Sunday, February 17, 2019

AFL WOMEN’S Power Rankings: Round 3 2019

This year’s top-ranked AFLW team so far, the Kangaroos, are pulling down the marks close to their goal line.

After three rounds of the 2019 AFLW season new team the Kangaroos are the best-performed team, with three wins and a percentage of 250.8. They have won their first three matches by four goals or more, including an impressive 31-point win against reigning premiers the Western Bulldogs on Friday night.

It may seem like the Roos are winning out all over the field, and to an extent they are (they’ve kept their opponents to an average of 20 points). However they only rank fourth in terms of getting the ball inside 50 metres – with just one more than Carlton, who they actually lost the inside 50 count to in their first match.

Where the Kangaroos have outranked teams is the rate at which they score once they get the ball inside 50 metres. They have scored 153 points from 86 inside 50 entries, or 1.77 points per inside 50. That rate is better than any men’s team from last season (though the West Coast Eagles when they had Josh Kennedy and Jack Darling playing was kind of an exception).

The Roos’ forward efficiency has been helped by their prowess at taking marks inside 50. They have taken 26 marks inside 50 from their 86 entries, at a ratio of about 30 per cent. The most potent AFL men’s offence in 2018 – Melbourne’s – had a ratio of less than 25 per cent.

What also matters is where the Kangaroos have been getting those marks inside their forward zone. In this year’s Champion Data AFLW Prospectus, Chloe McMillan noted that the two AFLW premiers so far tended to get more of their scores from kicks close to goal (i.e. less than 15 metres out). A kick for goal from 30 to 40 metres out isn’t really ideal for the kicking range of some AFLW players. Watch the highlights from Friday night’s game, and you’ll see that Courteney Munn didn’t have to roam far from the goal square for any of her four goals, nor generally did the other North players.

The AFLW Kangaroos are really doing what every football team hopes to do – mark the ball deep inside 50 and get an easy set shot close to goal. Next week’s match against Melbourne – who despatched the other of last year’s grand finalists, Brisbane, by almost 40 points on the weekend – could be a belter.

Monday, February 11, 2019

AFL WOMEN’S Power Rankings: Round 2 2019

The rankings agree – the new AFLW conferences are about as lop-sided as most people think they are.

This year, somewhat controversially, the AFL has split the AFLW teams into two conferences. Ostensibly this is because, following the introduction of two new teams, it is no longer the case that all of the teams will play each other during the course of the season (a decision that was itself controversial). Critics may of course note that all of the teams in the men’s AFL do not play each other the same amount of times, and conferences have not been introduced for that (yet).

Even before the season started some people noted that one conference – Conference A – looked to be stronger than the other. First, all of the odd-placed teams from last year (first, third, etc.) were placed in one conference, and all of the even-placed teams were placed in the other. Second, the stronger of the two new teams – North Melbourne/the Kangaroos – was placed in Conference A. Third, two of the Conference B teams – Brisbane, and particularly Collingwood – were significantly weakened by losing players to the Roos.

After the first two rounds, including the first round of ‘cross-conference’ matches, there is now a spate of commentary about how uneven the conferences appear to be. As noted in The Age:

“Melbourne sit bottom of Conference A with one win and a percentage of 119.1. The Brisbane Lions, with one win but a vastly inferior percentage of 73.4, are atop Conference B.”

And since the top two teams in each conference will play off in the finals:

“It doesn’t take too deep an inspection to see how this is an issue, with an increasing likelihood that mediocre Conference B teams will take finals spots ahead of better Conference A sides.

Are the two conferences really as uneven as they seem?

It’s probably not just the fixture

Particularly early in the season, the fixture can somewhat mask the ‘true’ strength of teams. Results obviously in part depend upon who you play and where you play them.

It’s not hard to see that it’s probably not just the fixture causing the disparity so far though. Every team in Conference A is well ahead of every team in Conference B following the first round of ‘cross-conference’ matches, and home ground advantage seems unlikely to be large enough to explain the difference. The five Conference A teams beat the five Conference B teams by an average of 20 points on the weekend, and none won by less than 13 points.

