ALLOWANCE COMPARISONS
by
J. N. Newman and G. S. Hazen
February, 1975
This research was carried out under the North American
Yacht Racing Union Ocean Race Handicapping Project,
M.I.T.
OSP Project No. 81535.
The generous support of the individual
donors to this program is gratefully acknowledged.
AB S TRACT
Race results from twenty-one major ocean races are analysed in
con-junction with eight different time allowance systems.
A two-parameter
family of speed curves is used to characterize the races and to derive a
new time-on-time system.
Numerical figures of merit are derived for
comparing the different time allowance systems.
Significant differences
are noted between American and British races, which appear to justify
the traditiönal choices of time-on-distance and time-on-time respectively.
Extensive computer listings of the race results are included in the
Page
L Introduction
1Description of Race Data
5Description of Time Allowance Systems
7Least-Square Speed Curves---
13Figures of Merit
- 17Description of Computer Progr
23Modified NAYRU Systems
25Sununary and Conclusions
28References
- --- -- 29Appendix - Listings of Race Results
72(see separate list on page iii)
Figures
Expected performance corresponding to the NAYRU
time-on-distance and NAYRIJ time=on_t:irné systems - 9
Least-square values of the parameters
C1and
C2for
twanty-one races
-- - - 15
Summary of figure of merit for NAYRU time-on-distance
formula and variations of the coefficient 0.6
26Summary of figure òf merit for NAYRU time-on-time formula
and variations of the coefficient 0.0567
27Tables
1.
Listing of twenty-one races
------
- 42..
Time allowance formulae ---
6Mean of magnitudes of four figures-of merit and all races - 20
TABLE OF CONTENTS
(continued)
Race Results
Race Name
Page
Appendix
Page of
Listing.Figure
Number
5Seine Bay
30
73 6Harwich-Hook of Holland
3276
7LeHavre-Royal Sovereign
34 79 8Morgan Cup
36 82 9Cowes-Dinard
38 86 10Channel
40
92 11 Fas tnet 42 97 12Honolulu
44 104 13Sydney-Hobart
46 107 14Anclote Key
48110
15St.Pete.-Ft.Lauderdale
50 114 16Ocean Triangle
52 117 17Lipton Cup
54121
18.Miami-Nassau
56125
19Nassau Cup
58 129 20Bermuda Race
60 133 21Chicago-Mackinac
62 138 22LeHavre-Royal Sovereign
64 144 23Morgan Cup
66
147
24Cowes-Dinard
68150
25Channel
70155
NONENCLATIYRE
A
Factor of time in corrected time equation
B
Factor of distance in corrected time equation
C
Corrected time
C1,C2
Coefficients of speed curve, equation (10)
D
Race distance
E
Elapsed time
k
Arbitrary constant in time allowance system related to
scratch rating
N
Total number of boats 1n fleet
n
Subscript to denote n-th boat in fleet
R
Rating
Scratch rating
TOT
Time-on-time
RACE STATISTICS AND TIME ALLOWANCE COMPARISONS
by
J. N. Newman and G. S. Hazen
1.
INTRODUCTION
As part of an on-going effort to gather race data and use this to
judge different handicapping systems, and particúlar aspects thereof, the
results from twenty-one major ocean races have been collected from the United
States, Great Britain, and Australia, for the 1973-4 sailing seasons.
These
results, consisting of yacht names, sail numbers, ratings and elapsed times,
have been punched onto cards in a form suitable for computer analysis.
Inorder to process this data, and future race results, an extensive computer
program has been developed.
The program accepts simultaneously eight
dif-ferent time allowance systems, fOr use with the prescribed rating of each
yacht, and comparisons are made of the race results under each time allowance
system.
The output from this program forms the bulk of the present report, and
will be described in more detail in the following sections.
The principal
results include for each race a listing of fleet positions and corrected
velocities under each time allowance, corrected velocities being defined
as the ratios of race distance to corrected time.
In addition to these
tabular listings which are contained in the Appendix, the results are also
displayed in graphical form on computer-generated plots of elapsed and corrected
velocities vs. ratings.
Additional steps have been taken to quantify these race results in
terms of least-square curve fits to the elapsed and corrected velocities of
each race.
The parameters defining these curves can then be utilized as
statistical descriptions of the races, and as quantitative measures of the
efficiency of the different time allowance systems.
For this purpose, the
elapsed velocities have been fit with a two-parameter "speed curve" of a
defining this speed curve are stu4ied to reveal differences and similaritiçs
between the various races, as well as to validate the NAYRTJ system for the,
different races.
The corrected velocity distributions are f:it with straight lines,
which would be horizontal and independent of
rating if the handicap.png
process was free of bias toward the large or small boats.
The slopes of
these straight lines are therefore regarded as "figures of merit" by which
to compare the different time allowance systems in each race.
interesting
relationships emerge, not only with respect to the time allowance systems,
but also in comparing the British and American race results.
Initially this investigation was confined to a few existing
time-allowance systems .inclu4ing the NAYRU time-on-distance (TOD) formula, the
RORC performance factor system (PFS) used for British races during
1973-4,
two earlier RORC systems, and the Cruising Yacht Club of Australia
(CYCA)system used during the samé period.
Inevitably other systems were included
f or comparison, particularly since the RORC was investigating existing and
proposed time allowance systems during the Pall of 1974 for use during the
1975 season.
As a result the above five existing time allowance systems
were supplemented by three proposed new
time-on-time formulae (TOT), including
a modified Australian system denoted as cYCA2, a.
modified RORC system denoted
as RORC2, and a new formula developed during
the course of this investigat.on
denoted as 'NAYRU TÔT.
The latter system, which will be described in some
detail in Section 3, has taken on added interest since it has been
adopted
by the RORC for use in British races during the 1975 season.
Certain individuals have been instrumental in supplying race
results
and time allowance formulae.
In this connection the authors wish to thank
Messrs. Roger Dobson of the RORC, Graham Newland of the CYCA, James
NcCurdi
of the CCA, Robert Potter of the LNYA, Jack Wirt of
thé SORC, and Gordon
Curtis of the TPYC.
Some additional race data has been gathered
from the
yachting press.
Corrections will be gratefully received and used to correct
and update the data bank of race results.
Work on this project was initiated during the first author's sahbatLcal
leave, and special thanks are due to his wife Kathy for programming assistance,
-3-to the University of Adelaide for use of their excellent computing facilities,
and to the Australian-American Educational Foundation and the John Simon
Guggenheim Memorial Foundation for fellwship support.
All of the
calcu-lations reported here have been performed using the facilities of the MIT
Information Processing Center.
