This article was downloaded by: [Queensland University of Technology]

On: 13 October 2014, At: 08:39

Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK

The Journal of Mathematical

Sociology
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/gmas20

Theory of Aces: Fame by

Chance or Merit?
a

M. V. SIMKIN & V. P. ROYCHOWDHURY

Department of Electrical Engineering , University

of California , Los Angeles, CA, USA
Published online: 16 Aug 2006.

To cite this article: M. V. SIMKIN & V. P. ROYCHOWDHURY (2006) Theory of Aces:

Fame by Chance or Merit?, The Journal of Mathematical Sociology, 30:1, 33-42, DOI:
10.1080/00222500500333258
To link to this article: http://dx.doi.org/10.1080/00222500500333258

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the
information (the Content) contained in the publications on our platform.
However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness,
or suitability for any purpose of the Content. Any opinions and views
expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the
Content should not be relied upon and should be independently verified with
primary sources of information. Taylor and Francis shall not be liable for any
losses, actions, claims, proceedings, demands, costs, expenses, damages,
and other liabilities whatsoever or howsoever caused arising directly or
indirectly in connection with, in relation to or arising out of the use of the
Content.

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden. Terms & Conditions of access and use can be found at
http://www.tandfonline.com/page/terms-and-conditions

Journal of Mathematical Sociology, 30:3342, 2006

Copyright # Taylor & Francis Inc.
ISSN: 0022-250X print/1545-5874 online
DOI: 10.1080/00222500500333258

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

Theory of Aces: Fame by Chance or Merit?

M. V. Simkin
V. P. Roychowdhury
Department of Electrical Engineering, University of California,
Los Angeles, CA, USA

We study empirically how fame of WWI fighter-pilot aces, measured in numbers of

web pages mentioning them, is related to their achievement or merit, measured
in numbers of opponent aircraft destroyed. We find that on average fame grows
exponentially with achievement; to be precise, there is a strong correlation ( 0.7)
between achievement and the logarithm of fame. At the same time, the number of
individuals achieving a particular level of merit decreases exponentially with the
magnitude of the level, leading to a power-law distribution of fame. A stochastic
model that can explain the exponential growth of fame with merit is also proposed.
Keywords: fame, merit, meme, stochastic process

An objective measure of achievement is difficult to define. As a result, the

question of how fame is related to merit is ill posed. Fortunately, there is
at least one case where an unquestionable measure of achievement does
exist. This is the case of fighter-pilots, for whom achievement can be
measured as a number of opponent aircraft destroyed. The website
(www.thearodrome.com/aces) contains names of all WWI fighter-pilot
aces1 together with the number of victories each had achieved.
These days there is an easily assessable index to fame: the number of
web pages (as found using Google) that mention the person in question
(Schulman, 1999; Schulman & Boissier, 2001). In the future we will refer
to it as the number of Google hits. This approach is similar to measuring
the impact of scientific papers in numbers of citations (Garfield, 1979) and
the importance of web pages in numbers of incoming hyperlinks (Brin &
1

An ace is a fighter pilot who achieved five or more victories.

Address correspondence to M. V. Simkin or V. P. Roychowdhury, Department of
Electrical Engineering, UCLA, Los Angeles, CA 90095 (310) 206-4975. E-mail: simkin@
ee.ucla.edu; [email protected]
33

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

34

M. V. Simkin and V. P. Roychowdhury

Page, 1998). Unfortunately, it is not enough to paste a name into Google

search window and record the number of search results, because many
aces have namesakes. For this reason search results were filtered to contain words like pilot or ace2. In many cases this method did not remove all
of the namesakes and Google search results had to be visually inspected to
complete the filtering. As it would be very time consuming to repeat this
procedure for all 1,849 registered aces, the study was limited to include
only German aces (392 total).

FIGURE 1 A scatter plot3 of fame versus achievement for 392 German WWI
aces. The correlation coefficient of 0.72 suggests that 0.722 52% of the variation in fame is explained by the variation in achievement. The straight line is
the fit using Eq. 2 with b 0.074.
2

The complete list of used filter words is: flying, pilot, ace, flieger, Jasta, Fokker, and
WWI.
3
There are many aces with identical values of both achievement and fame. Therefore
for display purposes random numbers between zero and one were added to every value of
achievement and fame. This way the scatter plot represents the true density of the data
points.

