2020 Season - Prediction Contest

Discussion in 'NWSL' started by cpthomas, Mar 6, 2020.

  1. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    It's about 6 weeks to the beginning of the season, so time for me to set up the programs for the 2020 Prediction Contest. Wow, time flies! Blissett is the returning regular season champion and thus the Shield bearer. The dreaded robot is the combined pre-season/regular season/playoffs champion.

    We had a little discussion after the dust settled about possible changes to the scoring system. I have a possible change for the regular season we might consider, based on the following table:

    upload_2020-3-5_23-19-20.png

    The table shows, of all the predictions last year, the % (and underlying number) that met each of the scoring categories. The highlighted numbers are a scoring system that approximates the percentages. In other word, the scoring system is an approximation of how hard it is to get each category right. The 4.0 - 1.7 - 0.8 - 0.6 possible scoring system compares to the 3 - 2 - 1 - 1 we've been using. Or, as an alternative that's a little simpler to remember and a little more aesthetic, the change could be to a 4 - 2 - 1 - 0.5. With these "simpler" points, regular season scores would go down about 15%, so the point spreads between competitors would shrink a little, but it shouldn't be by a lot. In terms of the proportionality among the pre-season, regular season, and playoffs prediction score spreads, I don't think it would make a big difference. Or, we could stay with what we've had for the last couple of years.

    Any thoughts about this or about other potential changes? SiberianThunderT, you're the expert on this, what do you think?
     

    Attached Files:

    CoachJon repped this.
  2. blissett

    blissett Member+

    Aug 20, 2011
    Italy
    Club:
    --other--
    Nat'l Team:
    --other--
    Hi. I don't have the time to ponder on the potential changes right now, but I just answered to acknoledge the creation of this thread and to subscibe to it to remain updated.

    As the shield bearer, I feel my personal opinion as a responsability :p: so I'll quietly think about it in the next days.
     
    CoachJon repped this.
  3. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    #3 cpthomas, Mar 6, 2020
    Last edited: Mar 6, 2020
    Scratch the earlier table, here's a revised table after some corrections:

    upload_2020-3-6_10-15-47.png

    In this table, the % of Correct Goals Scored for At Least One Team Predicted column includes both games in which the right number of goals was predicted for only one team and games in which the right number was predicted for both teams. In other words, roughly 50% of the time a predictor is going to get the score right for at least 1 team. The yellow highlighted numbers use that column's 50.2% as the base and represent the relative difficulty of the other predictions in relation to that. The bottom row rounds off the highlighted numbers to the nearest half.

    A simpler alternative would be a 5 - 2 - 1 - 1 score system. This would be exactly the same as the last few years, except it would increase the points for getting the exact score from 3 to 5, reflecting the difficulty of getting the exact right score. Applied to the 2019 predictions, the score spread for the regular season predictors would be in the same ball park as the combined score spread for the pre-season and playoff predictions. My inclination is we should go with this simpler alternative, to give a better reward for the much more difficult task of getting the exact score right. This would have resulted only in minor changes in the 2019 standings from the score system we've been using -- Blissett still would be the Shield bearer.
     
    CoachJon and blissett repped this.
  4. CoachJon

    CoachJon Member+

    Feb 1, 2006
    Rochester, NY
    Nat'l Team:
    United States
    Thanks for pondering and working to enhance the contest. For me, picking winners was my favorite part. My goals scored predictions were mostly pulled out of thin air. So having the WLD part have the least weight is a little disappointing. Regardless, I do think the 5-2-1-1 system is a good way to go.
     
  5. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    With a 5-2-1-1 system, the points earned in each category in 2019 would have been:

    Exact score correct (5 points per correct prediction): 575

    Goal differential prediction correct (2 points): 564

    WLD prediction correct (1 point): 566

    One team's exact score prediction correct (1 point): 639 (if changed from 1 point to 0.85, would be 543, which probably would be better)
     
  6. blissett

    blissett Member+

    Aug 20, 2011
    Italy
    Club:
    --other--
    Nat'l Team:
    --other--
    I am still pondering, but I am not sur if I like a so big gap between exact prediction and the rest. In the end, it is true that it's the most difficult prediction to nail, but it's also true that, most often than not, there probably is quite some luck involved in telling a 3-1 score apart from a 3-0 one.

    I guess exact score should be rewarded, but not in a way that punishes so harshly those who went close to it compared to who actually nailed it.

    This is independent from the data @cpthomasshowed about the rankings of last year not being so much affected by this change than one could figure; it's more of a "philosophical" objection: is there really a 3 points gap between nailing the score and nailing the goal difference? A draw is a draw, for instance: we reward 3 points more who could tell a 0-0 from a 1-1? Is there really a 4 points gap between a correct W/L/D prediction and an exact score?

    if it was me I'd go with something at least slightly tame, be even just a 4-2-1-1, although not completely justified by stats.
     
    CoachJon repped this.
  7. cpthomas

    cpthomas BigSoccer Supporter

    Portland Thorns
    United States
    Jan 10, 2008
    Portland, Oregon
    Nat'l Team:
    United States
    I can see an argument -- against 5 for the exact score -- being that getting the exact score isn't a matter of skill, it's simply a matter of luck.

    Another argument is that the current 3-2-1-1 as worked fine, so let's not change it.

    PS - I haven't made a final, final decision yet, but I think I'll be using Robot Version 2 for the coming season. Version 2 still will be based on historic results, but with a very different approach that could be better or could be worse. It will involve absolutely no human intervention, so strictly the Robot.
     
    CoachJon, SiberianThunderT and blissett repped this.

Share This Page