Seine Bay
Harwich-Hook of Holland
LeHavre-Royal Sovereign
Morgan Cup
Cowes -DinardChannel
Fastnet
Honolulu
Sydney-Hobart
Anclote Key
St.Pete.-Ft. Lauderdale
Ocean Triangle
Lipton Cup
Miami-Nassau
Nassau Cup
Bermuda Race
Chicago-Mackinac
LeHavre-Royal Sovereign
Morgan Cup
Cowes-Di"nardChannel
RACE NANE
DATE
SPONSORING ORGANIZATION
5/4/73
5/25/73
6/8/73
6/22/73
7/20/73
8/3/73
8/ii/73
7/4/73
12/26/73
2/i/74
2/6/74
2/14/74
2/23/74
2/25/74
3/1/74
6/21/74
7/24/74
5/17/74
6/21/74
7 /12/748/2/74
Royal Ocean Racing Club
ti t,
't
't i, t, t, f,Transpacific Y.C.
Cruising Y. C. of Australia
Southern Ocean Racing Conference
Table 1 -- Listing of twenty-one races
t, t, it t, i, ii t, t, t, t, i, t,
Cruising Club of America
Chicago Yacht Club
Royal Ocean Racing
Club
'i
t, Itit
t, t' t, tit'
't
i
i
'ti
it
t
t
ii
'i
t''t
't
't
't
t,'t
-5-2.
DESCRIPTION OF RACE DATA
The race data which has been analysed consists of the results for
twenty-one races held during the 1973-4 sailing seasons, in England, the
United States, and Australia.
These races are listed in Table 1.
Pertinent
data for each racé includes the race name and year, the length of the race
in nautical miles, and for each boat in the fleet the name, rating, and
elapsed time.
Elapsed times may be entered in hours, minutes and seconds,
or alternativly in decimal hours, and the program accepts which ever of
these two inputs Is given.
In those cases where the effective race, length
has been modified by the race cOmmittee this modified length is used.
Where
sail numbers are available for each boat these have been recorded, to facilitate
future identification of each boat on the computer tape ICR certificates.
The class or division in the race hs also been entered for each boat,
although this information is not used here.
No attempt is made to analyse boats separately within each class, or
to determine class finish positions, although this could be done simply by
running the program separately for each class.
Class performance may be
more significant than fleet performance from the racing standpoint, especially
in long races where separate weather systems may affect different portions
of the fleet.
But for purposes of judging the overall nature of the race,
and the equitability of different handicap systems, subdivision by class
seems less desirable.
The division between classes is arbitrary, and this
division may be crucial if the handicap system is not equitable throughout
the fleet.
Thus, in a situation where small boats are favored in the fleet,
one may expect that the smallest boats in each class will also
generally
be favored within their respective classes, and vice versa.
Under this
circumstance an individual boat's perfôrmance will be vitally affected if
it is placed at the top of one class, or alternatively at the bottom of
the next class.
Thus it seems clear that for purposes of handicap system
development and evaluation, one should strive for equitable handicapping
of the entire fleet, as opposed to each separate class.
By the same token,
one may anticipate that in most cases a handicap system which works well for
the fleet as a whole will be equitable for each class, and vice versa.
RORC
PFS
NAYRU
TOT
k Rh/2
i + 0.O57R:I
kl-A
1.5
-2.75R_h1'22.75
Ok
R1'2
[1.5 C/D +
k;l]R1/2
+ 1 - 2.75 C/D
k(R11'6- 0.96)
C/D
k[(R-8)''1 - 0.75].C/D
k(R1'2
+ 2.6)
C/D
k(.l999Rh/2
+ .34)
C/D
k
Rh/2
(1 + 0.057Rh/2)(C/D)
Table 2 -- Time allowance formulae.
Note that
R
is rating in feet, and
k
is an arbitrary constant
related to the scratch--boat-ratingj---R
,
suchthat
A
1.0
,B = 0
forFor the present analysis of race data
R5 = 29.0 feet.
Rh/2NAYRU
TOD
1.0
Ò.6(k
0..6 ± (cID-k)Rh/2
1.43
Rh/2
+ 2.6
RORC
TOD
l.O
k-R1#'2± 2.6
1.43 + (CID.-k)(Rh/2
CYCA2
TOT
k(Rh/6
- 0.96)
oCYCA
TOT
- 0.75]
ORORC
TOT
k(R1'2
+ 2.6)
oRORC2
TOT
k(.l999Rhi2
+ .34)
OFormula
A
V
3.
DESCRIPTION OF TIME ALLOWANCE SYSTEMS
The eight time allowance systems evaluated here are listed in Table 2.
These include two time-on-distance formulae, the RORC performance factor
system, and five time-on-time formulae.
All eight systems can be expressed
in terms of the common equation
C=AE+BD
(1)where
Cis corrected time,
Eis elapsed time,
D
is the race distance,
and
A
and
Bare coefficients depending on the particular time allowance
system and on the rating of each boat.
In the time-on-distance approach
A=l ;
for time-on-time systems B0 .
The complete formulae for these
coefficients are listed in Table 2 for all eight systems.
All time allowance systems are explicitly or implicitly related to
corresponding "speed curves", of hypothetical boat speed vs. rating.
Indeed,
-if al-1 boats f insh with idéntical corrected times.,
D ,in an "ideal" race,
it 'follows by solving (1) for the elapsed times of each boat that the velocities
will be
-=D/E-
A
CID - B
Thus, for a given time allowance system, (2) gives the velocity for a given
rating, such that the boat will perform .as expected in a given race.
Itshould be noted that this formula includes one descriptor
(CID)of the
race conditions, which will be large or small depending on whether the race
is slow or fast, respect-ively.
In add-ition,(2) contains two parameters
(A,B)
dependent upon the boat' rating, and thus its speed poténtial.
For
time-on-distance
örtime-on-time
thespecial choices of
A=l
orB=O
respectively restrict the flexibility of (2).
Table 2 includes the expected
velocity based on (2) for each of the eight time allowance systems.
As an exemple, let us first consider the NAYRTJ TOD formula, where Al.O
and
.B = O.6(Ç'"2 -
R_1/2)
,with
Rsthe rating of the scratch boat.
Thus, from (1), the corrected time is
C = E +
.6açh/2
-R'2)D
' (3)and from (2) the velocity of each boat is related to its corrected time
by the formula
1 (4) =C/D
0.61Ç1"2 + 0.6R1'2
Writing (4) in a slightly different form,
1.67 Rh/2
V
=
(l.67C/D -
çh/2)R1/2
(5)(In the present discussion 1.67 is used as a convenient approximation to
the inverse of 0.6.)
