Playing with Score Effects

Discussion in 'By The Numbers' started by Doctor No, Jan 9, 2019.

  1. Doctor No

    Doctor No Registered User

    Joined:
    Oct 26, 2005
    Messages:
    7,959
    Likes Received:
    1,371
    Trophy Points:
    139
    Home Page:
    An underlying hypothesis of score effects is that when a team is trailing, they take more shots on net in a deliberate trade between (increasing) quantity and (decreasing) quality.

    A consequence is that winning goaltenders will (a) face more shots on average, and (b) have a higher save percentage.

    This is a thread where I'll put interesting (to me) tables as I develop adjustments for my goalie data models.
     
    Day Man likes this.
  2. Doctor No

    Doctor No Registered User

    Joined:
    Oct 26, 2005
    Messages:
    7,959
    Likes Received:
    1,371
    Trophy Points:
    139
    Home Page:
    2017-18 NHL regular season, since it's complete. Shots and saves by period (just to get a baseline).

    We all understand that save percentages are lower in overtime (significantly). They are also lower (with more shots faced) in the second period, where the change is longer for both teams.

    SVSH
    1st 24,283 26,440 0.918
    2nd 25,102 27,678 0.907
    3rd 23,287 25,459 0.915
    OT 1,094 1,287 0.850
     
    Day Man likes this.
  3. Doctor No

    Doctor No Registered User

    Joined:
    Oct 26, 2005
    Messages:
    7,959
    Likes Received:
    1,371
    Trophy Points:
    139
    Home Page:
    2017-18 NHL regular season data, split by the score at first intermission. "I1D" reflects the goaltender's team position at the first intermission - for instance, "2" means that the goaltender's team is ahead by two at the intermission. Shots and saves in the table below reflect second-period data (so the period immediately following the intermission).

    I1DInstancesSVSH AV SH SV%
    -5 2 15 18 9.0 0.833
    -4 11 93 103 9.4 0.903
    -3 57 533 587 10.3 0.908
    -2 209 1,890 2,129 10.2 0.888
    -1 530 5,031 5,560 10.5 0.905
    0 924 8,978 9,897 10.7 0.907
    1 530 5,625 6,153 11.6 0.914
    2 209 2,164 2,388 11.4 0.906
    3 57 631 689 12.1 0.916
    4 11 115 125 11.4 0.920
    5 2 27 29 14.5 0.931
    I would say that the values between -2 and 2 are somewhat credible. There does appear to be this effect after 20 minutes, if not a large one.
     
    Day Man likes this.
  4. Doctor No

    Doctor No Registered User

    Joined:
    Oct 26, 2005
    Messages:
    7,959
    Likes Received:
    1,371
    Trophy Points:
    139
    Home Page:
    2017-18 NHL regular season data, split by the score at second intermission. Shots and saves below reflect third-period data (similar to the above).

    I2DInstancesSVSH AV SH SV%
    -7 1 7 9 9.0 0.778
    -6 4 22 27 6.8 0.815
    -5 24 199 213 8.9 0.934
    -4 62 520 576 9.3 0.903
    -3 140 1,034 1,153 8.2 0.897
    -2 302 2,332 2,578 8.5 0.905
    -1 473 3,820 4,162 8.8 0.918
    0 530 4,810 5,270 9.9 0.913
    1 473 4,929 5,353 11.3 0.921
    2 302 3,240 3,534 11.7 0.917
    3 140 1,494 1,633 11.7 0.915
    4 62 611 659 10.6 0.927
    5 24 237 256 10.7 0.926
    6 4 29 33 8.3 0.879
    7 1 3 3 3.0 1.000

    There does appear to be a similar (and larger) effect here. -1 doesn't seem to fit the overall pattern with respect to SV%. At least for the credible range, shots/period look to follow a clear pattern.
     
    Day Man likes this.
  5. Doctor No

    Doctor No Registered User

    Joined:
    Oct 26, 2005
    Messages:
    7,959
    Likes Received:
    1,371
    Trophy Points:
    139
    Home Page:
    One to-do is for me to adjust for team quality - for instance, stronger teams generally have higher shooting percentages (so opposing goaltenders generally have lower save percentages), and stronger teams are leading at the intermission more often than not.

    Empty net goals are excluded from the above.
     
    Day Man likes this.
  6. morehockeystats

    morehockeystats Unusual hockey stats

    Joined:
    Dec 13, 2016
    Messages:
    364
    Likes Received:
    62
    Trophy Points:
    46
    Occupation:
    sysadmin
    Location:
    San Jose
    Home Page:
    Did you see the adjacent thread where I posted the correlation graph between the Saves and SVP?
     
  7. Doctor No

    Doctor No Registered User

    Joined:
    Oct 26, 2005
    Messages:
    7,959
    Likes Received:
    1,371
    Trophy Points:
    139
    Home Page:
    Yes - it's interesting and matches loosely the numbers I've posted previously (once adjusting for opponent strength).

    Score effects is one of the reasons and this thread is an attempt to quantify that.
     
    Last edited: Jan 10, 2019

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice
monitoring_string = "358c248ada348a047a4b9bb27a146148"