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

Theory of Aces

35

Figure 1 is the result of this crusade. The correlation coefficient

between achievement and the logarithm of fame is 0.72. In contrast,
the correlation between achievement and fame (without logarithm)
is only 0.48. The significance of the correlation coefficient, r, is that
r2 is the fraction of variance in the data which is accounted for by linear regession. This means that about half (0:722 52%) of difference
in fame is determined by difference in achievement.
For a dozen of aces, achievement was not limited to the amount
of fights won. For example, Max Immelmann is more known for his
aerobatic maneuvers and Rudolf Stark for his books. Unfortunately,
several aces acquired additional fame for less esteemed achievements. This implies that the true correlation between achievement
and fame is more than the above-mentioned value of 0.72.
While fame is not perfectly correlated with achievementdifferent
measures of fame do perfectly correlate between themselves. Some 33
German WWI aces were famous enough to have books written about
them. Figure 2 shows the scatter plot of book-fame versus Googlefame. The correlation coefficient of 0.985 suggests that 0:9852 97%
of the variation in number of Google-hits is explained by the variation
in number of books.
The frequency distributions of achievement and fame are shown in
Figures 3 and 4 correspondingly. One can see that achievement is distributed exponentially and that fame has a power-law tail.
When the distribution of achievement, A, is exponential,
pA a expa  A;

and fame, F, grows exponentially with achievement,

FA expb  A;

then fame is distributed according to a power law. Clearly, elimination

of A from Eq. (1) using Eq. (2):
1
AF lnF;
b

dA

dF
bF

leads to:
a
pF Fc ;
b

a
c1 :
b

Substituting the values a 0:083 and b 0:074, obtained from the

least-square fits of the data (see Figs. 1 and 3 into Eq. (3) we get
c 2:1 which is quite close to c 1:9 obtained by fitting the actual
distribution of fame (see Fig. 4).

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

36

M. V. Simkin and V. P. Roychowdhury

FIGURE 2 Number of Google hits versus number of books for 33 German

WWI aces, who have at least one book written about them. The correlation
coefficient of 0.985 suggests that 0.9852 97% of the variation in book-fame
is explained by the variation in Google-fame. The data on the numbers of
books were provided by F. Olynyk.

Exponential growth of fame with achievement leads to its unfair

distribution. With 80 confirmed victories Manfred von Richthofen is
the top-scoring ace of the WWI. With 4,720 Google hits4 he is also
the most famous. The total amount of opponent aircraft destroyed by
German aces in WWI is 5050. At the same time there are 17,674
Google hits for all of the German aces. This means that Manfred von
Richthofen accumulated 4; 720=17; 674 27% of fame, while being
personally responsible for shooting down only 80=5050 1:6% of
4

The data used in the paper were collected around May 2003. Todays numbers of
Google hits are different.

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

Theory of Aces

37

FIGURE 3 The distribution of achievement (number of victories) obtained

using a sample of 392 German WWI aces. The straight line is the fit using
Eq. 1 with a 0.083.

opponent aircraft. On the opposite side 60 lowest scoring aces (with 5

victories each) together shot down 300 aircraft, or 300=5050 5:9% of
all aircraft destroyed. However, together they got only 463 Google hits,
or 463=17; 674 2:6% of fame.
A simple stochastic model can explain why fame grows exponentially with achievement. It is convenient to describe the dynamics of
fame in terms of memes (Dawkins, 1976) (we use this word in the
sense of a piece of information, which can pass from one mind to
another). We define the fame of X as the number of people who know
X, or, in other words, the number of memes about X. In practice we
cant count the number of memes, but we can count the number of
webpages. It is natural to assume that the number of webpages, mentioning X, is proportional to the number of memes.
The rate of encountering memes about X is obviously proportional
to the current number of such memes in the meme pool. We
will assume that when someone meets a meme about X, the probability that it will replicate into his mind is proportional to Xs

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

38

M. V. Simkin and V. P. Roychowdhury

FIGURE 4 The distribution of fame (number of Google hits) computed using a

sample of 392 German WWI aces. The straight line is the fit p(F) / F c with
c 1.9.

achievement (which thus plays the role of memes Darwinian fitness). The rate of the spread of a meme about someone with
achievement A is thus:
s nA:

Here n is an unknown independent of A coefficient, which comprises the effects of all factors other than achievement on meme
spread. The expectation value of the number of memes obeys the
following evolution equation:
dhFi
shFi nAhFi:
dt

If at time 0 there was only one copy of the meme the solution of
Eq. (5) is
hFti exp nt  A;

b nt:

which is Eq. (2) with

Theory of Aces

39

In case of the aces t is time passed since WWI. After substituting

Eq. (8) into Eq. (3), we get:

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

c1

a
nt

The value of c is consistent with the experimental data on aces if

nt a.
Eq. (7) gives the expectation value of fame for a particular value of
achievement. One can also ask what the distribution of fame is when
the value of achievement is fixed. The probability that after time t
there is just one copy of the meme is:
p1 t expst:
To have exactly 2 copies of the meme at time t, the meme should
have replicated at some intermediate time, s, and afterwards two
resulting memes should not replicate during time t  s. The probability of this is:
p2 t

sp1 sexp2st  sds

expst1  expst
It can be verified by induction that the probability to have F copies
of the meme at time t is:
pF t expst1  expstF1

After substituting Equations (5) and (8) into Eq. (9), we get:
pF A expbA1  expbAF1