The NAYRU TOD system is thus equivalent to the family of speed curves
given by
v=
-12
1 + C2R
1.67 Rh/2
(6)
where. C2
depends on the.speed of the race.
From (5),
C2and the correctèd
time are related by
C2 = l.67C/D - Ç"2
(7)or
C
0.6(C
+
ÇhI'2)D
(8)Therefore the parameter
C2and the corrected time
Care trniquely related
to each other, and either one is suitable basis for determining the winner
of a race, ¿lthough the correctéd time is a more readily understood parameter.
To pursue this viewpoint further, the family of speed curves defined by (6)
is plötted in Figure 1 (solid lines), for different values of
C2ranging
f rom O to 1/4 .
These define performance curves along which boats of
different rating müst sail, in order to finish with the same corrected times
under the NAYRU TOD system.
Alternatively, if all boats in a given race
are plotted on this Figure, the boat lying on the
highest of the family of
sôlid curves will be the winner under the NAYRTJ TOD time allowance system.
It may be noted from (7) that the speed-length ratio
(or more precisely,
the speed-rating ratio) expected by the NAYRU TOD system is
V/R1
2= 1+ C2R1/2
1.67I0
9
5
(jtf
/
3--I
20
30
9
40
50
RATING (FEET)
Figure 1 -- Expected performance corresponding to the NAYRTJ
time-on-distance (solid line) and NAYRU time-on-time (dashed
line) systems.This equation is displayed to emphasize the fact that the speed-length
ratio associated with the NAYRU TOD system is not a constant for all boats
in the fleet, or for any particular boat within the fleet.
Instead the
speed-length ratio is expected to vary both with the rating R ,
and with
the conditions of the race as represented by the parameter
C2 .This
important feature is not sufficiently appreciated, and critics of the
NAYRTJTOD time allowance system often state incorrectly that it will only apply
when the boats involved are sailing at a hypothetical assume4 speed-length
ratio.
A more general speed curve is obtained if the coefficient
0.6
isallowed to vary, corresponding to a change in the effective race distance
Under the NAYRU TOD system.
This is equivalent to changing the constant
1.67
in (6), and hence to the form'la
C1R112
=
1 + C2R112
(10)where
C1and
C2are cOnstants defining the particular speed curve.
Equation
l0) wilibe used here not only to relate to the NAYRU TOD system,
but to derive the so-called NAYRU TOT time-on-tie system, and, also, to
fit on a post-race basis the elapsed velocity distributions of the fleet
to a least-squares curve which can then be used to describe the
character
of each race.
Before dong so, however, we note that in its most general
f àrm (10) can not be used as the basis for a (pre-determined) time allowance
system since the twO unknown coefficients
C1and
C2can not be determined
uniquely and independently from the single equation (10) for a given velocity
V
and rating
R
To overcome this difficulty it is necessary to restrict
one or the other of the two independent
coefficients, such as is done in
the NAYRU TOD system when
C1is set equal to
1.67An alternative assumption is to fix
C2and allow
C1to vary with
the conditions of the raòe.
This complementary approach lads to a
time-on-time system, with
k
R1'2
A
=
1C2R1/2
(Il)
il
-C1
=
D/Ck
(13)since, with these substitutions, (2) and (lo) are Identical.
Here
k
isan arbitrary constant, related to
R5 ,whióh caribe conveniently chosen
H
so that A
1.0 when
R
R
-Thus, to obtain a on-time formula analogous to the NAYRU
time-on-distance system,
and
denoted here as the
NAYRUTOT formula, it is only
necessary to find a
uitable constant
C2to use in the formula (1).
Initially we selected the value of C2
= .057f-rom a preliminary graphical
fit. to race data.
Subsequently this formula has been adopted by the RORC
for üse during the 1975 season in British Races, but with the coefficient
0.057
changed to
0.0567
.This minor modification with
k
= 0.2424
in,the numerator, gives the time correction factor
0.2424
Rh/2
A
= 1
0.0567 R112
(14)such that
A
- 1.0002for a scratch rating
R529.0
.In earlier work
(Ref. 1) results were presented based on our original value of 0.057, but
subsequently the program has been changed so that the handicap system
identified here as NAYRU TOT has the coefficient set equal to
0.0567
and thus agrees exactly with the system to be used in the 1975 RORC races.
The family of speed curves equivalent to the
NAYRUTOT system are
given by
C1R/2
V
=
1 + 0.0567
(15)and these are plotted as dashed lines in Figure 1, for comparison with the
NAYRUTOD system.
In this case the corrected time could be replaced by
the parameter
C1for purposes of determining race results, the boat with
the largest value of
C,1being the winner of the race under the NAYRIJ TOT
system.
Figure 1 can be used to empare the expected results from
handi-capping under thè NAYU TOD and TOT systems, the two being coincident when
C1
=
1.67
and C2 =0.0567
.This condition corresponds to a rather fast
race, with boat speeds between
6 and9.5
knots for ratings in the range
level of performance from the larger boats, and hence favors the small end
of the fleet., by comparison to the TOD system.
Thus, in all but the fastest
racing conditions, large boatS will perform the best under the NAYRU. TOD
system, and small boats will be favored by the NAYRU TOT system by comparison.
This difference is typical of all timS-ori-time. vs
timeondistance comparisons,
as noted earlier in. Reference 2.
'The above discussion
applies only to. a relative comparison btween
the two time allowance systems, however, and it would be premature to conclue
that small boats are unduly favored in a race sailed under the TOT system,
or conversely that large boats are favored in 'a race sailed under the TOD
system.
An evaluation of this important subject
ust be based on the
exam-ination of actual race data, and In particular on the considerat-ion of what
constitutes in practice the most. relistic family of speed curves..
13
-4.
LEAST-SQUARE SPEED CURVES
In order to give a quant:itative descrptjon of the character of each
race, it is desirable to fit a suitable least-squares curve to the elapsed
velocities of all boats in the fleet
For this purpose the family of
speed-curves defined by (10) has been adopted, the procedure being to
determine the two coefficients
C1and
C2in a least-squares manner so
as to give an optimum fit to the elapsed velocities.
