Basic Principles of Probability

and Statistics

Lecture notes for PET 472

Spring 2010
Prepared by: Thomas W. Engler, Ph.D., P.E

Definitions
Risk Analysis
Assessing probabilities of occurrence for each possible
outcome
Risk Analysis

Probabilities and prob. distributions

Representing judgments about chance
events

Modeling
Geologic, reservoir, drilling
Operations, Economics

Decision criteria
EV, profit, IRR

Present to management for decision

Definitions
Sample Space
Complete set of outcomes
(52 cards)

Outcome
Subset of the sample space
(drawing a 5 of any suit)

Probability
Likelihood of drawing a 5
P(A) = 4/52

Definitions
Equally likely outcomes
Have same probability to occur

Mutually exclusive outcomes

The occurrence of any given outcome excludes the
occurrence of other outcomes

Independent events
The occurrence of one outcome does not influence the
occurrence of another

Conditional probability
The probability of an outcome is dependent upon one or
more events that have previously occurred.

Rules of Operation

Symbol

Definition

Expression

P(A)

Probability of outcome A occurring

P(A+B)

Probability of outcome A and/or B

occurring

P(A+B)=P(A)+P(B)-P(AB)

P(AB)

Probability of A and B occurring

P(AB) = P(A) P(B|A)

P(A|B)

Probability of A given B has

occurred.

Rules of Operation

Addition Theorem
P(A+B)=P(A)+P(B)-P(AB)

Example
12
P
(
A
)
outcome A drawing 4, 5, 6 of any suit

52

outcome B J or Q of any suit

P(A

B)

20
52

P ( B)

P(AB)

A
B
Venn Diagram

Mutually
Exclusive
events

8
52

Rules of Operation

Addition Theorem
P(A+B)=P(A)+P(B)-P(AB)

Example
12
P
(
A
)
outcome A drawing 4, 5, 6 of any suit

52

outcome B drawing a diamond

P(A

B)

22
52

P ( B)

P ( AB)

B
Venn Diagram

13
52

3
52

Rules of Operation

Multiplication Theorem
P(AB)=P(A)P(A|B)

Example
outcome A drawing any jack

P(A )

outcome B drawing a four of hearts P ( B | A )

on the second draw
P(AB)

4
52

1
51

1
663

Sampling without replacement

- observed outcome is not returned
- series of dependent events

4
52
1
51

conditional

Rules of Operation

Multiplication Theorem
P(AB)=P(A)P(B)

Example
outcome A drawing any jack, return
outcome B drawing a four of hearts
on the second draw
P( AB)

4
52

1
52

P(A )

4
52

P ( B)

1
52

1
676

Sampling with replacement

- observed outcome is returned to sample space
- series of independent events

Probability Distributions
A graphical representation of the range and likelihoods of
possible values of a random variable

a variable that can have more

than one possible value, also
known as stochastic or deterministic

Probability density function

f(x), frequency

Random variable

x, random variable

Useful method to describe a range of possible values. Basis

for Monte Carlo Simulation.

Probability Distributions

Range frequency Percent

50 - 80
4
20%
81 - 110
7
35%
111 - 140
5
25%
141 - 170
3
15%
171 - 200
1
5%
20
100%

Divide into intervals

Or bins

40%

35%

30%

25%

20%

15%

10%

5%

0%

50 - 80

81 - 110

111 - 140
Net Pay, feet

141 - 170

171 - 200

Percent

Histogram representation
Of statistical data

frequency

Data
Well No Net pay, ft
1
111
2
81
3
142
4
59
5
109
6
96
7
124
8
139
9
89
10
129
11
104
12
186
13
65
14
95
15
54
16
72
17
167
18
135
19
84
20
154

Frequency distributions

Probability Distributions
Range frequency Percent
50 - 80
4
20%
81 - 110
7
35%
111 - 140
5
25%
141 - 170
3
15%
171 - 200
1
5%
20
100%

Cumulative frequency distributions

minimum
Range
50
80
110
140
170
200

Cumulative
Percent
0%
20%
55%
80%
95%
100%

maximum

Benefits
1. Can easily read probabilities
2. Necessary for Monte Carlo
Simulation

Cumulative percent

100%
80%

60%
40%
20%
0%

50

100
Net Pay, feet

150

200

Parameters of distributions
A parameter that describes central tendency or average of the distribution
Mean, weighted average value of the random variable
Median value of the random variable with equal likelihood above or below
Mode value most likely to occur

A parameter that describes the variability of the distribution

Variance, 2 mean of the squared deviations about the mean
Standard deviation, square root of variancedegree of dispersion of distribution abut
the mean

