No changes to the system this week, so straight into the weekend’s scores…
To start us off, Dane wanted to know on Monday whether Todd Goldstein deserved even more than his 176 on the weekend.
CD: 176
SyntheticCoach: 200 (219)
Todd Goldstein (North Melbourne)
• Effective kicks: 11 => 44
• Effective handballs: 19 => 28.5
• Clangers: 2 => -8
• • of which frees-against: 1
• Ground-ball gets: 15 => 67.5
• Bounces: 0 => 0
• Goals: 1 => 8
• Behinds: 0 => 0
• Marks: 3 => 6
• • of which contested: 0 => 0
• • of which intercepts: 0 => 0
• Tackles: 1 => 4
• Frees-for: 3 => 12
• Hitouts-to-advantage: 9 => 45
• Goal assists: 2 => 7.0
• Spoils: 1 => 2
• Other one-percenters: 2 => 3.0
Total: 219
Blindspots:
• Gathers from hitouts are worth 2 points each. Player had 9 clearances – perhaps some came from that source?
• Player had 14 uncontested possessions that did not come from a mark. Perhaps some came from handball-recieves (1.5 points each)?
• SyntheticCoach knows nothing about when the game was on the line, let alone who did what when it was. This can have a massive effect on scaling.
In short: yes, he probably did.
According to SyntheticCoach, it wasn’t just Goldstein who was short-changed in this game: Ben Brown (CD: 136), SyntheticCoach 155 (169), Ben Cunnington 129, 141 (154.5), Jack Ziebell 125, 140 (153), and Shaun Higgins 121, 134 (146.5) were also scored less than what they should have been, according to my system.
Paddy Ryder, too, was given 126 (137.5) by SyntheticCoach, compared with 86 by CD. He was in a team that was badly beaten, however.
Moving on, the system also thought that Jack Viney’s effort (CD 113, SyntheticCoach 132 (145)) was under-appreciated, although given how bad Melbourne was on Friday night, the scaling was probably weighted heavily against them.
From my observations, sitting up in the stands, the 3 votes in Saturday’s Carlton vs St. Kilda game ought to have gone to Matthew Kreuzer, with the 2 votes going to Levi Casboult. SyntheticCoach broadly agrees with this perspective, although Champion Data took a slightly different view, awarding Casboult 108 points, compared to SyntheticCoach’s 128 (152).
Finally, Hawthorns Isaac Smith was awarded just 78 (92) by SyntheticCoach, compared with 111 from CD. The breakdown is somewhat helpful here:
Isaac Smith (Hawthorn)
• Effective kicks: 9 => 36
• Effective handballs: 11 => 16.5
• Clangers: 1 => -4
• • of which frees-against: 1
• Ground-ball gets: 4 => 18.0
• Bounces: 0 => 0
• Goals: 1 => 8
• Behinds: 0 => 0
• Marks: 6 => 12
• • of which contested: 0 => 0
• • of which intercepts: 0 => 0
• Tackles: 1 => 4
• Frees-for: 0 => 0
• Hitouts-to-advantage: 0 => 0
• Goal assists: 0 => 0.0
• Spoils: 0 => 0
• Other one-percenters: 1 => 1.5
Total: 92
Blindspots:
• Player had 14 uncontested possessions that did not come from a mark. Perhaps some came from handball-recieves (1.5 points each)?
• SyntheticCoach knows nothing about when the game was on the line, let alone who did what when it was. This can have a massive effect on scaling.
Throwing in the possible handball-receives gives Smith a potential extra 21 points. That still doesn’t quite get him to 111 points, but if we factor in the pro-winning-team scaling that likely came from this game being a complete blowout, his score probably isn’t too far off.
Were there any scores that seemed off to you on the weekend? Let me know in the comments, and I’ll let you know what SyntheticCoach has to say about it!
Leave a comment / Scroll to bottom
You’ve probably already mentioned this Salamander, but does the scaling back to 3300 points per game (or whatever that figure is) account for a lot of the sub par scores?
Yes and no. SyntheticCoach scales all the scores in each game so that they sum to 3300 (the scores in brackets are the pre-scaling scores); unlike Champion Data’s scores, however, SyntheticCoach’s scaling is linear. That is, it divides 3300 by the sum of the raw scores, and multiplies the raw scores by the result of this, before rounding. Champion Data, on the other hand, does their scaling in much more complicated fashion, based on who did what and when, as well as a whole bunch of other factors to which we mere mortals are not privy.
Getting back to your original question, I would say that a lot of the gaps between the real and synthetic scores are due to differences in scaling; in future, I intend to program the system to point out more potential causes of this. Another factor is that there are a couple of stats that are known to be part of the scoring formula, but we can’t currently measure: gathers from hitouts (2 points), and handball receives (1.5 points). Believe me, it drives me nuts that I can’t include these in the SyntheticCoach formula; for the time being, the breakdown functionality spits out a warning if a player racks up a lot of a publicly available stat that is likely to be correlated with one of these (clearances for gathers from hitouts, un-marked uncontested possessions for handball receives). Of course, by being absent from the raw scores, these in turn distort the scaled scores. In practice, however, the overwhelming majority of SuperCoach scores line up very, very closely with their respective SyntheticCoach scores, so it’s probably not as big of an issue as I am making it sound like.
Salamander why don’t you value all kicks and handballs the same? Effective ones are worth mire
According to the official information put out by CD, effective kicks are worth 4, and effective handballs 1.5 points, so those are the numbers I go with; likewise, as per the formula put out by CD, SyntheticCoach doesn’t hand out points for ineffective disposals.
Sorry I meant contested are worth more. Therefore why do you value all effective kicks and handballs the same when their contested versions are worth substantially more.
Ah, I see what you’re getting at. It’s not that the disposal is worth any more – the initial possession is. Ground-ball gets and contested marks cover that side of the equation. Gathers-from-hitouts are also worth 2 points, but we don’t have access to that stat (although if a player wins a lot of clearances, chances are that some will have been gathered from a hitout, so a player racking up a lot of clearances triggers a blindspot warning in the system).
Just like Cow Talk is associated with Father Dougal, Scoring Anomalies is all Salamander. Gone above and beyond with your analysis, Jack! Always an interesting read……
Thanks, Schwarz!