If the
N
boats in a given race are denoted by ratings
R,
and(elapsed) velocities V,,
with n=l,2,..N,
the dtrect least-squares
approach would be to minimize the sum of the squarés of the errörs
C,R,/2
i +CR'12
2n
However this leads to nonlineâr equations for
C1and
C2 ,and to avoid
the difficulties associated with diret solutions an alternative approach
is adopted, based on minimizing the sum of thé squares of the time differençes
-i +
n
CR''2
in
Ç' -
(l/C11Çh/2
E-.This is a more conventiOnal
tOb1em involving the least-!squàrés detérmination
o
thé two àônstants
= (1/C)
and
(C/C,)
such that
-
rC1'2
)2= minimUm:
-(16)
Differentiating (16) with respect to
o.and
ad' setting these derivatives
é4Ual. to zero we obtain two simúltaneoùs eqüations for
o. aid:-N
i-
)[ç1'/2]
= o
(17)R_1/2
-
N
Ç1
R1h:
[
ç1l2j2
N
1Ç'
where in all cases
denotes the sum over all
N
boats in the fleet.
Finally, substituting in the expressions for
C1and
C2ci
=(
R'J2
-
N
Ç1
R;1
-N
I
Ç1
Ç'
p_1/2Vi R_1/2
-
I Ç1 I
Ç1'
I Ç'
Çhu/2
-NI
Ç1
Rhh'2
(18) (19) (20) (21)These last two equations are the basis for the computations of
C and C2which are plotted in Figure 2 and listed in the Appendix for each race with
typical values of
V
computed for ratings 21, 29, 40 and 60 feet,. as well
aS the slope
dv/dR
evaluated at
R=29
.The computed values of
C1 andC2
have also been used in (10) to construct the least-square speed-curves
for each race which are plotted in the elapsed velocity Figures 5a-25a.
The computed values of
Ç1and
C2plotted in Figure. 2.can be compared
with the assumed values of thesE coefficients in the two NAYRU time allowancé
Systems.
The NAYRU TOD assumed value
C1 = 1.67
is shown as a horizontal
line in this Figure, and the NAYRU TOT value
C2 ' 0.0567as a vertical lin.
Races which lie on or near one of.. these twö lines will be
well handicapped
by the corresponding system.
In.geiera, the U.S. races are, scattered
more-or-less equally about the NAYRU TOD line, indicating that the
coefficient
H
0.6 is a reasonable mean valúe for american races.
British races and the
Hobart Race tend to lie substantially below this line,
indicating that for
these races
C1should be. reduced or the coefficient
0.6
increas.
Regarding the NAYRU TOT system, Figure 2 indicates that while the va1ue
C2 = 0.0567 is a reasonab1e average, valúe
for all of the races, most of the
I
3.0
2.0
1.0
O
-0.I
/
-
15
-x
U.S. RACES
O
BRITISH RACES.
HOBART RACE
/
X
CI.67
X
- -;Ox
(NAYRUTOD)
C20.O567
(NAYRU
OT)
0.1
0.2
0.3
C2
0.4
0.5
Figure 2 -- Least-square values of the parameters C1 and C2
U.S. races are to the right of this line and most of the
other
races
are to the left.
Thdeed, for five Brit±sh races and the
73 Hobart Race,
C2
is negative, indicating that thè speed length ratio of the larger boats
in these races exceeded that of the Sma-ller bòats.
Finally, it is Obvious from Figure 2 that the most significant
trend is for ail of the races to be distributed not along the horizontal
NAYRU TOD line hor along the verticàl NAYRU TOT 1ne, büt instead along
a diagonal line where
C1and
C2increase simultaneously.
A fair
appröximation is
C1 = 1 + 6.25 C2 as shown, and this particular straight
line gives a relatively good fit simultaneously to all 21 races.
Indeed,
it can be confirmed from Figure 2 that this relation between
C1and
C2gives a reasonable "colläpse" simultaneously for the British and American
races, i.e., the scatter of all points about this one unified line is
less than the individual scatter of American races about the NAYRU TOD
line and British races about the NAYRU TOT line.
Noreover, the diagonal
line shown is also able to account for the very unusual conditions of the
1974 Lipton Cup Race, (shown by the point at the upper right corner of
Figure 2), where differences in elapsed time throughout the fleet were
minimal due to a wind shift, with small boats favored overwhelmingly., and,
on the other exterme, the 1974 Channel and Le Havre Races (lower left cÖrner)
where large boats were heavily favored.
Unfortunately it is not possible to use the diagonal lIne in Figure 2
as the basis for a time allowance system based on the speed fórmula (IO),
sincé the resulting family of speed curves is singular at
R = (6.25)2= 39 feet.
Nevertheless the correlatión in Figure 2 is quite striking, and should be
17
-5.
FIGURES OF MERIT
in order to provide quantitative measures of the equitability of each
time allowance system, in each race, and ultimately a summary of these for
all races, figures of merit have been developed which will be described here.
In an ideal situation, the distribution of performance throughout the
fleet should be independent of rating, with large and small boats performing
equally.
Thus if the results are expressed in terms of the corrected
velocities, equal to distance divided by corrected time, and these corrected
velocities are plotted on a graph vs. rating for each. boat, as in Figures
5b-25b, a measure of the quality of the time allowance system is the extent
to which the points on this graph are distributed horizontally.
In order
to express this measure quantitatively we fit to all the points on the graph
a least-squares straight line., that is a straight line drawn on the
graph
which has a minimum mean-square error with respect to all the points on the
graph.
The slope of this straight line can then be used as a measure of
equitability between large and small boats.
If the slope is positive it
indicates that large boats are favored in the race, and vice versa.
Ifthe slope is. zero and the straight line is horizontal one may conclude that
there is no bias in either direction.
This is the basis for figure of merit
number one.
The mathematical procedure for determining figure ofmerit one is as
follows.
Given a set of cOrrected velocities
V
and ratings
R
, wen
.n
wish to fit to these a straight line
V=ct-I-R
(22)so as to minimize the mean-squared error
(Vn - -
R)2
As before, differentiating with respect to
and
and equating these
derivatives to zero gives the simultaneous equations
N1
(V -cx-R)=O
Rn
n
11=1fl
J
(23)Solving
for the unknowns
ci.and
from Cramer's rule,
V
R2-.1VRR
n
n
nfl
._nN
R2
-
(R)2
N
VR -.
V1R
'(25)N1R2- (R)2
Figure of merit one is the slope of (22), equal to
,and indicates the'
trend towards favoring large
(positive)
or small
(negative)
boats.
One possiblé objection to the above is.that there :Ls no attempt to.
discriminate..between "good" and "bad" boats in the least-squares fit to
the corrected velocities.
It is generally felt that the well-sailed boats
are more important in judging a handicapping system, and this suggests a
weighting of the least-squares fit to favor the good performers.
