Results 1 to 12 of 12

Thread: Men's free skate score standard deviation

  1. #1
    Forum translator Ptichka's Avatar
    Join Date
    Jul 2003
    Location
    Boston, MA
    Posts
    4,430

    Men's free skate score standard deviation

    This might just be interesting for math geeks, but nere it is nonetheless. Standard deviation is basically how much a set of numbers diverges. This is relevant for FS scores because a small standard deviation means all judges pretty much agree whereas a large deviation means there is no agreement. For example, scores of 7.25,7.5,7.5, 7.75 have standard deviation of 0.2, whereas scores 6.5, 7.25, 7.75, 8.5 have s.d. of 0.84. Mathman, feel free to explain it better

    I thought it would be interesting to see standard deviations in the component scores of men free skate. Note that I took a standard deviation for all components and all judges, so I had a set of 40 numbers for each skater (5 components times 8 judges). In general, standard deviation grows as we move down the ranks, that makes sense. What I found interesting, though, was that by far the largest standard deviation of the night went to Plushenko (Lysacek's is among the smallest). Also found it interesting that Weir's deviation wasn't all that big at all (I thought it was a matter of just some judges not liking him - not so).

    So, here it is, in order from largest deviation to the smallest; number in parenthesis indicates placement for the free skate.

    1. Plushenko (2) - 0.8953
    2. Ten (14) - 0.8581
    3. Kovalevsky (24) - 0.7530
    4. Pfeifer (20) - 0.7426
    5. Bacchini (22) - 0.7150
    6. Lindemann (23) - 0.6920
    7. Borodulin (12) - 0.6880
    8. Chipeur (21) - 0.6810
    9. Verner (17) - 0.6752
    10. Joubert (16) - 0.6691
    11. van der Perren (18) - 0.6613
    12. Schultheiss (13) - 0.6579
    13. Contesti (19) - 0.6373
    14. Fernandez (10) - 0.5901
    15. Weir (6) - 0.5568
    16. Lambiel (3) - 0.5367
    17. Brezina (11) - 0.5282
    18. Kozuka (8) - 0.5171
    19. Oda (7) - 0.5144
    20. Abbott (9) - 0.5042
    21. Lysacek (1) - 0.4413
    22. Chan (4) - 0.4367
    23. Takahashi (5) - 0.42957
    24. Amodio (15) - 0.4291

  2. #2
    Custom Title skatemom1122's Avatar
    Join Date
    Dec 2009
    Posts
    504
    Quote Originally Posted by Ptichka View Post
    This might just be interesting for math geeks, but nere it is nonetheless. Standard deviation is basically how much a set of numbers diverges. This is relevant for FS scores because a small standard deviation means all judges pretty much agree whereas a large deviation means there is no agreement. For example, scores of 7.25,7.5,7.5, 7.75 have standard deviation of 0.2, whereas scores 6.5, 7.25, 7.75, 8.5 have s.d. of 0.84. Mathman, feel free to explain it better

    I thought it would be interesting to see standard deviations in the component scores of men free skate. Note that I took a standard deviation for all components and all judges, so I had a set of 40 numbers for each skater (5 components times 8 judges). In general, standard deviation grows as we move down the ranks, that makes sense. What I found interesting, though, was that by far the largest standard deviation of the night went to Plushenko (Lysacek's is among the smallest). Also found it interesting that Weir's deviation wasn't all that big at all (I thought it was a matter of just some judges not liking him - not so).

    So, here it is, in order from largest deviation to the smallest; number in parenthesis indicates placement for the free skate.