10

Using Eq. (10) we can obtain the exact formula for the distribution
of fame in our model: see Appendix A.
The above derivation is almost identical to the one, used by Yule to
explain the power law in the frequency distribution of sizes of biological genera. Yule, 1925 considered two types of mutations: specific,
which lead to a new specie and occur at rate s, and generic, which
leads to a new genus and occur at rate g. The total number of genera
grows with time as exp(g  t). Therefore the number of genera of age
t is proportional to exp(g  t). The number of species in a genus of
age t is / exp(s  t). The problem is identical to ours with achievement,

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

40

M. V. Simkin and V. P. Roychowdhury

A, replaced with age of a genus, t, and factors a and b with g and s

correspondingly. Therefore it follows from Eq. (3) that the exponent
of the Yules distribution is c 1 g=s. Apart from Yules process
our findings are relevant to some problems studied recently in the
context of evolving networks [7] (see Appendix B).
According to Eq. (10), the probability to get a fame, which is ten
times more than what you can expect for your achievement is as small
as e10  5  105 , while the probability to get it ten times less is as big
as 1=10. This should give a confidence to the famous and relief to the
unknown.

APPENDIX A
To obtain the distribution of fame we integrate the product of the
distribution of achievement (Eq. (2)) and conditional probability of
fame given the value of achievement (Eq. (8)):
pF

dApF ApA

dAexpbA1  expbAF1 aexpaA


Z
a
a 1
a
a
1; F

dxxb 1  xF1 B
b 0
b
b


a
a C b 1 CF



b C a1F
b

Here B and C are Eulers Beta and Gamma functions. Using known
properties of C function one can obtain the large F asymptotic of the
above:

a
1
C
a
b
pF /
a
b F b1

APPENDIX B
The problem of finding distribution of fame when the value of achievement is fixed is identical to that of finding degree distribution of nodes
of the same age in a growing network. Krapivsky & Redner, 2001 solved
this problem (see section IV.A of their paper) and found the distribution

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

Theory of Aces

41

to be exponential in agreement with our results. Note that our derivation, which followed that of Yule, is simpler.
The problem of finding distribution of fame for all values of achievement is identical to that of finding degree distribution of nodes of the
same age in a growing network, in which the nodes have exponentially
distributed multiplicative fitness. Such network was studied using
numerical simulations in, (Bianconi & Barabasi, 2001) but no analytical results were obtained. One can use results of this paper to show
that the degree distribution of nodes of the same age in such network
obeys a power law.
Our results are also relevant to theory of citing. Redner (1998)
observed that the distribution of citations to papers published during one year and cited next 17 years follows a power law. This
observation cannot be explained by a cumulative advantage model
(Yule, 1925; Simon, 1955; Price, 1976), in which the rate of citing
a particular paper is proportional to the number of citations it
had already received. Krapivsky and Redner (2001) have shown that
cumulative advantage leads to exponential distribution of citations
to papers of the same age. The results of the present paper suggest
that Redners observation can be explained by a cumulative advantage model with multiplicative fitness (Bianconi & Barabasi, 2001),
if we assume that the distribution of fitness is exponential. One
may speculate that this fitness is related to merit, as it was in
the case of aces.

REFERENCES
http:==www.theaerodrome.com=aces.
Bianconi, G. & Barab

asi. (2001). Competition and multiscaling in evolving networks.
Erophys Lett., 54, 436, http:==arxiv.org=abs=cond-mat=0011029.
Brin, S. & Page, L. (1998). The anatomy of a large scale hypertextual web search engine.
Computer Networks and ISDN Systems, 30, 107. Available online at: http:==www.
db.stanford.edu=
backrub=google.html
Dawkins, R. (1976). The Selfish Gene. Oxford University Press.
Dorogovtsev, S. N. & Mendes, J. F. F. (2002). Evolution of networks. Advances in
Physics, 51, 1079.
de S. Price, D. (1976). A general theory of bibliometric and other cumulative advantage
process. Journal of American Society for Information Science, 27, 292.
Garfield, E. (1976). Citation Indexing, John Wiley: New York.
Krapivsky, P. L. & Redner, S. (2001). Organization of growing random networks. Phys.
Rev. E, 63, 066123, http:==arxiv.org=abs=cond-mat=0011094.
Redner, S. (1998). How popular is your paper? An empirical study of citation distribution. Eur Phys. J. B, 4, 131, http:==arxiv.org=abs=cond-mat=9804163.
Schulman, E. (1999). Can Fame Be Measured Quantitatively? Annals of Improbable
Research, 5, No. 3, p. 16 (available online at: http:==members.bellatlantic.net=

vze3fs8i=air=fame.html); Schulman, E. and Boissier, S. (2001). How Should

42

M. V. Simkin and V. P. Roychowdhury

Downloaded by [Queensland University of Technology] at 08:39 13 October 2014

Fame Be Measured Quantitatively? http:==members.bellatlantic.net=
vze3fs8i=air=

fame2.html
Simon, H. A. (1955). On a class of skew distribution functions. Biometrika, 42, 425.
Yule, G. U. (1925). A mathematical theory of evolution, based on the conclusions of
Dr. J.C. Willis, F.R.S. Philosophical Transactions of the Royal Society of London B,
213, 21.