A
a< b

B
a= b

Parameters of distributions

Computing mean and standard deviation

1. Arithmetic average of discrete sample data set

N
x
i 1
N

N number of equally-probable values

N
(x
i 1

N
17.6
2.87

Core porosity and permeability

Depth
4807.5
4808.5
4809.5
4810.5
4811.5
4812.5
4813.5
4814.5

k,md
2.5
59
221
211
275
384
108
147

,%
17.0
20.7
19.1
20.4
23.3
24.0
23.3
16.1

4815.5
4816.5
4817.5
4818.5
4819.5
4820.5
4821.5
4822.5
4823.5
4824.5
4825.5
4826.5
4827.5
4828.5
4829.5
4830.5
4831.5
4832.5
4833.5
4834.5
4835.5
4836.5
4837.5
4838.5
4839.5
4840.5
4841.5
4842.5
4843.5
4844.5
4845.5
4847.5
4847.5
4847.5

290
170
278
238
167
304
98
191
266
40
260
179
312
272
395
405
275
852
610
406
535
663
597
434
339
216
332
295
882
600
407
479
139
135

17.2
15.3
15.9
18.6
16.2
20.0
16.9
18.1
20.3
15.3
15.1
14.0
15.6
15.5
19.4
17.5
16.4
17.2
15.5
20.2
18.3
19.6
17.7
20.0
16.8
13.3
18.0
16.1
15.1
18.0
15.7
17.8
20.5
8.4
17.6
2.87

Parameters of distributions

Computing mean and standard deviation

2. Values listed as frequencies in groups

n x
i i
n
i

i index to denote number of intervals

n frequency of data points in each interval
x midpoint value of each interval

n (x
i
i
n
i

i
1
2
3
4
5
6
7
8

Porosity
ni
interval frequency
7 x < 10
1
10 x < 12
0
12 x < 14
1
14 x < 16
10
16 x < 18
12
18 x < 20
8
20 x < 22
7
22 x < 25
3
42

Applicable for large data sets

Results are approximate

pi
xi
prob. midpoint
0.024
8.5
0.000
11.0
0.024
13.0
0.238
15.0
0.286
17.0
0.190
19.0
0.167
21.0
0.071
23.5
1.00

mean deviation variance

0.202
85.342
2.032
0.000
45.402
0.000
0.310
22.450
0.535
3.571
7.497
1.785
4.857
0.545
0.156
3.619
1.592
0.303
3.500
10.640
1.773
1.679
33.200
2.371
17.74

8.96
2.993

Parameters of distributions

Computing mean and standard deviation

3. Discrete probability distributions

x

drilling costs probability of range

p x
i i

p (x
i i

midpoint

$M
100.0
105.2
111.5
130.6
136.3
148.2
165.2
168.7
178.5
183.7
190.0

$M
0
0.007
0.040
0.229
0.093
0.225
0.278
0.035
0.066
0.021
0.007

102.6
108.4
121.1
133.5
142.3
156.7
167.0
173.6
181.1
186.9

pi is the probability of occurrence of the xith value

of the random variable

EV

xi*pi

(x i- )2

p(x i)(x i- )

$M

$M

($M) 2

($M) 2

0.7
4.3
27.7
12.4
32.0
43.6
5.8
11.5
3.8
1.3
143.1

0.7
4.5
29.9
12.7
33.3
45.9
5.9
11.8
3.9
1.3
149.9
15.8

1641.3
1208.5
486.8
93.4
0.7
184.6
568.2
929.5
1443.0
1912.9

10.7
48.3
111.5
8.7
0.2
51.3
19.9
61.3
30.3
13.4
355.6
18.9

Parameters of distributions

Computing mean and standard deviation

4. Cumulative frequency distribution

x

midpoint

drilling costs probability of range

$M
100.0
105.2
111.5
130.6
136.3
148.2
165.2
168.7
178.5
183.7
190.0

1.0

Cumulative probability

0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
100.0

120.0

140.0

160.0

Drilling Costs, $M

180.0

200.0

$M
0
0.007
0.040
0.229
0.093
0.225
0.278
0.035
0.066
0.021
0.007

102.6
108.4
121.1
133.5
142.3
156.7
167.0
173.6
181.1
186.9

EV
$M
0.7
4.3
27.7
12.4
32.0
43.6
5.8
11.5
3.8
1.3
143.1

xi*pi
$M
0.7
4.5
29.9
12.7
33.3
45.9
5.9
11.8
3.9
1.3
149.9
15.8

(x i- )2
($M)

1641.3
1208.5
486.8
93.4
0.7
184.6
568.2
929.5
1443.0
1912.9

p(x i)(x i- )
($M) 2
10.7
48.3
111.5
8.7
0.2
51.3
19.9
61.3
30.3
13.4
355.6
18.9

Types of distributions

Normal
Lognormal
Uniform
Triangle
Binomial
Multinomial
hypergeometric

Types of distributions

Normal

Characteristics

Cumulative frequency

Define by and
f(x)
Mode=mean=median
Curve is symmetric
x
Cumulative frequency graph is s shaped
Can normalize and obtain area (probability) under
the curve.