In order
.to provide alternative types of weighting, three substantiallydifferent
approaches have been adopted, which are identif led here as figures of merit
two, three, and four.
In figure of merit two the. weight
actors are related
linearly to the corrected fleet positions, i.e., each boat is given a weight
factor equal to one plus the number of boats it has. beaten.
The mathematical scheme for accomplishing this. requires that each. of
the
N
boats be ranked according to its fleet positIon (denoted as
IP
in
the computer listings).
Assuming this has been done, with
n=ithe race
winner, etc., and
n=N
the last boat, the appropriate
weighted errors are
1
N
T(N-n+l) (V-ct-R)2
We then proceed as before to find
andso that this sum Is minimized.
The resulting value of
defines figure of merit two.
Note that the last
boat to finish the race has a small but non-zero influence on this figure
of merit.
(Here, as in all figures of merit, credit is given for beating
a DNF boat, but the DNF boat is iiot included
in
the least-squares fit.)
In
figure of merit three a more brutal approach is taken, in that all
boats lying below the middle of the fleet are eliminated and the remaining
(24)19
-half of the fleet is fit as in figure of merit one, the slope of this new
straight line being defined as figure óf merit three.
Finally, as an
alternative but equally arbitrary approach, all of the boats below the
straight line corresponding to figure of merit one are eliminated and the
remaining half of the fleet is fit to give figure of merit four.
Note
that figures of merit three and four both eliminate approximately one. half of
the fleet, but the elimination procedures differ unless figure of merit one
is zero.
Haviig introduced f our separate figures of merit, it remains to decide
which of the four is most significant for á comparison of the different
time allowance systems. Fortunately, while significant dif:ferences exist
between the individual figures of merit for each race, as shown in the
separate tabulations in the Appendix,
these differences do not persist
when the figures of merit are first averaged over all the races and then
cómpared.
For this purpose Tàble 3 shows the mean, for all twenty-one races,
of the magnitudes of each figure of mer:it for each time allowance system.
From Tble 3 the eight time allowance systems can be ränked, with the smallest
figure of merit corresponding to the best. time 'allowance system, etc., and
in general similar conclusions follow regardless of which figure of merit
is used.
In particular, regardless. of which of the four figures of merit is
used, the Performance Factor System is the worst (i.e., largest figure of
merit) of the eight time allowance systems, and with only one exception the
time-on-time systems are superior to the time-on-distance.
The exceptional
case is figure of merit three, where the NÁYRIJ TOD system is slightly better
than the RORC2 TOT system.
In view of the arbitrary nature of figures of merit three and four,
and the relatively continuous weighting of figure of merit two with respect
to fleet performance, the latter has been adopted for comparative purposes in
the results tO be described later, but all four figures of merit are tabulated
-in the Appendix for each race and time allOwance system.
Restricting our attention to figure of merit two, Table 4 presents the
values of this parameter for all eight time allowances and twenty-one races.
Also shown are the mean values of the magnitude. of, this figure of merit,
Figure
ofMerit
NAYRIJ:RORC
RORC
CYCA2
CYCA
RORC
RORC2
TOD
TOD
PFS
TOT
TOT
TOT
TOT
Table 3 -- Mean of magnitudes of four figures of merit and all races
0.0226
0.0219
.0.0288
0.0186
0.0173
0.0208
0.0188
0.6203
0.0195
0.02(46
0.0172
0.0162
0.0186
0.0175
30.0095
0.0098
0.0125
3.0.085
0.0092
0.0085
0.0096
40.0209
0.0211
0.0271
0.0165
0.0160
0.0185
0.3169
Table 4 -- Summary of figures of merit
NAYN'J
TOD -RONCTul)
NONCCYCA
PFS
TOTCYCA
TOE NOIÈC TOTROC2
îoî
\1AYRUTot
I 9,3
SE1E iAY NACE
0.0284
0.0243
0.0361
0.0i3i
0.015M
0.0225
0.0165
0.0164
HANWICH-HOOK
0.0313
0.0279
0.0424
0.0i(s
0.0147
0.0217
0,01b4
0.019
LE
HAVNE - NOVAL
0.0200
0.0161
0.0295
0.0069
0.0039
0.0116
0.005.1
0.0047
i1)N(,AN CUP
0.0214
0.0118
0.0310
0.0081
0.0059
0.012?
0.0060
0.0059
COWt.S-U I NANI)
0.079
0.0219
0.0300
0.02(3
0.0244
0.0328
0.0246
0.0244
CHANNEL NACE
- 0.01,7 -0.0221 -0.OIYÓ -0.-008f -0.uIlO -0.0042 -0,0116 -0.011s
FASTNET NACE
0.0229
0.0194
0.0301
0.0130
0.0112
0.0158
0.0115
0.0115
rIONULULU NACE
0.0036 «0.0119 -0.0023 -0.0043 -0.001M
-0.0067 -0.0118 -0.0070
SYL)NE Y-HO8AN T
0.017.
0.0129
0.0177
0.0193
0.0.189
0.0216
0.0166
0.0177
1i4
ANCLOTE
-0.0011 -0.0051
0.0042 -0,0138 -0.0131 -0.0130 -0.0179 -0.0161
T. LAU[)LNUALE
0.0040 -0.0008
0.008/ -0.0029
-0.0021
-0.0027 -0.0071 -0.0050
OCtAN TNIANGLE
0.0130
0.0066
0.0166
0.0082
0.0096
0.0081
0.0037
0.0060
L1tUN CUP
-0.0311 -0.03/9 -0.025s -0.0389 -0.0381 -0.0317 -0.0437 -0.0416
MIAMI-NASSAU
- 0.0143 -0.0258 -0.0204
-0.0026
-0.0016 -0.0023 -0.0087 -0.005'.
NASSAU CUP
- 0.0008 -0.00s5
0.0044 -0.0101 -0.0095 -0.0102 -0.0160 -0.0133
8ENMiJDA NACE
0.01d9
0.0113
0.0203
0.0171
u.00n
0.0163
0.0114
0.OlsI
CH1C4bO-MACr INAC
u.0029 -0.0010
0.0066 -0,0040 -0.0057 -0.0004 -0.00o8
-0.0064
LE
HAVNE - NUYAL
0.05/9
0.0549
0.0116
U.0U5
0.0381
0.0449
0.0387
0.0385
MONGAN CUP
0.0208
0.0145
0.019/
0.0:241
0.0185
0.0337
0.0228
0.0213
CUWES/DINARI)
-
0.00b7 -U.u0b
0.0043
-0.0208-0.0264
-0.0134
-0,0208 -0.0228
CHANNEL NACE
0.0651
0.0615
0.OTs2
0.Osid
0.0503
0.0.58?
