Ever wonder if 10,000 hours is really the baseline? Did some set of events make the difference in someone getting to the olympics? WorldRowing has captured data on races and athletes for decades but little has been done to analyze the data across countries, athletes and races to the possible outcomes and where investment might be best to identify impact to rowing at a high performance level. Data for each race, for each athlete is stored on the WorldRowing website. Varying amounts of personal information, plus race information, is available for athletes spanning several decades. This talk investigates the athlete race data from WorldRowing.com and demonstrates an end to end walk through of a data analysis problem using Python.
2. Who are we?
Lou Harwood
CTO/Co-Founder at
Skedaddle (use code
LOU for $5 off your first
ride!)
Andrew Fisher, PhD
Biomedical Engineering
Researcher at Boston
University
Disclaimer: Neither of us are
affiliated with WorldRowing.com
4. Project Creation
Several Olympic rowers were discussing the current state of USRowing, and
trends in athlete development
Several statements were made about the transition rate from the Under 23
national team to the Senior national team (Olympic team)
7. Competitions/Regattas
Multiple teams compete head-to-head
Heat, quarter-final, semi-final, final
scheme
National teams organized primarily by
age
under 18 (junior), under 23 (U23), and
senior (open) levels
2009 World Rowing Championships. Photo Credit: Ainunau, flickr
11. List of Boats Within Data
Sculling
1x Single
2x Double
4x Quad
Sweep
2- Pair
2+ Coxed Pair
4- Four
4+ Coxed Four
8+ Eight
Photo Credit: Dellavasca, Wikimedia Commons
49. Demographics of World Rowing Population
Junior athlete count: 18542
U23 athlete count: 8335
Worlds athlete count: 12452
Olympian athlete count: 5101
Count Junior & U23: 4029
Count U23 & Worlds: 3403
Count Worlds & Olympics: 3356
Count U23 & Olympics: 1059
Count Junior & Olympics: 1413
Ratio of Juniors who did U23: 0.217
Ratio of Juniors who became Olympians: 0.076
Ratio of U23 who became Olympians: 0.127
Ratio of Olympians who did Juniors: 0.277
50. Data Analysis - Junior Team Representation in
Olympic Boats
Top Nations for Olympians
with Junior Team Experience
USA 1158
GBR 668
AUS 581
CAN 531
NED 530
ITA 522
FRA 474
CHN 445
DEN 415
NZL 361
ITA 94
GBR 86
GER 79
FRA 78
ROU 61
AUS 55
USA 51
POL 49
NED 46
BUL 42
Top Nations for
Olympians
51. Same Trend in Male vs Female?
Male
['USA'] 701
['GBR'] 415
['AUS'] 369
['ITA'] 363
['NED'] 331
['GER'] 330
['FRA'] 316
['CAN'] 274
['DEN'] 253
['NZL'] 220
Female
['USA'] 457
['CAN'] 257
['GBR'] 253
['CHN'] 232
['AUS'] 212
['GER'] 200
['NED'] 191
['ROU'] 183
['FRA'] 129
['RUS'] 126
Male
['ITA'] 81
['GBR'] 67
['FRA'] 57
['GER'] 41
['POL'] 40
['ESP'] 32
['USA'] 32
['NED'] 32
['AUS'] 30
['SUI'] 30
Female
['GER'] 38
['ROU'] 37
['AUS'] 25
['BUL'] 24
['FRA'] 21
['USA'] 19
['GBR'] 19
['GER' 'GDR'] 15
['NED'] 14
['ITA'] 13
Top Nations for Olympians with
Junior Team Experience
Top Nations for
Olympians
52. Junior Team Composition of Top Olympic 8+ Boats
Junior Exp.
Yes
No
Junior Exp.
Yes
No
%ofBoatwithJuniorExperience
53. U23 Team Composition of Top Olympic 8+ Boats
U23 Exp.
Yes
No
U23 Exp.
Yes
No
%ofBoatwithU23Experience
59. Next Steps
Analyzing retention of athletes from Junior and U23 development
Comparing success (medals) at various regattas to team composition
Improve the scraping process, store data in a database
Feature Engineering
Network Analysis
60. Race as a Symphony
“He suggested that Joe think of a well-rowed race as a symphony, and himself as
just one player in the orchestra. If one fellow in an orchestra was playing out of
tune, or playing at a different tempo, the whole piece would naturally be ruined.
That’s the way it was with rowing. What mattered more than how hard a man
rowed was how well everything he did in the boat harmonized with what the other
fellows were doing.”
― Daniel James Brown, The Boys in the Boat: Nine
Americans and Their Epic Quest for Gold at the 1936
Berlin Olympics
62. Weighted Synergy Graphs
“In this work, we are interested in a model of team performance that goes beyond
the sum of single-agent capabilities. We understand that there is synergy among
the agents in the team, where team performance at a particular task depends not
only on the individual agents’ capabilities, but also on the composition of the team
itself”
S. Liemhetcharat and M. Veloso. Weighted synergy
graphs for effective team formation with heterogeneous
ad hoc agents. Artificial Intelligence. 208. (2014). 41-
65.
http://dx.doi.org/10.1016/j.artint.2013.12.002
# Testable hypotheses:
# 1) We could expect that the male rowers predate female rowers
# 2) Female rowing received boon from Title IX (USA). So after taking affect in 1972 the number of female rowers would increase
# due to collegiate sports. This would translate to a starting birth year of ~1954
['Length of athlete table: ', 45343]['Tallest in set: ', 185186.0]['Shortest in set: ', 1.0]['Number of ridiculous heights (>3m): ', 87]['Number of ridiculous heights (<1m): ', 13]
Let’s clean this up by showing median age for each birthyear
U.S. fields the most Olympians (over the entire WR dataset), however U.S. is 7th for Olympians with Junior team experience
By the way, GDR was the country code for West Germany.
Looking at top 6 finishing 8+ boats (A final) for the past 7 Olympics