1-4p+4p^2}}{2p} = \frac{1 \pm \sqrt{(1-2p)^2}}{2p} = \frac{1 \pm (1-2p)}{2p}$$ $$x = \frac{2-2p}{2p} = \frac{1-p}{p} \quad \text{or} \quad x = \frac{2p}{2p} = 1$$Since $x = P(D_1) \in [0,1]$ and the two solutions agree when $p=1/2$, the probability of death given you are at location $1$ is
If $p = P(\text{Stepping right to safety}) \le \frac{1}{2}$ then $P(D_1) = 1$
If $p = P(\text{Stepping right to safety}) > \frac{1}{2}$ then $P(D_1) = \frac{1-p}{p}$
For example if the probability of stepping to safety is $p=4/5$ which is not that "bad" then the solutions for $P(D_1)$ will equal $0.25$ or $1$ whereas if $p=1/5$ then the solutions for $P(D_1)$ will equal $1$ or $4$.
Articles
Personal notes I've written over the years.
- When does the Binomial become approximately Normal
- Gambler's ruin problem
- The t-distribution becomes Normal as n increases
- Marcus Aurelius on death
- Proof of the Central Limit Theorem
- Proof of the Strong Law of Large Numbers
- Deriving Multiple Linear Regression
- Safety stock formula derivation
- Derivation of the Normal Distribution
- Comparing means of Normal populations
- Concentrate like a Roman
- How to read a Regression summary in R
- Notes on Expected Value
- How to read an ANOVA summary in R
- The time I lost faith in Expected Value
- Notes on Weighted Linear Regression
- How information can update Conditional Probability
- Coupon collecting singeltons with equal probability
- Coupon collecting with n pulls and different probabilities
- Coupon collecting with different probabilities
- Coupon collecting with equal probability
- Adding Independent Normals Is Normal
- The value of fame during and after life
- Notes on the Beta Distribution
- Notes on the Gamma distribution
- Notes on Conditioning
- Notes on Independence
- A part of society
- Conditional Expectation and Prediction
- Notes on Covariance
- Deriving Simple Linear Regression
- Nature of the body
- Set Theory Basics
- Polynomial Regression
- The Negative Hyper Geometric RV
- Notes on the MVN
- Deriving the Cauchy density function
- Exponential and Geometric relationship
- Joint Distribution of Functions of RVs
- Order Statistics
- The Sample Mean and Sample Variance
- Probability that one RV is greater than another
- St Petersburg Paradox
- Drunk guy by a cliff
- The things that happen to us