0.0510
0.0508
MEAN 0f
MAuNITUL)E
81TISti RACES
O.02U
0.0264
0.0354
0,0211
0.02Do
0.0246
.0.0205
0.0203
USACES
0.0100
0.0114
0.01210.0115
0.0113
0,0108
0.0142
0.0129
ALL kACES
0.0203
0.0195
0.0246
0.01(2
0.0162
0.0186
0.0175
0.0170
For the average over ali races, the five time-on-time systems are
signi-ficantly better than the two time-on-distance formulae.
Among the five
time-on-tim
Systems the CYCA f orula is best, followed in òr4er by NAYRU TOT,
CYCA2, RORC2, and RORC.
When the méans are computed separately for the
British and U.S. races, however, significant differences emerge.
For the
British races time-on-time is superior, with YCA and NAYRU TOT the best.
But for the American races the time-on-distance f örmulae aré tore efféctive,
with NAYRU TOD the best of the eight.
Thus ve again see that American and
British races must be analysed separately.
Small differences between the figures of merit shown in Table 4 should
not be regarded as significant, since the definition of this figure oferi
introduces some degree of arbitrariness.
Moreover, one can envisage
small-modifications of the coefficients in the various time allowance systems
which would chaige the figures of merit and, if judged solely by this or
another measure, woùld lead to changes in the relative iâerit of the time
allowance systems.
But. large differences between the figures of merit shown
in Table 4 are thought to bé significant, and relevant to the selection of
equitable time allowance systems.
23
-6.
DESCRIPTION OF COMPUTER PROGRAM
In order to process the race results outlined in Section 2, using
the different time allowance systems listed in Table 2, a program has
been developed using the programming language PL/l and the computing
f acilities of the MIT Information Processing Center, notably an IBM 370-168
digital computer and a CALCOMP plotter.
As the CALCOMP and MIT-supplied
plotting subroutines are written in FORTRAN, the optimizing, PL/l compiler
was chosen in order that its interlanguage interpreting capabilities could
be utilized.
In the absence of the optimizing compiler, however, the same
language interfacing capabilities could be accomplished using the standard
PL1F compiler.
Brief ly, the program accepts as input on computer cards of specified
format the race data, including two cards to describe the race ánd one
card for each boat.
The eight time allowance systems are treated as an
external subroutine which can be changed readily without re-compiling the
main program.
Considerable effort was expended in order to make the program as
versatile and simple to use as possible.
Consequently, the first card in
the input data deck is for describing the execution options desired.
These
are entered using the PL/l data format, and all or any of the options may
be specified.
In the absence of explicit specifications, each option's
default value will be used.
On a primary level there are options to include
printed, punched and plotted output.
If plotted output is specified two
additional options are availáble controlling labeling
and size of plots.
As has already been mentioned, the time allowance systems can be changed
by means of changing a single subroutine.
In conjunction with this change
the second card of the input data deck contains the eight titles to be used
to describe the time allowance systems.
Following thede two control cards the data is entered race by race.
The first card in each race deck contains its title and date, followed by
a card specifying the race length and scratch rating used.
Next each boat
is entered on one card with fields for the boat's name, sail number, class,
rating and elapsed time (decimal hours or hours, minutes and seconds).
DNF boats are indicated on the input cards by an elapsed time of zero
The
end of a raca is designated. by a card with
a blank chàiacter in
o1umn one,
and a non-blank character in column. two.
Similarly, the eùd of the last
race is flagged by a cárd with blanks in both the first änd second columnS.
In summary, the essential functions performed by the program are:
Read in race data from cards.
Convert elapsed time to decimal hoùrs if necessary.
Compute eight corrected times and corresponding corrected velocities
using
R5
29 feet.
Determine fleet positions of each boat under each time allowance system.
Print, if desired, a listing for the fleet of each boat's name, sail
number, rating, elapsed time, and fleet position followed by corrected
velocity for each of the eight systems.
Compute and print summary data including figures of merit, and
least-squares values of
C1and
C2for Speed equation (10) as well as
representative values of velocity and derivative of velocity with
respect to rating from this speed curve.
Cönstruct if desired, CALC0
plots of elapsed and corrected velocities
as well as the least-squares speed curve and figures of merit.
Punch, if desired, siary data.
25
-7.
MODIFIED
NAYRTJ SYSTS
Questions regarding the optimum value of the coefficient
0.6in the
NAYRU TOD system are well known, and this coefficient has been varied over
the years both for general use and for particular races.
Similarly, the
optimum value of the coefficient
0.0567
in the NAYRIJ TOT sysem may be
questioned.
In order to address these questions the eight time allowance
systems listed in Table 2 were replaced temporarily, in the program described
in Sectidn 6, by sequences of eight NAYRU TOD systems where the coefficient
0.6
was varied between
0.45
and 0.8in steps of
0.05
,and by sequences
of eight
NAYRU TOT
systems where the coeffIcient
0.0567
was replaced by a
range of values between
O and0.1
.All
twenty-one races were re-run
with these two modified programs, to determine the effect on the corrected
times of these changes.
To avoid voluminous output from these computer
runs, only the figures of merit were tabulated, and the magnitudes of figures
of merit
two averaged for all twenty-one races to provide quantitative
measures of the optimum values of the coefficients in the two systems.
The results of these additional investigations are plotted in Figures
3 and 4 for the
NAYRUTOD and TOT systems, respectively, as curves of the
figure of merit vs. change in the coefficient.
Recalling that the optimum
figure of merit is a minimum value, it follows that the coefficients of the
two time allowance systems should be chosen to coincide with the minimum
points of the curves in Figures 3 and 4.
Figure 3 shows that the optimum
value of the coefficient in the
NAYRTJTOD system is very close to
0.6
forthe average of all U.S. races, but is on the order of
0.8for the British
races and
0.75
for all 21 races.
Figure 4 shows the corresponding optimums
for the coefficient of the
NAYRIJ TOTformula, with U.S. races a minimum near
0.09
,British races near
0.01 ,and the average of all 21 races a very
flat insensitive minimum near
0.04
.Also we note that if the
NAYRTJ TODsystem is used with an optimum coefficient for U.S. races, it will give
approximately the same average figure of merit
0.01
as does the
NAYRU TODsystem.
But for British races the optimum TOT system is substantially better
than the optimum
TODsystem.
These results confirm our earlier conclusions
based on the comparisons of the coefficients
C1 and C2in Figure 2, as
05
Q6
07
08
COEFFICIENT
Figure 3 -- Summary of figure of merit for NAYRU time-on-distance
.o.o3
w
Lio O02
w
IL
004
o..o i
-
27
-O05
o.'
