Digest: March 17

Digest: March 17

Post by katherine william » Wed, 18 Mar 1998 04:00:00

Triathlon Digest: Tuesday, March 17, 1998
Vol. 2, No. 62
Published by Triathlon Central

In This Issue:
        * Announcement: African Regional Triathlon Champs (from Liesbeth
Stoltz, South Africa)
        * Results: South African Triathlon Championship (from Liesbeth Stoltz)
        * Results: Copa America Triathlon (from Cecilia Manazzoni, Argentina)
        * Results: Powerman Honolulu (courtesy of the Running Room, Honolulu)
        * Rankings: Statistician John Walker responds to comments

Notes from Katherine:
        -- Phil Briars in Auckland has forwarded a correction on the
Ironman New Zealand results. Tim DeBoom, credited with fourth place, was
actually a DNF - everyone moves up one spot below him. Tenth place belongs
to Yugi Fujiwara from Japan.
        -- The announcement that should have come from ITU on Monday,
regarding a second-tier schedule of 30 races worldwide, has been delayed,
apparently, until Wednesday.

- --------------------
From: Liesbeth Stoltz

Subject: African Triathlon Championships

The African Triathlon Championships and regional race for the African
continent will take place on 28 March at Swakopmund on the Skeleton Coast
of Namibia. If you can visualise the contrast of a desert meeting an ocean:
this is it! Huge red sand dunes from the desert meet the cold water of the
South Atlantic.

Four countries (Namibia, Zimbabwe, South Africa and the island Mauritius)
will send teams of four men and three women each to race in the Elite race.
Wild cards will be given to athletes from outside Africa who want a rare
experience of Africa.

The route will consist of a three lap swim, a flat out and back cycle route
(4km will be on a salt surface) and a flat one lap run loop. It will be a
draft-legal race.

The entry fee is US$ 30 and the nearest international airport is at Windhoek.

Enquiries:
Bert Laaser or Urda Hoffmann at fax (264) 61-220403

(from Liesbeth Stoltz in South Africa)
- --------------------
From: Liesbeth Stoltz

Subject: Results for South African Champs

SOUTH AFRICAN TRIATHLON CHAMPIONSHIPS

Elite and Junior men:
1 Glen***(Central Gauteng) 1:47.05
2 Greg von Holdt (Central Gauteng) 1:48.03
3 Raynard Tissink (Central Gauteng) 1:49.30
4 Richard Woolridge (KwaZulu-Natal) 1:49.34 (junior)
5 David Hyam (Eastern Cape) 1:49.40
6 Claude Eks*** (KwaZulu-Natal) 1:51.14
7 Glen Errington (Central Gauteng) 1:51.16
8 Lieuwe Boonstra (Western Cape) 1:52.09
9 Paul Siebert (Central Gauteng) 1:53.22 (junior)
10 Carl Storm (KwaZulu-Natal) 1:56.36 (junior)

Elite and Junior Women:
1 Kim Carter (Central Gauteng) 2:02.29
2 Lizel Moore (Australia) 2:05.04
3 Trudi Sanders (KwaZulu-Natal) 2:07.29
4 Fiddy Gey van Pittius (Natal) 2:08.26
5 Rowena Curr (Western Cape) 2:09.05 (junior)
6 Louann Rivett (Central Gauteng) 2:11.18
7 Tracy Weare (Western Cape) 2:12.35 (junior)
8 Kathleen Shuttleworth (Western Cape) 2:13.49
9 Ulrica Roux (Western Cape) 2:14.01
10 Christina Scannell (Western Cape) 2:15.00 (junior)

Notes: The South African Triathlon Championships took place at Midmar Dam
near Howick in the scenic Natal Midlands, this weekend. Unconventional rules
saw the Elites and Juniors racing in the same wave, on a draft-legal
out-and-back route. With water temperature of 21C the juniors were in
wetsuits while the elites raced without wetsuits.
        In the women's race, variation was added in the form of Lizel
Moore, an Australian, and Trudi Sanders, who left South Africa as an
average age-group athlete and started training with a group in Cronulla
near Sydney. For her efforts she was awarded with a third place in the
elite race and an opportunity to race the African Regional Championships.
The women's race was dominated by Kim Carter, the French Grand Prix
champion for 1997.

