SCNC1111

Scientific Method and Reasoning

Lecture 4
Guesstimation
(How to Make Educated Guesses)

Dr. William M.Y. Cheung

1
 Guesstimation
• Guesstimation = Guess + estimation
– Estimate the order of magnitude of a quantity
• Need not be exact
– Without using detailed/complicated information
• E.g. no reference to the internet or official statistics
– Quick
• No hours and hours of calculations

• E.g.:
– How many people in the world are using the bathroom at
this moment?

2
 Why Guesstimate?
Human-Powered Helicopter

Drawing of helicopter by Leonardo da Vinci

source: leonardo-da-vinci-biography.com
3
 Human-powered helicopter
• The Igor I. Sikorsky Human-Powered
Helicopter Competition from the American
Helicopter Society:
• A US$ 250,000 award
• A helicopter powered by a human
• must lift off the ground, hover for at least 60
sec, reach an altitude of 3 m during the flight,
and stay within a 10 m2 area
• Question: Will you take part in this
competition?
4
 Human-powered helicopter
• On June 13th, 2013, the
Canadian AeroVelo team
successfully flew their
human-powered
helicopter that meet the
requirements, and
hence officially claimed
the Igor I. Sikorsky
Human-Powered
Helicopter Prize.
http://www.wired.com/autopia/2013/07/human-powered-helicopter-prize/

5
 Why Guesstimate?
1. As a feasibility check
– Imagine you work in a laboratory…
Will it We can
work? perform an
experiment to
How large is measure this
this effect? effect!
How big should our
experimental set-up be in
order to be able to detect
this effect?
How long will the May not have the
experiment take so as exact answers unless
to obtain trustworthy the experiment has
results? been done…

– We need to guesstimate!
6
 Another Example

Can you charge your

smartphone by typing?

http://www.wired.com/wiredscience/2011/07/can-you-charge-your-phone-by-typing/

7
 Why Guesstimate?
2. To identify factors relevant to the problem
– Not all factors are important…
No! But it
should not be
important…

Have you considered the

Riemann quantum
mutational spectroscopic
effect in your experiment?

– Great if you can show that

• The effect you want to measure: ~ 0.1 mm
• This extra effect: ~ 0.0001 mm
– No need to be exact in this case!
8
 Why Guesstimate?
3. Necessary when quick but rough answers are
needed without many data available
– Many top software firms, management consultants, and
investment banks use similar guesstimation questions in
job interviews to judge the intelligence and problem-
solving skills of their applicants.

• Examples:
– What is the size of the market for disposable
diapers in China?
– How many golf balls does it take to fill a 747 plane?

source: amazon.com
9
 A Classic Example
Question: How many piano
tuners are there in Chicago (or in
Hong Kong, or in the United
States, or in the whole world)?

Can obtain a good estimate by

1. Starting with some basic
information;
2. Making reasonable
source: wikipedia
assumptions;
3. Breaking down the problem
into manageable parts.
10
 Fermi Questions
• Such a question is called a Fermi
question, named after Enrico Fermi
(1901—1954), an Italian physicist
and Nobel laureate.

• Fermi estimated the power of an

atomic bomb by Enrico Fermi
source: wikipedia
1. Dropping small pieces of paper
during the testing of the bomb;
2. Observing how far the paper
pieces had been blown away by
the blast wave.
He was correct within a factor of 2!
11
 Solution to the piano tuner problem
The following is a solution to the “piano tuners in Chicago”
problem (taken from Ecological Numeracy by Robert A.
Herendeen):

Break down the problem into smaller parts, with relevant

assumptions that we are more confident with:

How many households?

• Chicago city population: 3,000,000 (2007 figure is 2,836,658).
• On average, each household has 3 persons (2.6 for US, from
US census 2006).
• Number of households in Chicago: 3,000,000 ÷ 3 =
1,000,000.

12
 Solution to the piano tuner problem
How many tunings per year?
• Number of regularly tuned
pianos in households: 1 in 15
households.
• Average number of tunings for a
regularly tuned piano: once per
year.
• Number of pianos in households source: wikipedia
that are tuned regularly:
1,000,000 ÷ 15 ≈ 70,000,
resulting in 70,000 tunings per
year.

13
 Solution to the piano tuner problem
How many pianos can one tuner tune per year?
• Length of time (including travel time) to tune a piano: 2 hours.
• Each piano tuner works full time: 8 hours in a day, 5 days in a
week, and 50 weeks in a year.
• Each piano tuner works for 8 × 5 × 50 = 2,000 hours per
year, which yield 1,000 tunings per year.

Hence there are 70,000 ÷ 1,000 = 70 full-time piano tuners in

Chicago.

How would you improve this estimate?