COEFFICI ENT
Figure 4 -- Summary of figure of. merit for.NÄYRtJ
time-on-timé.
8.
$t1NMARY AND CONcLUSIONS
.. ,.
Based onthe aata bank of race results, and a computer program for
.
analysing these in conjunction with varoús time allowance systems, it has
been possible to perform exteiisive investigations qf the race data and
various time allo-ance systems..
Further work along the stne lines can be
carried out Tjt
future race data, and a similar program can be used to
evaluate modifications of. the rating rule in conjunction. wLth the existing
race data.
It is by no means certain that the twenty-one races studied to
date are a sufficiently large sample to be statistically relevant, and the
analysis of additional race data should be made before drawing f irth conclusions.
Neve±theleàs the sample jncluded here is more extensi.ve than any kflown to
have been studied. systematically in the past, and some tentative conclusion
can be set forth for short-range purposes and discussion.
generali differeñcês bétwee
races are moré importaflt than differences
between time-allowance systems.
Thus race results are not stroly depeident
on which of the eight time alIowanòe systems is used, to. the same extent
that the results depend on race conditions
Thi
is apparent quantitatively
from Table 4 where the variation of the figures of merit is much greater
among the different races than it
is among the various time allowance systms
for-a given racé.
In this context unjustified criticism has been directed
at particular time allowance systems, as in a magazine account of the 1974
Morgan Cup Race where the victory
of "BES" is
attributed to the Performance
Factor System; from the results for this race shown in the Appendix it can
be confirmed that the same boat would win under all but one of the eight
tithe-allowance systems.
.In order to provide a quantitative description of the varying race
conditions we have used a two-parameter family of "speed curves" derIved as
a generalization from the NAYRTJ time-on-distance formula.
The two parameters
C1. and
C2. which thereby characterize a particular race are plotted in
Figure .2, and listed, in the Appendix.
Figure. 2 displays vividly the differences
between BrLtish and Ameridan racés, and the associated advantages of
time-on-time vs. time-on-time-on-distance systems respectively.
But a unifying trend is
also apparent in Figure 2, na.ely a relátion between the two parameters
9-British and American races, and also for some unusual race conditions such
as were met in the 1974 Lipton Cup Race.
This type of analysis suggests
that we are still far short of an optimum time allowance system.
Our study of time allowance systems has led to the development of a
new time-on-time formula, based on the same general family of speed curves
as the NAYRU time-on-distance system.
This new time-on-time system, denoted
for convenience as the NAYRU TOT system, has been selected by the RORÇ for
use in Britain during the forthcoming season, including the Admiral's Cup.
Thus our comparison here of the results to be expected from this system, in
comparison with the NAYRU TOD system and other British an4 Australian time
allowance formulae, may be used to give some indications of the results to.
be anticipated from the 1975 RORC races.
Detailed computer listings for each of the twenty-one races and eight
time allowance systems are included in the Appendix, and the essential results
of these computations are plotted in a more digestable form in Figures 5-25,
which are shown on the following pages.
Yachtsmen curious to see how they
would have fared "if only the race had been sailed under the right time
allowance system" may wish to study the Appendix.
Others can evaluate the
time allowance systems and race ch.aracteristics in the more descriptive form
of Figures 5-25, where anonymity of boats may also contribute to the
objec-tivity of the conclusions.
REFERENCE S
"Handicapping Systems for Ocean Racing Yachts", by J. N. Newman, paper
to be presented to the Royal Institution of Naval Architects 20th March 1975
meeting and published in the July 1975 issue of The Naval Architect.
"A Fundamental Approach tó Ocean-Racing Handicap Rules", by J. N. Newman,
Proceedings of the Symposium on Yacht Architecture, HISWA, Amsterdam, 1973,
p. 187.
>-I..
C)
D
w
>
H-3
D
-2
i. I I I I I I f20
30
110
50
60
70
HATING
(FEET)
Figure 5 -- Velocity plots for the 1973 Seine Bay Race.
Figure a
(above) shows the actual velocity of each boat in knots, defined
as race distance divided by elapsed time.
Also shown in this
f igre is the least-square speed curve represented by equation
10, with the coefficients C1 and C2 determined from equations
20-21. Figure
b (opposite page) shows the corrected velocities
of each boat in knots, defined as the race distance divided
by corrected time, under each of the eight time allowance systems
listed in Table 2.
Also shown in this figure are the
least-square straight lines corresponding to figure of merit two,
these straight lines being weighted for each boat by its fleet
position.6 D D H-D
z
-6 D -D H- LI-6 (n -Lj_ X X1973 SEINE BAI RACE
31
- H-D - H-L) >-L)1973 SEINE BIA RACE.
2Ó 30 Lb 50 60 70 20 O- LIb 50
RATING CEET)
RATING
(FEET)F-D
H
X X X1.973 HRHkICHHI OF H.O.LLRND RCE
H'
I IFigure 6
Velocity plots for the 1973 Harwich-Hook Q
Holland
Race
Figure a (above) shows the actual velocity of each boat
in knots, defined as race distance divided by elapsed time..
Also shown in this figure is the least-square speed curve represented
by equation 10, with the coefficients C1 and C2 determined from
équations 20-21.
Figure b (opposite page) shows the corrected
velocities of each boat in knôts,. definéd as the race dïstañce
divided by corrected time, under each of the eight time allowance
systems listed iñ Table 2.
Alo showr in this figure are the
least-square straight lines corresponding to figure of merit
two, these straight lines being wéighted for each boat by its
fleet position.20
30
LID
.50
60
70
-
33
-1973 HPRNICHHOK OF HOLLAND ARCE
1973 HPR1JICHHOO
OF HOLLANO ARCE
ARTING (FEET)
ARTING [FEET]
173 LE HAVRE-ROYAL SOVEREIGN HACE
I II--
I i XF-5
-XXX
xX
Xz
X XX X2-J.
