Tuesday, 6 March 2018

AFLW 2 - ONE OUT, WHO IS NEXT

Not many games to go now, and the weekend's results sees;
that it is mathematically impossible for Collingwood to make the finals.
that with a percentage of 50.5, Carlton's season is all but over (that percentage along with needing losses to all their challengers makes it a very very steep hill climb).
that next week we have an 'elimination final' between Adelaide and Fremantle in Darwin. Its 5th vs 6th with 2 wins each, and Freo a draw as well.


Results:
Once again three of the four tipped correct. The elusive 4/4 evades again.


PREDICTIONRESULTRATINGS CHANGE
4pts6pts+1-2
9pts18pts-13+12
11pts35pts+10-32
10pts8pts+4-1



Current Rankings:
The below rankings are up to date and are an insight into the finals contenders. Dogs and Lions on top, Crows and Mons there or there abouts.

Then a gap back to the other teams that have tentative claims. An exception would be GWS, who are undefeated over the last three rounds and making a finals charge.


RANKINGSPre:R 6
TEAMPTSΔPTSΔRK
1W Bulldogs/FFC1067-1-
2Brisbane Lions1057-1-
3Adelaide103610+1
4Melbourne10301-1
5Collingwood9764+1
6Carlton953-32-1
7GWS92713+1
8Fremantle924-12-1



Forecasting:
All the next week's predictions below and it looks a bit lop-sided.
Adelaide in Darwin get the nod with no HGA but a high rating as defending (W) premier.

HOMEAWAYVENUEPREDICTION
AdelaidevFremantleMarrara12pts
Brisbane LionsvCollingwoodMoreton15pts
GWSvW Bulldogs/FFCManuka14pts
CarltonvMelbournePrinces8pts

Brisbane should win against the Magpies given the relative strength of teams and necessities to win.
Melbourne have to win to maintain a top two shot.

This week will shape the finals, big time.



Over in the season predictions, each teams expected finishing window is getting tighter.


FMI RANKING-BASED PREDICTIONS (2018)
LADDER POSITION12345678
1W Bulldogs/FFC97121
2Brisbane Lions16618591
3Melbourne1144426104
4Adelaide16223115772
5GWS44161838181
6Fremantle51526221419
7Carlton6514222528
8Collingwood2863549

The Bulldogs look good for a finals berth, with Brisbane the other most likely to play in the Grand Final. Melbourne and Adelaide also stake reasonable claims to that date as well.


Which begs an interesting question... where would a Bulldogs-hosted AFLW Grand Final be played in this new-found finals allocation meritocracy?

2 comments:

  1. Your methodology is mysterious to me. For example, I'm looking at your "rankings based predictions", which you ran more or less one hundred times to get a probability spread (which is wonderful - good science!). But then we see 99 total outcomes when you add the first place column...101 in the second place column...99 in each of the 7th and 8th place columns...99 for Carlton, 100 for GWS, and 101 for the Bulldogs! What gives?

    I also run a ratings-based system for both AFL and AFLW (as well as American and Canadian football, etc.). It's a pure ELO-based system, where a team which improved upon its forecast gains the same number of points as its opponent lost. For example, last week's games mandated the following changes in my ratings: Melbourne up 1.0, Brisbane down 1.0; GWS up 3.2, Freo down 3.2; Adelaide up 2.4, Carlton down 2.4; and Collingwood up 0.3, Bulldogs down 0.3. (Fifty is our norm, rather than 1000, not that it matters.) I'm not sure how you justify Adelaide gaining ten points but Carlton losing 32. I'd be very curious to see the mathematics involved, although I suppose that's not something of general interest to your readers.

    I really appreciate the 100-ist prediction ladder in particular: that's something my system doesn't easily produce for my readers. Thanks for all you do on your site!

    ReplyDelete
  2. Hey thanks for reading! Let try to answer your questions.

    "But then we see 99 total outcomes when you add the first place column...101 in the second place column...99 in each of the 7th and 8th place columns...99 for Carlton, 100 for GWS, and 101 for the Bulldogs! What gives?"
    I am a simple man who likes the look of a whole number in the table. So in the above instances, numbers are rounded and dropped (or boosted). The raw data will have a team's chance of finishing in a position as 0.15%. I will delete that point for the table because 'white space' looks better in the chart than '<1%'.

    "Collingwood up 0.3, Bulldogs down 0.3. ... I'm not sure how you justify Adelaide gaining ten points but Carlton losing 32."
    Yes, I can see how this is off-putting. What I have allowed for is a 'form' characteristic. So if a team performs well, then an additional ratings shift is added, and built on week on week. A loss / under-performance in a good run of wins reduces that bonus.
    This is also applied for team performing badly... the longer the streak goes on, the bigger the penalties applied week on week. In a way it provides a
    - smoothing to a team that is going well but has one bad game (/going really badly and gets an upset win), and
    - quicker reaction / shift in ratings for teams that hit a purple patch of form.

    Hope that makes sense. Good luck with your modelling Is it public? Send me a link - there are a few of us that run ELO models who would be interested to see your system in operation.

    ReplyDelete