    1. Plushenko (2) - 0.8953
    2. Ten (14) - 0.8581
    3. Kovalevsky (24) - 0.7530
    4. Pfeifer (20) - 0.7426
    5. Bacchini (22) - 0.7150
    6. Lindemann (23) - 0.6920
    7. Borodulin (12) - 0.6880
    8. Chipeur (21) - 0.6810
    9. Verner (17) - 0.6752
    10. Joubert (16) - 0.6691
    11. van der Perren (18) - 0.6613
    12. Schultheiss (13) - 0.6579
    13. Contesti (19) - 0.6373
    14. Fernandez (10) - 0.5901
    15. Weir (6) - 0.5568
    16. Lambiel (3) - 0.5367
    17. Brezina (11) - 0.5282
    18. Kozuka (8) - 0.5171
    19. Oda (7) - 0.5144
    20. Abbott (9) - 0.5042
    21. Lysacek (1) - 0.4413
    22. Chan (4) - 0.4367
    23. Takahashi (5) - 0.42957
    24. Amodio (15) - 0.4291
    HA! Plushenko first....some love him, some hate! Happy to see most agreed with Lysacek, Weir, and Takahashi's placement.

  3. #3
    Custom Title
    Join Date
    Jan 2010
    Posts
    2,709
    Quote Originally Posted by Ptichka View Post
    This might just be interesting for math geeks, but nere it is nonetheless. Standard deviation is basically how much a set of numbers diverges. This is relevant for FS scores because a small standard deviation means all judges pretty much agree whereas a large deviation means there is no agreement. For example, scores of 7.25,7.5,7.5, 7.75 have standard deviation of 0.2, whereas scores 6.5, 7.25, 7.75, 8.5 have s.d. of 0.84. Mathman, feel free to explain it better

    I thought it would be interesting to see standard deviations in the component scores of men free skate. Note that I took a standard deviation for all components and all judges, so I had a set of 40 numbers for each skater (5 components times 8 judges). In general, standard deviation grows as we move down the ranks, that makes sense. What I found interesting, though, was that by far the largest standard deviation of the night went to Plushenko (Lysacek's is among the smallest). Also found it interesting that Weir's deviation wasn't all that big at all (I thought it was a matter of just some judges not liking him - not so).
    Or do an SD of SP, LP scores in the past few competition
    Plushenko is closer to 0.
    Lysacek is closer to 1. Lysacek magically got 90 in the SP and 167 in the LP. His SD deviated too much from normal score. Plushenko, on the other hand, doesn't.

    It's how you play with number and present them. So take with a grain of salt. =).

  4. #4
    On the Ice lilbit's Avatar
    Join Date
    Oct 2003
    Posts
    28
    I think the main reason for Plushenko's large standard deviation is not so much the differences among the judges but rather the differences from one component to the next, across all judges.

    For instance, both Plushanko's and Lysacek's marks were lowest in Transitions and highest in Performance/Execution. But the spread between the two was 1.50 for Plushenko (7.25 to 8.80 -- the Joe Inman effect ) and only 0.55 (7.95 to 8.50) for Lysacek.

    One could do a two-way analysis of variance (ANOVA) and split out the part of the standard deviation due to differences among the judges and the part due to differences among the components.

    Ptitcka, can you give a link to the scores for the individual judges? On the ISU and Olympic Games sites I could only find summaries.
    Last edited by lilbit; 02-19-2010 at 05:49 PM.

  5. #5
    Constable , Costume Police colleen o'neill's Avatar
    Join Date
    Nov 2006
    Posts
    1,797
    That's really illuminating. Interesting that Patrick is also among those with the least divergence..

  6. #6
    Custom Title
    Join Date
    Jan 2010
    Posts
    2,709
    Quote Originally Posted by colleen o'neill View Post
    That's really illuminating. Interesting that Patrick is also among those with the least divergence..
    Illuminated one thing: judges got a memo to be supremely generous to hometown boy.
    Patrick Chan scored a personal best for THAT LP. It would be shocking if his wasn't among those with the least divergence.