.1:
L
i20
30
140
50,
RATING (FEET)
Figure 7
VelocIty plots for thè 1973 Le Havre-Royal Sovereign
Race
Figure a (above) shows the actual velocity of each boat
in knots, defined as race distance divided by elapsed time
Also shown in this figure is the least-square
speed curverepre-sented by equation 10, with the coefficients C1 and C2 determined
from equations 20-21
Figure b (opposite page) shows the corrected
velocities of each boat in knots, defined as the race
distancedivided by corrected time, under each of the eight time
allowancesystems listed in Table 2
Also shown in this figure are the
least-square straight lines corresponding to figure
of merittwo, these straight lines being weighted for each boat by its
f leét position.WI:
70
7
1973 LE HRVE-RTRLSDVEREIGN flRCE
-
35
-I I I
21J 30 40 50. 60 70
RFITJNG (FEET)
1973 LE -RVRE-R0YRL SOVEE.TGN ARCE
20 30 14Q 50
ARTINC (FEETÌ
H-1973 MORGAN CUP RACE
X
X
X
'igure 8 -- Velocity plots for the 1973 Morgan Cup Race.
Figure a
(above) shows the actu1 velocity of each boat in knots, defined
as race distance divided by elapsed tizne.
A]so shown in this
figure is the least-square speed curve represented by equation
10, with the coefficients Cl and C2 determined from equations
20-21.
Figure b (opposite page) shows the corected velocities
of each boat in knots, defined as the race distance divided
by corrected time, under each of the eight time allowance systems
listed in Table 2.
Also shown in this figure are the
least-square straight lines corresponding to figure of rneit two,
these straight lines being weighted for each boat by its fleet
position.I I I I I I
20
30
Lb50
60
70
20 30. LIO 50 60
- 37
j5.T9
6-RATING
[FEET)
RATING [FEET]
70 20 . 30 LIO
.50
60 . 70-J
3
Cr).
F
)QX
z
xx
S-,-,*
XX
X X X>-
x.
F-r-'
D
XÇ.)
X X. ,- X
LU
X XF-
_X X
XD
1973 CONES-DINAR.D HRC
X X X X X X X I --
I : I - I20
-30
110
50
60
70
HATING
(EEET)
Figure 9 -- Velocity plots for the 1973 Cowes-Dinard Race.
Figüre a (above) shows the actual velocity of each boat in knots,
defined as race distance divided by elapsed timé.
Also shown
in this figure is the least-square speed curve represented by
equation 10, with the coefficientS C1 and C2 determined from
equations 20-21.
Figure b (opposite page) shows the corrected
velocities of each boat in knots, defined as the race distance
dividèd by corrected time, under each of the eight time allowance
systems listed in Table 2
Also shown in this figure are the
least-square straight lines corresponding t9 figure of merit
two, theSe straight lines being weighted for eadh boat by its
fleet poSitionX
U)
D
O)D.
:- *
C1P2 TOT
oi -J BD!RC PFS. NAYHU TOD cil ai:
+'
y'
''
t V * :+ + V t V, * *+*_V.
t. .
1
-t'V
HORC. TOD I I I20.
30
1973 CHANNEL RACE
Q
X X XH
I: i::.
f I I50
HATING
(FEET)
X60
Figure 10 -- Velocity plots for the 1973 Chanñel Racé.
Figure a(above) shows the actual velocity of each boat in knots, defined
as race distance divided by elapsed time
Also shown in this
figure is the least-square speed curve represented by equation
10, with the coefficients C1 and C2 determined f romequations
20-21.
Figure b (opposite page) shows the corrected velocities
of each boàt i.i
knots, defihed as the race distance divided
by corrected time, under each of the eight time allowance systems
listed in Table 2.
Also showi in this figure are the
least-square straight lines corresponding to figure of merit two,
these straight lines being weighted for each boat by its fleet
position.(D -J w D D
z z
nl w D D pl (D -J w D Dz z
nl f-J
D D nl NAYFU TOT ACRC2 TOTROFIC Toi
CIClI TOT
)
O) a O) O)U)
D
z
OED
D3
1.973 FASINET RPE
X2Ò
30
140
50
60
70
HATING
(FEET)
Figure 11 -- Velocity plots for the 1973 Fastnet
Race. Figure a(above) shows the actual velocity of each
boat in kñots, defined
as race distance divided by
elapsed time.Also shown in this
figure is the least-square speed curve represented by equation
10, with the coefficients Cl and C2
determined from equations
20-21.
Figure b (opposite page) shows the corrected
velocitiesof
ach boat in knots, defiñed as the race distance
dividedby corrected time, under each of the eight
time allowance systemslïsted in Table 2.
Also shown in this figure are the
least-square straight lines corresponding to figure of merit two,
these straight lines being weighted for
eách boat by. its fleetOE L)
*. I.»
bLI
.' D DF
L) D +r
* t1973 FRSTNET RRCE
3,0 LI0 50 aO RAT 1N 'IFEET]= 43
- >-L) 6-q1973 FRSTNET R.PE
I 'I .1 F. F--D -* =5--- * 30 ' LIO 50 ' 'SO ' 70 -, RT'1'NÒ .(FET)1973 HONOLULU RACE
-
MOCÓ RATINGS
.50.
60
RAT ING
(FEET)
Figure 12 -- Velocity plts for the 1973 Honolulu Raqe.
Figure a(aboyé) shows the actual velocity of each bOat in knots, defined
as race distance divided by elapsed time
Also shown in this
figure is the least-square speed curve represented by equation
10, with the coefficients C1 and C2 determined from equations
20-21.
Figure b (opposite page) shows the corrected velocities
of each boat in knots, defined as the race distance divided
by correctéd time, üñder éach of the eight time allowance Systems
listéd in Table -2.
Also shown in this figure are the
least-square straight lines corresponding to figure of merit two,
tiese straight lines being weighted for each boat by its fleet
position..D D Li
7-Li Li 5 20 30 L0 50RATING
[FEET)1973 HONOLULU RACE - MOO. RATINGS
t. I I t I I
:.
:
45
-
7-6-
'+4+,.
Li D D -5 7 71973 HONOLULU RACE - MOO. RATINGS
,
I I -t -4 I
60 70 20 30 Lb .. 50 60 70
x
XXX
XXX
XXX
X1973
SYDNEI-HOBRHT RPÌCE
X* X X X X X X X X X X X X X XFigúre 13 -
Velocity plots for the 1973 Sydney-Hobart Ràce.
Figure a (above) shows the actual velocity of each boat in knots,
defined as race distance divided by elapsed time.
Also shown
in this figure is the least-square speed curve represented by
equation 10, with the coefficients C1 and C2 determined f röm'equations 20-21. Figure b (opposite page)
hows the corrected
velocities of each boat in knots, defined as the raöe distance
divided by corrected time, under each of the eight time allowance
systems liSted in Table 2.
Also shown in this figure are the
least-square straight lines corresponding to figure of merit
two, these straight lines being weighted for each boat by its
fleet position.
20
30
LIO
50
60
7q
RRTING (FEED
U-)F
z
XX X X X, X XR)