ABCDEFGHIJKLMNOPQRSTUVWXYZ
1
2
Quick Summary
3
Mens Teams
Top 5 Scorers Rating
RankAverage RatingRank Change
4
Laval7132114260
5
McMaster7095214199
6
Guelph700631401-1
7
8
Women's Teams
Top 5 Scorers Rating
RankAverage RatingRank Change
9
Laval6977113950
10
Guelph6894213791
11
STFX685931372-1
12
13
14
NameTeamRankNameTeam
15
16
Andrew DaviesMcMaster1Siona ChisholmSTFX
17
Max TurekMcMaster2Jessy LacourseLaval
18
Andrew PeverillSt Mary3Catherine BeaucheminLaval
19
Jean-Simon Desgagnes
Laval4Lauren McNeilUBC Okanagan
20
Thomas NobbsUBC5Kiana GibsonGUELPH
21
Jonathan TedeschiLaval6Jenna McFadyenSaskatchewan
22
Mitchell UbeneGuelph7Jade BerubeLaval
23
Felix Lapointe-PiloteLAVAL8Paige ChisholmPEI
24
Matthew BeaudetMCGILL9Sadie-Jane HicksonGUELPH
25
Alex DroverMcMaster10Joanna BrownUBC Okanagan
26
Jacob BenoitSTFX11Olivia CooperAlberta
27
David GirardinLAVAL12Murielle AglotMONTREAL
28
What is this and how does it work?
Nicholas BannonGuelph13
Chloe Fleurent-Gregoire
McGill
29
Pierre-Yves Normandin
LAVAL14Jade RobertsMemorial
30
This is a sheet that compiles all of the races of the 2021 Usports XC season and builds a rating system similar to what you would see used for chess players. Due to the long distances, potential travel restrictions and the variability of both XC courses, weather, distances, etc. it can sometimes be very difficult to track who is performing well. This sheet will try to remedy this problem by providing an up-to-date rating for each athlete competing this season. Ratings all start at 1000, and 1000 will likely be the ongoing average of the competitors. A difference of 800 between competitors signifies that there is effectively a 100% probability that the higher rated runner will beat the other. Ratings do take time to adjust, and especially in the early weeks the ratings should be seen as provisional at best. Ratings (R') are determined by four main metrics. First is your previous rating (R). Second is a measure of how far away you are from the average. I have decided to use a Z-Score to simulate this, and then pull a probability value from it (S). Third is based on how many races you have run (K) (over time, ratings will change less and less), and fourth is a comparison of your performance against your expected performance (E). As the season progresses the ratings will become more accurate. One of the major assumptions is that the average of each field in each region is roughly the same. This may not hold true in the long run but stay tuned and we will find out together!Updated Nov 6: Ratings are finalized! Congrats to everyone who has finished their season and good luck to all competitors racing in Quebec and Ottawa in the coming month! Laval is looking pretty dominant right now with their depth in both the men's and women's teams, however they are going to face some stiff competition, especially coming out of Guelph. I expect it to be a really solid race on both the men's and women's side for first in the individual titles. Jesse vs Kiana vs Joanna will be quite the show, while on the guys side Thomas throwing it down against Matthew and Andrew should be interesting. Anticipating more views in the coming days, reach out to me if you've enjoyed following this! Its hard to gauge engagement through a google sheet. Going forward I'm interested in putting together a small team to manage this in future years, if you are interested in being a part of this let me know! As always any and all feedback is appreciated! See you in Quebec!
31
32
33
34
35
36
37
38
39
R' = R + K(S - E)
40
41
Wikipedia:
42
https://en.wikipedia.org/wiki/Elo_rating_system
43
44
Trackie Results:
45
https://www.trackie.com/USports/XC/Results/
46
47
Trackie's Rankings (lol):
48
https://www.trackie.com/USports/XC/Rankings/
49
50
If you are reading this and see that results are missing, please send me the link to them via messenger or instagram!
51
-Seb Saville
52
53
54
ProbabilityDifference
Is this actually a good estimator and predictor?
55
1800From limited statistical analysis, yes! R values show the correlation of data, and after doing a comparison of expected Z-scores to actual Z-scores, the R value for men so far is about 0.78, and for women is about 0.89. These are both very high levels of correlation. Something interesting that has come up is that people with higher ratings tend to perform much closer to their predicted score, generally within 10 percentile points (often much less, average of about 4%) when over a rating of 1200. This can be seen pretty clearly on the two graphs below! All in all, really happy with how the results of this are turning out! Thanks for all of the interest that everyone is showing!
56
0.99677
57
0.9366
58
0.8240
59
0.7149
60
0.672
61
0.50
62
0.4−72
63
0.3−149
64
0.2−240
65
0.1−366
66
0.01−677
67
0−800
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100