  7. #7
    Custom Title Mathman's Avatar
    Join Date
    Jun 2003
    Location
    Detroit, Michigan
    Posts
    28,354
    test

  8. #8
    Custom Title
    Join Date
    Mar 2004
    Location
    Canada
    Posts
    517
    Quote Originally Posted by Ptichka View Post
    This might just be interesting for math geeks, but nere it is nonetheless. Standard deviation is basically how much a set of numbers diverges. This is relevant for FS scores because a small standard deviation means all judges pretty much agree whereas a large deviation means there is no agreement. For example, scores of 7.25,7.5,7.5, 7.75 have standard deviation of 0.2, whereas scores 6.5, 7.25, 7.75, 8.5 have s.d. of 0.84. Mathman, feel free to explain it better

    I thought it would be interesting to see standard deviations in the component scores of men free skate. Note that I took a standard deviation for all components and all judges, so I had a set of 40 numbers for each skater (5 components times 8 judges). In general, standard deviation grows as we move down the ranks, that makes sense. What I found interesting, though, was that by far the largest standard deviation of the night went to Plushenko (Lysacek's is among the smallest). Also found it interesting that Weir's deviation wasn't all that big at all (I thought it was a matter of just some judges not liking him - not so).

    So, here it is, in order from largest deviation to the smallest; number in parenthesis indicates placement for the free skate.

    1. Plushenko (2) - 0.8953
    2. Ten (14) - 0.8581
    3. Kovalevsky (24) - 0.7530
    4. Pfeifer (20) - 0.7426
    5. Bacchini (22) - 0.7150
    6. Lindemann (23) - 0.6920
    7. Borodulin (12) - 0.6880
    8. Chipeur (21) - 0.6810
    9. Verner (17) - 0.6752
    10. Joubert (16) - 0.6691
    11. van der Perren (18) - 0.6613
    12. Schultheiss (13) - 0.6579
    13. Contesti (19) - 0.6373
    14. Fernandez (10) - 0.5901
    15. Weir (6) - 0.5568
    16. Lambiel (3) - 0.5367
    17. Brezina (11) - 0.5282
    18. Kozuka (8) - 0.5171
    19. Oda (7) - 0.5144
    20. Abbott (9) - 0.5042
    21. Lysacek (1) - 0.4413
    22. Chan (4) - 0.4367
    23. Takahashi (5) - 0.42957
    24. Amodio (15) - 0.4291
    Ptichka, first thank you for the effort. I think there is flaw in this analysis. Your analysis is based on an assumption that a skater is even in all 5 elements of the PCS, so you suppose that they should be all in line with each other. But it is not the case. For example, Plushenko's transition is much lower than the other four, and performance was high compared to the other three if I remember it correctly. That might explain why he's among the highest in SD. So it might make more sense to do it by each element (SS, TR, ect). In my opinion it is right to have different means for different element as ,for example, a skater could get different choreography/composition and transition scores according different program, but his SS and the other should not change that much. Just a suggestion.
    Last edited by hongligl; 02-19-2010 at 06:46 PM.

  9. #9
    Constable , Costume Police colleen o'neill's Avatar
    Join Date
    Nov 2006
    Posts
    1,797
    Memo - schmemo. Apparently not all of them got the memo to give the returning hero the gold, no matter what..I guess any memos about Patrick have priority then... and so early in his career.

    the lack of rationale , not to mention the bile , is too much for me. Enjoy.
    Last edited by colleen o'neill; 02-19-2010 at 06:37 PM.

  10. #10
    Custom Title
    Join Date
    Jan 2010
    Posts
    2,709
    Quote Originally Posted by colleen o'neill View Post
    the lack of rationale , not to mention the bile , is too much for me.
    Lack of rationale? As in personal best for that LP. Enjoy the best program ever by Patrick Chan, according to the judges.

  11. #11
    Forum translator Ptichka's Avatar
    Join Date
    Jul 2003
    Location
    Boston, MA
    Posts
    4,430
    Ok, here is Plushenko's deviation broken into components:
    Skating skills - 0.6627
    Transitions - 1.1164
    Performance - 0.5559
    Choreography - 0.6744
    Interpretation - 0.8047

  12. #12
    Forum translator Ptichka's Avatar
    Join Date
    Jul 2003
    Location
    Boston, MA
    Posts
    4,430
    In fact, I now did a spread for all components AND averages for GoE deviations. Table here: http://ptichkafs.livejournal.com/46416.html
    Not only is Plush's component standard deviation very high (though calculated this way, it's not THE highest - that honor goes to Ten), but the average of standard deviations for each of the GoE's is the highest at 0.7560 (though Takahashi's at 0.7522 is about the same) - the norm is far lower.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •