Vous êtes sur la page 1sur 29

How Low Should You Go?

A Method to Calibrate Estimates


of P99 Prospect Reserves
Robert Otis
Paul Haryott

www.roseassoc.com

Acknowledgement
Thanks to Chevron Corporation
for use of data in the development
of this presentation

Background
Prospect reserve and many reserve
parameter distributions are lognormal
Data from a lognormal
distribution plot as a
straight line on logprobability graph paper

Probit scale
Log scale

Background
Prospect reserve and many reserve
parameter distributions are lognormal
Mid-point is the P50;
slope is proportional
to standard deviation

Probability scale is
greater than
convention P10 is a
large number; P90 is
a small number

Background
Prospect reserve and many reserve
parameter distributions are lognormal
Volatility or uncertainty
is another term from
for
standard (or
variance
deviation
standard
(or
variance) and
deviation)
but is
can
most
be
expressed
easily
quantified
most easily
by the
as the P10/P90
P10/P90
ratio ratio

The larger the P10/P90,


the more uncertainty or
the higher the variance
(standard deviation) of the
distribution

Background
Prospect reserve and many reserve
parameter distributions are lognormal

A straight line is defined by two


points thus, any lognormal
distribution can be completely
defined with two independent pieces
of information, e.g., mean and
variance, P10 and P90, P01 and
P10/P90, etc.

Pre-Post Drill Discovery Sizes


8th To 14th Licensing Rounds
Norwegian North Sea
Size of Discovery,
MMBOE

10,000

1,000

100

10

1
1

10

100

1,000

10,000

Expected Size before Licensing, MMBOE


Norwegian Petroleum Directorate, 1997

Observations
The P01 (high side) of most distributions is
fairly well constrained
Field size distributions can use the largest
field which is usually found early in the
exploration phase
Area distributions can use the closing contour
on structures and reservoir extent on
stratigraphic traps
Average net pay distributions can use a high
N/G and the thickest reservoir with an
appropriate geometric correction or averaging

Observations
Given a constrained upside or P01, we
can acknowledge a significant downside
by several approaches
Consider the smallest field in the trend and
force the distribution to honor both the
extreme upside (P01) and the extreme
downside (P99) connect with straight line
Assume an appropriate P10/P90 and pivot
from the P01 do a credibility check on the
estimated P99 and adjust as needed

Appropriate P10/P90?
One guideline (from the mid 1980s)
suggests a range of values based on type
of prospect and typical field size
distributions
Development 2 to 7
Extension 5 to 25
Trend 10 to 120
Rank 55 to 250
Frontier 120 to 650

Unfortunately, these values are not, for the


most part, supported with published data

What We Want To Do And Why?


What
Describe a method or process to calibrate
P10/P90 to appropriately represent the
uncertainty of a distribution

Why
Differing opinions are abundant, but are
mostly based on a wide variety of experience;
few, if any, are validated with data
We would like to see a data supported
process that allows a company to develop
their own rule(s) of thumb

Finding An Appropriate P10/P90


Because each company has a different
portfolio, there are no hard rules each
company will need to determine it
independently
To do this, each company needs a
history of consistent estimates of pre-drill
distributions and post-drill actual results to
use for calibration
If you have a history, apply the following
process

The Process
Compile your history wells require two
independent values to define the pre-drill
distribution and the post drill mean
Estimate the P01 from each pre-drill distribution
Assign a common P10/P90 to all wells and see
where the actual prospect result falls on each
distribution
Evaluate results using a percentile histogram
method from Otis and Schneidermann (1997)
Do this for several P10/90 ratios and see which
one is most appropriate

Percentile Histograms
52 MMBOE in
50-60 Interval

P01
P05

25 MMBOE in
90-100 Interval

P01

100

193 MMBOE in
0-10 Interval

P05

P01
P05

P10

P10

P10

P20
P30
P40

P10

P20
P30
P40

P20
P30
P40

P20
P30
P40

P60
P70
P80

P60
P70
P80

P60
P70
P80

P60
P70
P80

P50

P50

P50

P90

P90

P90

P90

P95

P95

P95

P95

1000

P99

P99

P99

10

P01

P05

P50

76 MMBOE in
10-20 Interval

10

100

1000

10

100

40-50

30-40

20-30

1000

P99

10

100

1000

3
2
1
0
90-100 80-90

70-80

60-70

50-60

10-20

0-10

Each percentile interval has a 10% probability


If the post-drill result from each pre-drill distribution
is random, the result is a uniform distribution

Diagnostics
50%
40%
30%
20%
10%
0%

50%
40%
30%
20%
10%
0%
90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Heavy on the downside too


optimistic
50%
40%
30%
20%
10%
0%

Heavy on the upside too


pessimistic
50%
40%
30%
20%
10%
0%

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Heavy on both up - and downsides


need to widen ranges

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Uniform distribution - acceptable

Lets Look at Some Data


Chevron international exploration wells
from 1989-1993 (total of 50 discovery
wells)
Prospects include trend, rank and frontier
wells
No appraisal wells

P10/P90 Ratio

P10/P90 For Years 1989 - 1993


20
18
16
14
12
10
8
6
4
2
0

Look Back
1989

1990

1991

Wells

1992

1993

Percentile Histograms
(1989-91)
(1992-93)
Average P10/P90 = 3.7

Average P10/P90 = 7.8

Actual Post Drills - Pre-Lookback

Actual Post Drills - Post-Lookback

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

P10/P90 Ratio

20
18
16
Too many results 14
12
below P90
10
8
About right on 6
upside (P10) 4
2
0

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Look Back
1989

1990

1991

Wells

1992

1993

Essentially
balanced

Observations
1989-91 wells had average P10/P90 of ~4
but needed to include more downside
1992-93 wells had P10/P90 of ~8 and had
reasonably balanced results
What would the histograms have looked
like if we used a variety of P10/P90s?

Range of P10/P90 Values


What would these wells look like with a
P10/P90 = 7?
P10/P90 = 10?
P10/P90 = 20?
P10/P90 = 50?
P10/P90 = 100?
P10/P90 = 250?

All Wells
Average P10/P90 Ratio = 7
Actual Post Drills

Post Drills w/7 P10/P90

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Not bad - a nice balanced distribution


Pre-drill, 31% of the wells were above P50
With P10/P90 = 7, 56% of the wells were
above P50

All Wells
Average P10/P90 Ratio = 10
Actual Post Drills

Post Drills w/10 P10/P90

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Although not bad, indications are that we


are becoming pessimistic on the downside
(P99 too low)
With P10/P90 = 10, 73% of the wells were
above P50

All Wells
Average P10/P90 Ratio = 20
Actual Post Drills

Post Drills w/20 P10/P90

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Worse than the original, and way too


pessimistic of the downside (P99 way too
low)
With P10/P90 = 20, 84% of the wells were
above P50

All Wells
Average P10/P90 Ratio = 50
Actual Post Drills

Post Drills w/50 P10/P90

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%
90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

90- 80- 70- 60- 50- 40- 30- 20- 10- 0-10
100 90 80 70 60 50 40 30 20

Point made . . .
With P10/P90 = 50, 93% of the wells were
above P50

Observations
For this data set, the process indicated
P10/P90 ratios for reserves vary between 5 and
10
Less than 5 too many results fall below P90
Greater than 10 too many results fall above P10

Quick Look process


Check number of results above and below pre-drill
P50
Rule of Thumb: Greater than ~70% in either bin
check your process

Let the process determine the appropriate


P10/P90 ratio for your company each portfolio
is different and needs to be examined.

Observations
The process should be applied to reserve
parameters as well (e.g., area, average net pay)
to determine appropriate P10/P90 ratios
For this data set, appropriate P10/P90 for area
ranged between 3 and 4
For this data set, appropriate P10/P90 for average
gross pay ranged between 6 and 8 however, a
balanced histogram was reached only by allowing to
P01 to be twice the original P01

Take-Aways
Reserves and key, high variance reserve
parameters have consistent P10/P90 ratios
capitalize on this observation
The high side is usually well constrained, and,
together with an appropriate P10/P90 ratio,
can define a very reasonable distribution as a
credibility check
Use your P10/P90 ratios as QC and if
aberrations are observed ask Why?
Appropriate P10/P90 ratios can only be
determined by examining your company history
and the basins explored of pre and post drill
results maintain these religiously

So, How Low Should You Go?


Low enough to have ~10% of your post
drill outcomes in the P100-P90 range of
your pre-drill distributions
When data are available, high enough so
that the distribution honors values in the
80% confidence interval (P90-P10)
This can be done by developing and using
your companys appropriate P10/P90 as
a calibration check

How Low Should You Go?


A Method to Calibrate Estimates
of P99 Prospect Reserves
Robert Otis
Paul Haryott

Thank You For Your Attention


www.roseassoc.com

Vous aimerez peut-être aussi