The rankings put 46 per cent of their weight on the results of the most recent two matches. Accordingly, Conference A sides now occupy five of the top seven positions in the rankings. Further, the two Conference B sides in the top seven – Brisbane and Collingwood – who as mentioned above were weakened by player movements, are falling faster than anyone.

It may well be that a significant test of the relative strength of the conferences comes in the final cross-conference match of this weekend. If Melbourne, the last-placed team in Conference A, can win away against Brisbane, the first-placed team in Conference B, then the comments that the sixth and seventh ‘best’ teams could get into the finals may gain even more momentum.

Sunday, February 3, 2019

AFL WOMEN’S Power Rankings: Round 1 2019

The Kangaroos were impressive in their first match against Carlton. Their big win however may be more about the weakness of the Blues than the strength of one of the AFLW premiership favourites.

Coming into this AFLW season, new club North Melbourne/Tasmania/the Kangaroos were considered to be one of the league’s most formidable teams, on the back of their strong recruiting from other clubs.

They certainly played that way in their first match against Carlton, beating the Blues by 36 points. Many of their top-line recruits – including Emma Kearney and Jenna Bruton from the Western Bulldogs, and the ex-Collingwood quartet of Emma King, Jasmine Garner, Jess Duffin, and Moana Hope – met expectations in their first outing for their new club. Case closed then, right? After that, can anyone stop the Kangaroos this year?

Early last season a six goal win against the Blues – or any team – would certainly be a signal of a strong AFLW side. Indeed, the 12-goal thrashing of Carlton last season by eventual premiers the Bulldogs was possibly the first major sign they were one of the top flag contenders.

The thing is now though is that almost everyone has easily beaten Carlton recently. Carlton has lost its past six matches by an average of 35 points – about the same as North’s winning margin on the weekend – in a league where the average margin has tended to be about half of that amount. Even after adjusting for opposition strength and home ground advantage (see chart below), Carlton’s average adjusted net margin over that span is -29 points. Compared with a league average score of around 30 points the Blues have had 40 points or more scored against them in all of those matches, and have only once got past 22 points themselves.

This is not to take away too much from the Kangaroos’ win, or to deny that they seem to be one of the teams with the strongest chance of winning the premiership. It’s just to say that – unfortunately for the Blues’ players – what the Kangaroos did to Carlton on the weekend has recently been within the reach of most AFLW teams.

The rankings’ tips for Round 2 are below. The prediction for the Collingwood v Melbourne match should be treated with a great deal of caution. As mentioned before on this blog Collingwood finished last season really well, but they lost several of their important players between seasons, particularly to North Melbourne. Their one-point loss to Geelong on the weekend is not in itself enough to drastically change their ranking, but I would be hugely surprised if they are still so highly rated after the next couple of weeks.

Thursday, January 24, 2019

The Wooden Finger Is Finally Getting Into the AFL Tipping Business

After almost a decade of my AFL Power Rankings, I’m caving in. (Or I’m expanding my horizons, if you want to look at it that way.)

I created my Power Rankings simply so I could get a better indication of each team’s strength than the ladder was showing me. I’ve never been that interested in using it to publically tip the results of matches. (Privately, I’ve done so for several years in various office football tipping competitions – not that I ever expect it to win the competition.)

But seemingly EVERY other ‘rankings system’ does tips. And it is interesting to see them. So yes, starting this year the Wooden Finger is getting into the AFL tipping ‘business’.

How does it work?

To get the predicted net margin of the home team:
To get the expected probability that the home teams wins, I am using a standard deviation of 36 points or six goals for the men’s league, and 18 points or three goals for the women’s league. Maybe it should be a bit more, but since I do all of my HGAs in terms of goals, and for my purposes it doesn’t make much difference, I’m going with those.

AFLW Round 1 predictions

Let’s start with my predictions for Round 1 of the new AFLW season, which is kind of nice in that there are less ranking systems that do AFLW tips.