Types of distributions

Normal

Given a set of data how do you know whether it

is normally distributed?
Shape of curves
median = mean

Cumulative frequency

Examples: porosity, fractional flow

f(x)

Types of distributions

Lognormal
mode

Characteristics

median

Cumulative frequency

Define by and
f(x)
Modemeanmedian
Curve is asymmetric
x
Cumulative frequency graph exhibits rapid rise
Can transform to normal
variable by y=ln(x)

Types of distributions

Lognormal

Examples:
permeability
thickness
oil recovery (bbls/acre-foot)
field sizes in a play
mode
median

f(x)

Types of distributions

Uniform

Cumulative frequency

Characteristics:
f(x)
all values are equi-probable
specify min and max
allows for uncertainty
used in Monte Carlo simulation

min

max

max

100%

min

Types of distributions

Triangle

Cumulative frequency

Characteristics:
f(x)
all values are equi-probable
specify min and max
allows for uncertainty
used in Monte Carlo simulation

M, most likely

L, low

H, high

max

100%

min

Types of distributions

Triangle

Convert to cumulative frequency plot:

normalize to a 0 to 1 scale:
Define m as:

x'

x L
H L

M L
H L

M, most likely

For x m, cumulative probability is given by:

P( x )

(x )
m

For x > m,

P( x ) 1

f(x)

(1 x )
1 m

L, low

H, high

Types of distributions

Triangle

Example

f(x)

Estimated costs to drill a well vary from a minimum of $100,000

to a maximum of $200,000,with the most probable value at $130,000.

M, 130

Convert the probability distribution to a cumulative

frequency distribution

L, 100
x, random

x'

cumulative

variable

normalized

probability

100
110
120
130
140
150
160
170
180
190
200

1.0

0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0

0.000
0.033
0.133
0.300
0.486
0.643
0.771
0.871
0.943
0.986
1.000

0.8

Cumulative probability

(drilling costs)

H, 200

0.6

0.4

0.2

0.0
100

120

140

160

Drilling Costs, ($M)

180

200

Types of distributions

Binomial

Describes a stochastic process characterized by:

1.
2.
3.
4.

Only two outcomes can occur

Each trial is an independent event
The probability of each outcomes remains constant over repeated trials
Binomial probability equation is given by:
P(x )

n x
C p (1
x

p)

n x

where
x = number of successes (0 x n)
n = total number of trials
p = probability of success on any given trial
and the combination of n things taken x at a time
C

n
x

n!
x! ( n x )!

Types of distributions

Binomial

Example
Your company proposes to drill 5 wells in a new basin where the chance of
success is 0.15 per well
What is the probability of only one discovery in the five wells drilled?
What is the probability of at least one discovery in the 5-well drilling
program?
1.0

P(x)

discoveries

0
1
2
3
4
5

0.4437
0.3915
0.1382
0.0244
0.0022
0.0001

Cumulative

0.9

P(x)

0.8

0.4437
0.8352
0.9734
0.9978
0.9999
1.0000

0.7

Cumulative

0.6
P(x)

Number of

0.5
0.4
0.3
0.2
0.1
0.0

Number of discoveries

Types of distributions

Multinomial

Describes a stochastic process characterized by:

1.
2.
3.
4.

Any number of discrete outcomes

Each trial is an independent event
The probability of each outcomes remains constant over repeated trials
Multinomial probability equation is given by:
P ( x , x ..., x )
1 2,
r

x x
x
n!
1
2
r
p p
... p
2
r
x ! x !... x ! 1
1 2
r

where
r = number of possible outcomes
x1 = number of times outcome 1 occurs in n trials
x2 = number of times outcome 2 occurs in n trials
xr = number of times outcome r occurs in n trials
n = total number of trials
pr = probability of outcome r on any given trial

Types of distributions

Multinomial

Example
Your company proposes to drill 10 wells in a new basin where the chance
of success is 15% per well
What is the probability of obtaining 7 dry holes, 2 fields in the 1-2 mmbbl
range and 1 field in the 8-12 mmbbl range?
outcome
range
mmbbl
1-2
2-4
4-8
8-12
probability
of dry hole

probability
of
outcome
0.08
0.04
0.02
0.01
0.150
0.850

number of trials (wells) in program

probability of dry holes
probability of 1-2 mmbbl
probability of 2-4 mmbbl
probability of 4-8 mmbbl
probability of 8-12 mmbbl

n=
x1 =
x2 =
x3 =
x4 =
x5 =

10
7
2
0
0
1
0.7%

Types of distributions

Hypergeometric

Describes a stochastic process characterized by:

1.
2.
3.
4.

Any number of discrete outcomes

Each trial is dependent on the previous event (sampling without
replacement)
The probability of each outcomes remains constant over repeated trials
Hypergeometric probability equation for two possible outcomes:
d

P(x )

where

N d
1
1
C C
x
n x
N
C
n

n=number of trials
di = number of successes in the sample space before the n trials
xi = number of successes in n trials
N = total number of elements in the sample space before the n trials
Cab = the number of combinations of a things taken b at a time.

Types of distributions

Hypergeometric

Example
Our company has identified ten seismic anomalies of about equal size in a
new offshore area. In an adjacent area, 30% of the drilled structures were
oil productive.
If we drill 5 wells (test 5 anomalies) what is the probability of two
discoveries?
number_s a mpl e
number_pop
popul a ti on_s
s a mpl e_s

n=
N=
d1 =
x1 =

5
10
3
2
42%