Sie sind auf Seite 1von 22

Monte Carlo Simulation

Probability Fundamentals
• Event
A desired outcome

• Probability
The chance of getting a desired outcome
Probability Example
Joe is tossing a die. What is the probability of getting the
two dots on the top side, once the die lands on the table?

Event
Getting the two dots on the top side

Probability
1
6
Discrete Probability Distribution
An analyst is predicting the stock price of GoProLogistics. He estimates
that the probability of reaching $92 per share is 10%, the probability of
reaching $93 per share is 30%, the probability of reaching $94 per
share is 20%, and the probability of reaching $95 per share is 40%.

Probability
45%

40%

35%

30%

25%
Probability
20%

15%

10%

5%

0%
$92 $93 $94 $95
Probability for Continuous Data
An analyst is predicting the stock price of GoProLogistics.
He estimates the stock can reach any price between $80
and $100 per share. For example, the price can be $80 per
share, or $92.18 per share, or $99.56 per share, etc.

The probability of reaching any value between $80 and


$100 per share is equal.

What is the probability of reaching $93.75 per share?


Uniform Probability Distribution
Normal Distribution
Monte Carlo Simulation Example
The ABC company is developing a new smartphone named
KoolCal. The company wants to predict the first year profits
of this new phone by considering the following factors:

• Selling price per unit (p). The price for the new phone is
$249 each.
• First-year administrative and advertising costs (ca).
Currently, this cost is $1,000,000.
• direct labor cost per unit (ci).
• parts cost per unit (cp).
• first-year demand (d).
Current Analyses
• Base-Case Scenario

• Best-Case Scenario

• Worst-Case Scenario
Base-Case Scenario
• The base-case estimates of the direct labor cost per unit,
the parts cost per unit, and first-year demand are $45,
$90, and 15,000 units, respectively. Using these
estimations, the profit projection is:

• Profit = (249 - 45 - 90)*(15,000) - 1,000,000 = 710,000

• Thus, the base-case scenario leads to an anticipated


profit of $710,000.
Best-Case Scenario
• The best-case estimates of the direct labor cost per unit,
the parts cost per unit, and first-year demand are $43,
$80, and 30,000 units, respectively. Using these
estimations, the profit projection is:

• Profit = (249 - 43 - 80)*(30,000) - 1,000,000 = 2,780,000

• Thus, the best-case scenario leads to an anticipated profit


of $2,780,000.
Worst-Case Scenario
• The worst-case estimates of the direct labor cost per unit,
the parts cost per unit, and first-year demand are $47,
$100, and 0 units, respectively. Using these estimations,
the profit projection is:

• Profit = (249 - 47 - 100)*(0) - 1,000,000 = -1,000,000

• Thus, the worst-case scenario leads to an anticipated loss


of $1,000,000.
Weaknesses of the Current Analyses
• The analyst only estimated the possible events (e.g.,
direct labor cost, part cost, and demands in the base
scenario, best scenario, and worst scenario), but she
didn’t consider the probability of the events.

• The ignorance of probability makes the current analyses


very subjective, and the predictions could have missed
other possibilities.
Monte Carlo Simulation
• Before the data analysis, we find out each unknown factor’s
probability distribution.
• Expand the possible scenarios from 3 to a large number, e.g.,
1000 scenarios or 2000 scenarios. The larger, the better.
• In each scenario, for each unknown factor, we let the analysis
software generate a random number (denoted as Δ) that is
between 0 and 1.
• We bring each random number Δ to each unknown factor’s
probability distribution, and find the value corresponding to the
random number in the probability distribution. The set of values
for the unknown factors is called a trial.
• After one trial, we repeat this process above to get other trials
in the rest of the scenarios. We consider all trials, so that we
can show the probability distribution of the profit.
Direct Labor Cost Per Unit

Direct Probability
Labor
Cost
$43 0.1
$44 0.2
$45 0.4
$46 0.2
$47 0.1
Part Cost Per Unit
Demand
1000 Scenarios
• In the first scenario, we let the analysis software generate
a random number Δ for direct labor cost per unit, part cost
per unit, and demand.
• Corresponding to each random number Δ, we can find the
direct labor cost per unit, part cost per unit, and demand
that are most likely to occur.
• Use the predicted direct labor cost per unit, part cost per
unit, and demand to calculate the profit in the first
scenario.
• Repeat the steps above to get the other 999 scenarios.
• Predict the probability distribution of profit.
Example
In the first trial, the analysis software generate the following random
numbers for the three factors.
Factor Random Number
Direct Labor Cost per Unit 0.4
Part Cost per Unit 0.3
Demand 0.6
Corresponding to these random numbers, the following events are
most likely to occur.
Factor Predicted Value
Direct Labor Cost per Unit $45 (Because $45 is most likely to occur at
the probability of 0.4 according to the direct
lab cost’s probability distribution)
Part Cost per Unit $86 ($80+0.3*(100-80)=$86)
Demand 16141 (Use =NORM.INV(0.6,15000,4500)
function)
Thus, the estimated profit in the first trial is
Profit = (249 - 45 - 86) * 16,141 – 1,000,000 = 904,638
Random Value in Uniform Probability

If the software gives us a random number 0.3, what is the


corresponding value?
Corresponding value = lower bound + (upper bound –
lower bound) * the given probability = 80 + (100-80)*0.3 =
86
Repeat the same steps in other 999
scenarios, and get different estimated
profits
Advantages of Monte Carlo Simulation
• Simulate realities that have different scenarios
with different probabilities.

• Simulate realities in which real experiments are


impossible (e.g., nuclear weapon development).

• Address the analysts’ narrow subjective opinions.

Das könnte Ihnen auch gefallen