Two notes of caution though, related to the entry of two new teams into the AFLW competition this year. First, the ranking points for the two new teams – Geelong and North Melbourne – are of course not based on any previous matches, but are sort of based on their premiership odds. Second, the entry of those teams has resulted in a fair amount of player movement, particularly from Collingwood to North Melbourne. It may take a couple of weeks for the ranking points to adjust to the ‘new’ strength of each playing group.

So there you go. My heart still lies in ‘explaining’ performance rather than predicting it. Nevertheless I hope these tips add another interesting and informative dimension to the rankings each week.

Tuesday, January 22, 2019

Which AFL Club Has The Easiest Fixture in 2019?

When the 2019 AFL fixture was released Champion Data tweeted this table, ranking each team’s fixture from hardest to easiest.

Champion Data rated Collingwood as having among the hardest fixtures, and St. Kilda as having among the easiest. So did Rohan Connolly. So did HPN Footy. So did FMI. Spoiler alert: I’m going to as well.

Note the shadings in the table above though. Those shadings indicate how Champion Data ranked the strength of each team – e.g. the ‘strongest’ teams are West Coast, Richmond, Melbourne, and Geelong. Champion Data copped some criticism in the comments under their tweet about those rankings. Oh boy, did some people get annoyed about that.

In particular, there was criticism that Geelong – who finished eighth in 2018 – was ranked in the top group, while runners-up Collingwood were not. Some people were incredulous that a team that was in the last finals spot and then eliminated in the first week could be ranked above another team that made the Grand Final. Yes, it may sound a little strange, but on this one my rankings agree with Champion Data.

Few would dispute that Collingwood had a better finals series than Geelong last year. Over the season as a whole though I’d say that Geelong was better, or at least they were similar. In the home and away season Geelong scored about as many points as Collingwood, and conceded about 150 points less. The Cats won two less matches, but had a much tougher fixture than the Magpies. Collingwood played lower teams Brisbane, Carlton, and Fremantle twice, while Geelong’s only ‘easy double-up’ was Gold Coast. Switch those fixtures around, and the Cats probably finish higher going into the finals.

One point of all this is that, while I don’t know exactly how Champion Data come up with fixture assessment, I expect it is very similar to how I do it. My ‘groupings’ would also be similar, with the exception that I would have Essendon rather than North Melbourne ranked in the top eight after its strong finish to 2018.

The other point is that, while the fixture won’t in itself make you a ‘good’ team, it can sometimes make a fair difference to where you finish on the ladder.

Rating each team’s fixture in 2019

My method for rating the fixtures is to add up the ranking points of all the team’s opponents over the season, while adjusting for home ground advantage. This means that the fixture rating is determined by:
  • which five teams the team plays twice, as well as that a team plays every team at least once except itself; and
  • net home ground advantage over the season.
As Champion Data’s table implies, it is the ‘double-ups’ that matter the most. I rate St. Kilda as having the easiest fixture this year, as it plays Carlton, Fremantle, and Gold Coast twice. On the other hand, I rate Collingwood as having the hardest ‘double-ups’, as it plays Melbourne, Richmond, and West Coast (and Essendon) twice.

Of course, it is likely that some teams will perform very differently in 2019 to 2018. In 2018 teams that played Adelaide and St. Kilda twice had easier fixtures than I initially expected, while teams that played Melbourne and North Melbourne twice had tougher fixtures.

St. Kilda’s ‘80-point’ advantage in the fixture compared with an average team isn’t by itself going to make up their -500 point differential from last year. If the Saints significantly improve though it may well help them with a finals push. It probably helped Richmond to a better finals spot when they improved in 2017, and as mentioned above it probably helped Collingwood last year.

St. Kilda finally wins out in the fixture

I’m in no way a Saints fan, but they were definitely overdue for some ‘love’ from the fixture. They had one of the worst fixtures last year, and the year before that, and in 2015 and 2014 as well.

St. Kilda’s tough luck in the fixture in recent years was well covered in Squiggle’s article ‘How The Fixture Screwed St. Kilda’. In recent years the Saints have had tougher ‘double-ups’ than a team that finished in their position typically would. Also while net home ground advantage is usually fairly minor, it has mattered more in St. Kilda’s case. The Saints have tended to get less home matches against non-Victorian opponents compared with the amount of times they have had to travel across the border.