- --------------------
From: Cecilia Manazzoni

Subject: Results for Copa America Triathlon

6ta. SEMANA INTERNACIONAL DE TRIATLON
Sunday, 15 March in Mar del Plata, Argentina
Copa America Olimpic Distance (1,5 Km, 40 Km, 10 Km)

Men
1. Oscar Galindez (ARG), 1:44:40
2. Daniel Fontana (ARG), 1:44:42
3. Raul Lemir (ARG), 1:47:29
4. Carlos Cremaschi (CHI), 1:47:30
5. Tie, Eligio Cervantes (MEX), 1:48:43 and
5. Carlos Probert (MEX), 1:48:43
7. John Freddy Tibocha (COL), 1:48:48
8. Ricardo Cordero (COL), 1:50:30
9. Francisco Pontano (ARG), 1:50:45
10.  Ronan Pavoni (ARG), 1:51:19

Women
1. Maria Morales Rendon (COL), 2:01:31
2. Maria Clara Veron (ARG), 2:04:54
3. Maria ***ia Coronel (ARG), 2:05:36
4. Nidia Kontratavicius (ARG), 2:06:35
5. Patricia Correa (ARG), 2:07:37
6. Fiorella D'Croz (COL), 2:09:38
7. Paola Brignole (ARG), 2:11:10
8. Nancy Alvarez (ARG), 2:11:36
9. Gisela C***(ARG), 2:11:54
10. Silvia Villanueva (ARG), 2:12:23

- ---------------------
Results for Powerman Honolulu, 15 March
8k R, 60k B, 15k R
Courtesy of the Running Room shop in Honolulu

Men
1. Michael Tobin, Boise, ID, 2:52:06
2. Tony Olsen, New Zealand, 2:56:46
3. John McEvoy, Tucson, AZ, 2:58:50
4. Chris Willis, St. Helena, CA, 2:59:15
5. Louis De La Torre, Kona, HI, 3:00:00
6. Brendan Brazier, Victoria, BC, Canada, 3:00:54
7. Bruce Erskine, USA/Australia, 3:01:23
8. Tim Monaco, Encinitas, CA, 3:01:38
9. Gary Flacke, Santa Barbara, CA, 3:02:46
10. Brian Taylor, Norman, OK, 3:04:23
11. Mike Murphy, Vancouver, BC, Canada, 3:04:34
12. Jeff Sanders, Kehei, Maui, HI, 3:06:01
13. Peter Hursty, Honolulu, HI, 3:12:12

Women
1. Jane Despas, Australia, 3:25:24
2. Debbie Hornsby, Honolulu, HI, 4:14:02

- --------------------
From: John Walker

Subject: Response to comments on rankings

Katherine has passed along some well-thought-out comments on the rankings
we recently published in the Triathlon Digest.  I'll try to summarize
the comments and respond to them.  While I can't expect to convince
everyone that this is absolutely the best method (I can't even convince
myself of that), I can hopefully explain why things are done the way they are.

But before I go any further, I want to explain a little terminology that
I use.  I will refer to ratings as the points that an athlete has (5000,
6234, etc.).  I refer to the ranking as how an athlete's rating compares
with the ratings of all the other eligible athletes (1, 2, 3, 15, 72, etc.).

Comment #1: A number of athletes appear to be losing points after winning
races. Even worse, some athletes appear to lose points when winning against
strong fields, and gain points when winning against weaker fields, which
is counter to the idea that there are more points to gain at tougher races.

Answer #1: My response to this requires a description of two things: a
description of the strength of field, and some startup issues related to
getting an initial guess for an athlete's rating.

Strength of Field:
The strength of field numbers are just a weighted average of the ratings
of all the athletes in the race.  They give more weight to the higher-ranked
athletes than to the lower-ranked ones.  The premise is to give one number
that reflects both the quality and the size of the field.

A race against a single person rated 9000 is going to be harder to win than
one against 10 people rated 6000.  But a race against 10 people rated 5100
is going to be easier to win than one against 300 people all rated 5000 (one
of them is bound to have a great day).  No single number can perfectly
do this, but it should at least take it into consideration.

Now that I'm finished describing the strength of field, I'm going to make
the somewhat surprising revelation that it doesn't make any difference.
Consider two races.  Race A has 4 athletes, rated 5000, 5500, 5800, and
6000. Race B has 4 athletes, 3 rated 5500 and one rated 5800.  The athletes in
both races have the same average rating (for simplicity, I'm not using
weighted averages in this example).  But the athlete rated 5800 is most
likely going to place 2nd in race A and 1st in race B.

An average (straight or weighted) does not give enough information to
accurately predict how an athlete is expected to finish.  So the
rating/ranking program instead looks at the ratings of every athlete in the field
when adjusting the ratings.  The strength of field is computed separately.  I
initially computed it to help me analyze the actions of the program, since I am
not familiar enough with all the athletes to trust my own judgment on
the strength of all the races.

Although this may not be immediately apparent, points are not created or
destroyed based upon the results of any race.  Points are exchanged
among competing athletes.  If you do better than expected, you essentially
take points from the athletes who did worse than expected.  I have
tried to be especially careful about this so the ratings as a whole do
not grow astronomically large or extremely small over time (one of the
earlier methods I tried had this problem, which became apparent only
after I ran thousands of computer-simulated races).

