In case any of you missed it, last week I introduced SyntheticCoach, a way of taking what is publicly known about the SuperCoach scoring formula, and what relevant data is publicly available, to try to model what a player’s ‘correct’ SuperCoach score should be. With thanks to Jamhead in last week’s comments, the formula now includes spoils, smothers, and other one-percenters, as well as goal-assists. The previous week, BBQbarbs suggested that running bounces are worth 1 point each; it would not surprise me in the slightest if this were the case, but as yet, I have not found a source to back this up (to be fair, I haven’t looked very hard – I’ve had some other things going on this week). If I can confirm that this stat is also part of the formula, and that it is worth 1 point apiece, it will be added to the system by this time next week. Lest anybody think that the formula is flipping and flopping every week, rest assured that it will stabilize in the coming weeks, as I nail down the final details of what we currently know (and are able to model) of the scoring system.
There has been a little bit of confusion in the comments over the last couple of weeks over how Champion Data scores kicks, so I thought I would clear that up here. As far as the SuperCoach formula is concerned, it’s quite simple: effective kicks are worth 4 points, ineffective kicks 0, and clanger kicks -4. The confusion comes from the way in which CD defines an effective kick:
A kick of more than 40 metres to a 50/50 contest or better for the team or a kick of less than 40 metres that results in the intended target retaining possession. [emphasis added]
I hope that’s cleared that up.
To start us off this week, some were curious on the weekend about Tom Rockliff’s relatively low score of 97, especially given his Dreamteam score of 150, on the back of 36 touches. SyntheticCoach gave him 118 (un-scaled: 133), so, at face value, those complaining may have a legitimate point. However, other commenters pointed out that he did much of his good work in the second half, when the game was already won. This highlights a blindspot in the SyntheticCoach system: it currently has no way of measuring scaling done in favour of actions taken when the game was on the line. To do this, it would need to know not only when the game was on the line – which it could easily be taught, as we have access to the score worm data – but also when each stat was recoded – data to which we do not currently have access. Long-term, this is something I would like to fix, but it would require logging of statistical data at regular intervals in-game, which would require significant work to set up. My intention is to do this eventually, but for the time being, there are much more practical – and much more easily implemented – improvements to be made to the system.
Moving on, a number of people were wondering about Brodie Grundy’s high score of 145, compared to James Sicily’s 141, when the latter, to most eyes, was a clear BOG. As it turns out, as far as SyntheticCoach is concerned, they were both robbed! It awarded Grundy 156 (173), and Sicily 165 (182).
Meanwhile, in his always-excellent Round Review, Dane wanted to know if Sydney Stack (106) had been given somewhat favourable treatment, particularly when compared with teammates Bachar Houli (87) and Brandon Ellis (80). Computer says no…
…t really. SyntheticCoach awarded Stack 101 (113), whilst Houli was given 84 (94), and Ellis 90 (100). The only discrepancy that really stands out there is Ellis, and 10 points across a whole game is nothing that cannot potentially be explained by scaling, or perhaps some stat of which my system is ignorant.
Dane also wanted to know how Liam Jones got to 47 from 5 disposals at 40 percent (including a truly woeful 25 percent kicking efficiency), plus 4 clangers. SyntheticCoach gave him 41 (43), so his real score probably wasn’t too far off the mark. So how did he get there? In raw, un-scaled terms: 12 spoils (24 points), 2 intercept marks, of which 1 was contested (12 points), 3 ground-ball gets (13.5 points), 1 effective kick, 1 effective handball, and 1 tackle (9.5 points), held back by 4 clangers (-16 points), for a total of 43 points.
Finally, and somewhat amusingly, in the same match, Champion Data awarded Marty Hore 3 points, whilst SyntheticCoach, in both its raw and scaled forms, gave him a big, fat, 0. Perhaps CD was feeling generous? Either way, as an owner, I’ll gladly take the extra 3 points!
Were there any scores that seemed off to you on the weekend? Or are you interested in a full stat-by-stat score breakdown for a particular player? Let me know in the comments below, and I’ll let you know what SyntheticCoach has to say about it.
Leave a comment / Scroll to bottom