Again, the fixture by itself didn’t make St. Kilda miss the finals last year, and it likely won’t get them in the finals this year if they don’t otherwise improve. It’s good for them though, after recent years, to have a fixture that is a bit kinder to their chances.

Wednesday, January 16, 2019

Hello Quinn Wheatley

AFL WOMEN’S Power Rankings 2019: Ranking AFLW’s Two Newest Clubs

In a month’s time the AFL Women’s (AFLW) competition will kick off its third season. At the same time I will start up the third season of my weekly AFLW ‘Power Rankings’. These rankings aim to give a more accurate reflection of the ‘current’ strength of each team than the ladder does. They adjust each team’s results for the strength of its opponents, as well as whether they played the match home or away.

For my rankings though the start of the third AFLW season brings with it a new challenge. Two new teams will be joining the competition – the Geelong Cats and the North Melbourne Tasmanian Kangaroos. (Next year will be even more challenging.) How should I rank these two new clubs?

Initial ranking points: look at the premiership odds

I faced this type of situation before when I was trying to work out how to initially rank the eight clubs in the AFLW’s first year. Back then I decided the teams should not be rated as equal, since the ‘consensus’ view – as reflected in bookmakers’ premiership odds – was that they were quite different in their ability.

This also looks to be the case this year for Geelong and North Melbourne. North Melbourne are rated as a very strong side, and have the second-lowest odds to win the premiership, after last year’s premiers the Western Bulldogs. Meanwhile Geelong are rated as a relatively weak side, with the longest odds to win the flag.

Therefore I am going to start North Melbourne off on a relatively high rating of about +5 points, which is what the teams that were initially the highest ranked started on in 2017. Conversely I will start Geelong on a relatively low rating of about -7 points, which is the same as the lowest ranked team at the beginning of 2017. (There is a bit more behind my decision, but I don’t think the full method is either interesting or ‘scientific’ enough to go into all of the specifics here.) All of the teams’ ranking points are also adjusted slightly, to keep the average number of ranking points across teams at zero.

This will start North Melbourne off in fourth spot on the rankings, and Geelong in second last. The rankings update pretty quickly though because of the small number of matches, with 25 per cent of a team’s ranking determined by its most recent match alone. Hence a much better indication of the strength of these teams should emerge within a couple of rounds of the new season, as it did in 2017.

Home ground advantage: think GWS

As there is only a small history of AFLW matches to draw from, I base my adjustments for home ground advantage (HGA) on the adjustments I use for the men’s league (see table below). While I haven’t done a big review of my men’s league HGA adjustments since those rankings began in 2010, they are still based on far more data than we have for the women’s league to date.

In the men’s league I give Geelong the same HGAs as NSW teams against Victorian clubs whenever the Cats are playing in Geelong. Therefore, my HGA for the Cats in Geelong will be based on that of GWS. Following this rule and using my HGAs from last year would give Geelong an HGA of four points when playing against Victorian or NSW teams in Geelong, and eight points when playing against other clubs.

What to do for North Melbourne, which will play its home games in Tasmania? In the men’s league my HGAs for Hawthorn and North Melbourne when playing in Tasmania are smaller than their HGAs when playing in Victoria, as they play less often there. That doesn’t seem right for a genuine ‘Tassie Kangaroos’ team though. Alternatively, I could give the Roos the same HGAs as the other ‘Melbourne clubs’ – that is HGAs against non-Victorian teams at home, but no HGAs against other Victorian teams.

However I’m going to take the view that, since Tasmania is their permanent ‘home turf’, North’s HGA against other Victorian teams in Tasmania should at least be as much as the Cats in Geelong. After all, Victorian teams have to travel further to Tassie than they do to Geelong. On the other hand, I don’t think I should make them any higher than when Victorian teams travel to Sydney. Let’s make North Melbourne’s HGAs the same as Geelong’s and GWS’ then.