But even though points are exchanged, this doesn't mean the strength of
field will remain constant when the same athletes compete.  Examine what
happens when Al and Bob, both with ratings of 5000, race each other.  If Al
wins, then their ratings might be 5100 and 4900 afterwards (I'm making those
numbers up).  The field strength of a rematch race would be higher because Al's
higher rating would be weighted more heavily than Bob's lower rating.
This is despite the fact that the total number of rating points has
remained constant.

Startup Issues:
In trying to develop the rating system, one of the trickiest issues was
assigning initial ratings to athletes.  Rather than add in my bias (or
anyone else's) by assigning initial ratings based upon my perception
of the 1000+ athletes, I decided to let each athlete's first few races
do this.  To allow ratings to quickly settle into a rather appropriate
level, they are not adjusted in quite the same way for an athlete's first
few races.  Initially, the ratings are allowed to fluctuate pretty wildly.
In the rankings announcement on Saturday, these are described as
"provisional" ratings.  After that, the ratings are bound a little more
tightly.  These are the "transitional" ratings described in the announcement.
Only after that does the athlete get a "standard" rating and the adjustments
completely follow the rules described.  I expect this rapid initial
adjustment to be of increasing importance, once the system has been around
a while.  This should allow new athletes to be assigned a rating reflective
of their ability much more quickly.

All of the examples I have seen where an athlete's rating has gone down
after winning a race have been when the athlete was in this "startup"
phase.  Once an athlete has a standard rating, this will not happen.
Currently, this happens after 5 races, but that is subject to change
as I investigate fairer ways of handling startup issues.  Please note
that nearly all the description of how the ratings should react are
based upon athletes who already have standard ratings.

Comment #2: Long-course specialists appear to be at a disadvantage because
they don't race as often.  People like Van Lierde, Hellriegel, Zack, and
Bowden aren't included.  Could we include long course athletes with fewer than 5
races?  Could we weight long races more heavily?

Answer #2: You guys have scored another bullseye by finding another thorny
issue.

I would really love to try to include athletes with fewer than 5 races,
but I really don't feel that those ratings would be accurate.  Perhaps
once there are more ranked athletes, the initial ratings of new athletes
will adjust more quickly and we can do that.  But at the current time,
we just don't have enough history of race results for me to feel
comfortable with it.  But since we don't throw away race results after 1 year,
people like Hellriegel and Bowden should get added soon.

To start with, I'll admit that these all-inclusive rankings are
inherently somewhat unfair to long-course specialists.  More precisely,
all-inclusive rankings are somewhat unfair to anyone who does any
races besides his/her specialty.  Because there are limits even to what
a triathlete's body can endure, you find more long-course specialists
doing shorter races than the other way around, putting them at a
disadvantage in an all-inclusive ranking.  My preferred solution to this
would be to have separate long-course rankings, which include only the
results of long races.  Then Lothar Leder's long-course rating would
not be affected by his relatively lower finishes in ITU World Cup and
Formula 1 races.  But I would really like to get most of the kinks
worked out of this one set of ratings before I start adding another set.

As to weighting long races differently, I have been adamantly opposed
to that idea, figuring the stronger races will still make themselves
known.  But having the question explicitly asked has made me reconsider.
Because people cannot do long races as often, they tend to take each
one more seriously.  Shouldn't someone's rating be allowed to move as
much as a result of 3 IM races as 5 Olympic-distance races?  My ears
are open to input.

Comment #3: Is it really fair to weight more recent races more heavily than
older races?  Aren't IM Hawaii results just as important now as they
were two months ago?

Answer #3: There are two general methods of computing rankings: treating all
races over a given time period equally, or weighting more recent ones more
heavily.  Consider the two following questions: "Who was the best triathlete
in 1997?" and "Who is currently the best triathlete?"  Using an example,
consider Al and Bob.  Al's race finishes in 1997 were 1-2-3-4-5-6-7, and
Bob's finishes (in the same races) were 8-7-6-5-4-3-2.  Given these
results, I would say that Al was the better triathlete in 1997. But I would
also say that Bob was the better triathlete at the end of 1997.

Both answers are perfectly valid, but they are to different questions.  More
heavily weighting recent races more closely approximates the answer to
the question, "Who is currently the best triathlete?"

While I have the opportunity, I would like to point out one thing.  While
rankings and ratings are nice, let's not forget that there is a whole lot
more depth to every athlete than a single number.  Just because Bob is
rated 100 points above Al today doesn't mean he is clearly the better
athlete.  What if Al doesn't race well in the heat?  Or if Bob doesn't like
hills (but races hilly courses anyway)?  Or if every piece of shattered
glass has eyes only for Carl's tires?  Or if Doug is racing while sick or
injured?

I think you get the idea.  I think this is especially important when
considering athletes for things like awards and sponsorship. Decisions like that
are too important to entrust to a single set of numbers.

                         -- John

- --------------------
End of Triathlon Digest, which is sponsored by Triathlon Sourcebook.