14
 Refining the estimate
• The above estimation is just for households. Adding to this are pianos in
schools, churches, bars, …
• For schools, school age population in Chicago, which is from, say, 5 to 17
years, would be
17 − 4 ÷ 65 = 1/5 of the total population in Chicago, if all ages
were equally represented and all people die at age 65. Therefore there are
3,000,000 ÷ 5 = 600,000 students.
• If each school has about 500 students, there would be
600,000 ÷ 500 = 1,200 schools.
• Suppose each school has 2 pianos. This yields 2,400 pianos.
• Next we consider churches, bars, orchestras, … We may lump all these
pianos to the school pianos, to make it 10,000 pianos, so that the total
number of regularly tuned pianos in Chicago be raised to 80,000.
• This results in increasing the estimate of 70 full-time tuners to 80.

15
 Modifying assumptions
• Some assumptions may not be accurate, and need changes.
• E.g., If we modify the assumptions to:
• 1 piano in 20 households
• On average every piano is tuned once every 2 years
But keep that each tuner can tune 1,000 times a year, this will
yield an estimate of
1,000,000 ÷ (20 × 2 × 1,000) = 25 full-time tuners
• Adding institutional pianos might push this up to 30 or 40 full-
time tuners.
• At this point we have estimates within the range of 25 to 80
tuners. We can go on and on to make further refinement!

16
 How accurate is our estimate?
• So, how many piano tuners are in Chicago?

• In Chicago proper, there are 25 to 40 Yellow Page phone

numbers under “piano tuning – repair-refinishing” and at
least 4 more in the suburbs. Note that, more than one tuner
could use a common phone.
• The Chicago Piano Tuners’ Guild has 66 members. The
University of Illinois Music Library estimates that there are
two active tuners not in the guild for every tuner that is a
member; this would yield 198 tuners.

• There are other pathways of getting estimates.

17
 Tips and Cautions
• (https://www.youtube.com/watch?v=FQUcou
n99Vk)
1. Use round numbers in calculations or in final
answers.
– E.g. use 7 million for HK population, not 7,184,541.
2. Avoid long, complicated intermediate steps.
– Calculators may not even be needed!
3. Do not rely too much on internet/literature
search.
– This is not the point!

18
 Tips and Cautions
4. Ensure your assumptions make sense.
– E.g. everyone spends $1000 per visit to the hair stylist?
5. Try to explain your basic data.
– E.g. why men cut their hair 10 times a year?
– At least argue why this is not unreasonable.
6. Try to estimate the answer to within a factor
of 10.
– Good enough to make most decisions!
– E.g. how much will you pay for a video game?
• $30 - $300? Good enough!
• Let’s take $100.

19
 Tips and Cautions

7. Give upper and lower bounds.

– E.g. how many clowns can fit into a Volkswagen Beetle?
• 1 - 100!

20
 Tips and Cautions
8. The Use of geometric mean to find averages
in Guesstimation.
– Sometimes we know quite well about the upper and lower
bounds of a quantity
• Eg. the number of customers having noodle at the CYM Canteen
this morning: 1 - 100
• because you were there this morning and you did a simple survey
– In these cases, we estimate the average using the
Arithmetic mean:
1+100
1 ≈ 50 100
2

21
 Arithmetic vs Geometric Mean
8. But in Guesstimation, we would like to
guesstimate an average within, say, a factor
of 10.
1+100
1 2
≈ 50 100
× 50 ×2

50 may be a bit too “biased” towards 100…

• With geometric mean:
1 1(100) ≈ 10 100
× 10
× 10
Sounds good!

• Geometric mean can be a better “average” in guesstimation.

22
 Some Useful Numbers
The following information will be useful to your
guesstimation tasks:
• World population (and population of certain countries and
cities)
• Dimension of the earth (and dimension of certain planets,
distance between the earth and the sun, etc.)
• Area/dimension of certain countries and cities
• Some physical constants (e.g., acceleration of gravity at the
surface of the Earth, speed of light, speed of sound, …)
• Scale of the universe (e.g., http://htwins.net/scale2/)
• Conversions between units
• ……

23
 Exercises
1. How many people in the world are sneezing at this moment?
2. The world record (as of Oct 2011) of the number of digits of
𝜋 computed is 10 trillion. Suppose these digits are to be
printed in a book (say, about A5 size, with normal size
typeface). How thick will this book be?
3. How fast is the Earth travelling around the Sun?
4. At what speed is the equator of the Earth spinning?
5. If we stand together shoulder to shoulder, can all people in
the world fit in Hong Kong?

24
 Exercises
6. “Human Chain” Rally: How many people will be needed to
form a “human chain” by joining hands to span the entire
coastline of Hong Kong Island?

Hong Kong Island

source: hong-kong-travel.org

25
7. How many people were there in the gathering?

26