Note though that in the men’s league I assume Geelong is not disadvantaged against Victorian teams when they play at other Victorian venues. In AFL, Geelong plays a fair amount of games at both the MCG and Docklands. In AFLW though that looks like it will not apply. Therefore I will also add a disadvantage for the Cats (and Roos) when they play elsewhere in Victoria.

One other thing – in general, I’m not going to use the same HGAs as I have the first two years. When I first came up with my women’s league HGA adjustments I didn’t know what the volume of scoring would be. Given the scoring we have seen so far, and since I am already making some adjustments to last year’s rankings because of the new clubs, I am going to take this opportunity to reduce them slightly.

In matches where GWS, Geelong, and North Melbourne play the ‘Melbourne clubs’ I’m going to reduce the HGA from four points to three points. In all other matches involving clubs from different cities I’ll reduce the HGA from eight points to six points.

These new adjustments do not make much difference to the ranking points, but seem more in line with the scoring we’ve seen over the first two AFLW seasons (indeed, they may still be a touch high). The table below lists the new HGAs, and compares them to the men’s league.

Revised rankings: Magpies now on top, but they are not the same team in 2019

As mentioned above, these revisions make little difference to the rankings (see table below). Collingwood regain the top spot it lost to the Brisbane Lions after the Grand Final. The Lions’ showing in the GF against the Dogs in Melbourne is now considered slightly less impressive with the reduced HGA.

While the Pies’ top ranking may seem peculiar given they only finished mid-table in 2018, it is because they finished the year off really well. They had strong wins against Melbourne and Adelaide, an away win against Brisbane, and a close loss against the Bulldogs. Given that Collingwood is in the weaker half in the premiership odds, maybe one should be thinking about the Magpies as a ‘good bet’ then?

Not so fast … what the premiership odds likely reflect that the rankings do not is that Collingwood lost some top players to North Melbourne, including Jess Duffin, Jasmine Garner, Moana Hope, and Emma King. The Magpies have also lost AFLW Rising Star winner Chloe  Molloy for the season due to injury. Top-end talent matters more in AFLW than the men’s league, and the forward prowess that we saw from Collingwood in the second half of 2018 is likely to be significantly curtailed this season.

Indeed you could make an argument for a more significant shake-up of the rankings given the off-season player movements. As I said above though, if these player movements do result in large changes in performance the system will correct quite quickly. For the most part the revised rankings are not too far off what ‘the market’ is saying about how teams are expected to perform this season, so I’ll leave the tinkering there.

Monday, December 10, 2018

AFL Statistics Series #2: Scoring a Behind – ‘Scoreboard Impact’ or a ‘Missed Opportunity’?

One point for ‘trying’

What is the ‘value’ of a behind in Australian rules? On the scoreboard it is of course one point scored for your team. That’s less than the six points for scoring a goal, but better than no points and potentially the difference between winning and losing.

A behind scored by an individual player is also recorded against the name of that player, as part of their contribution to the team’s score. In fantasy football leagues a behind makes a minor but positive contribution towards a player’s fantasy points total. So while it isn’t as good as a goal, it seems like something at least.

Is it a positive though? Every supporter has known the agony of his or her team losing a match through inaccuracy in kicking for goal. It’s even more agonizing when the players are missing shots for goal that are considered relatively easy.

Shots on goal are hard to come by, and six behinds are needed to obtain as many points as just one goal. Furthermore scoring a behind gives the ball back to the opposition for a kick into play, in contrast to the roughly ‘fifty-fifty’ chance a team has of getting the ball back again from the centre bounce that follows a goal.

With that in mind should we really consider kicking a behind favourably? Should a behind be seen more as ‘impacting the scoreboard’ or a ‘missed opportunity’?

How existing player rating systems credit behinds

As I said in my first post in this series part of thinking here is perhaps to arrive at a new ‘player rating’ system. Both the Australian Football League Fantasy and SuperCoach (Champion Data) ratings give a player a point for scoring a behind. HPN’s PAV system also gives positive credit to a player for any point he scores, though as with the other systems a behind will have only a minor effect on a player’s rating.

In the AFL Player Ratings system however a player can lose rating points for missing a shot at goal. Rating points in this system essentially depend upon how a player’s action affects a team’s expected score. For example, a player taking a shot from 15 metres out would in most cases be expected to score a goal, so scoring a behind instead means the outcome was a lot worse than expected. The ‘penalty’ for missing is less if the player was taking a more difficult shot – say, from 60 metres out.

Kicking a behind could then lead to a net negative effect on the player’s rating. In the extreme if the player who missed the shot on goal gets no credit at all for creating the shot then the player has only hurt his team. On this view, it would be a similar situation to a player undoing the good work of his teammates by kicking the ball to the opposition, and it is well-known that a player loses points for this under the SuperCoach rating system at least.

Reading through the explanation of the AFL Player Ratings system for the first time did alter my view of what a behind was worth. Or maybe it just returned me to a more intuitive state of being a fan in the stands watching a player on my team miss a shot on goal – shaking my head and cursing as he ‘blew’ all the hard work of my team getting the ball up the field for a scoring shot only to get one point out of it.

Attributing a team’s points to player accuracy

Players can help their team score by contributing to the creation of scoring shots, and by taking scoring shots. Crediting players for the former is going to take a bit of work. I think though it’s relatively easy to give credit for the latter.

OK, a player is taking a shot for goal – what is the change in the team’s expected points from him converting or missing the shot? Under the AFL Player Ratings system this depends upon the expected points for the position on the ground that the player takes the shot from. However I only have public data and I don’t know where the shot was taken from. Therefore, let’s define the extra points from scoring shots as following:

Extra points created by player from taking scoring shots = Points scored by player – (League average points per scoring shot, excluding rushed behinds * Scoring shots by player)

In 2018 the league average points per scoring shot, excluding rushed behinds, was 3.85 points. Therefore, if a team creates a scoring shot and I don’t know where on the field the scoring shot was created from, I’m going to assume the value of creating one scoring shot is 3.85 points. These points can be attributed amongst the players who contributed to creating the shot, including perhaps to the player who took the shot itself.

But what is the value of simply taking the shot? That is, let’s ignore the player’s role in creating the shot, the metres gain from kicking to goal, and what happens after the score is kicked. Under the system above each successful shot by a player on goal adds 2.15 points, with the other 3.85 points going to the players that created the shot. On the other hand, if a player misses a shot he can be said to have subtracted 2.85 points from his team’s total.

This simple system does ignore shots at goal that go out on bounds on full, but I cannot get these from public statistics. Probably the bigger weakness though is it does not account for the difficulty of the player’s shots on goal, as the AFL Player Ratings system does. For example, Lance Franklin converted shots on goal in 2018 at about the league average rate. Franklin though is well known for converting longer and more difficult shots than your average forward. (One might also argue that the difficulty of shots varies at a team level – i.e. some teams create better shots than others.)

If we just use this simple system though, which players created the most value last season from scoring shots converted, and which players lost the most value? This will depend upon the accuracy of the player’s shots and his volume of scoring shots. Hence, the top players in terms of extra points added from converting scoring shots in 2018 included leading goalkickers such as Ben Brown, Tom McDonald, Jack Riewoldt, Luke Breust, and Tom Hawkins (see table below).

The ‘worst’ goalkickers do not ‘destroy’ quite as much value as the best goalkickers create (see table below). According to this system Jarryd Lyons lost about 28 points for his team in 2018 through his inaccuracy, less than half of the 63 points Ben Brown created.

The line from ‘good’ to ‘bad’ converter is relatively thin. There is little overlap among the ‘best’ and ‘worst’ converters if the same calculations are done for 2017 (though Ben Brown topped the list in both years), with Christian Petracca even flipping between the two.

On a per game basis the points gained or lost from simply converting scoring shots may seem relatively small. Even Ben Brown is only credited for less than 3 extra points per game from his accuracy. However in a league where a team scores on average 80-90 points per game – meaning that each of a team’s 22 players contributes on average about four points per game – goal accuracy can be quite significant.

It is even more significant in evaluating a player’s contribution to an individual game. A player scoring four or five behinds without scoring a goal could very well obliterate every positive contribution he has made for the game. Few ‘possession chains’ a team or a player is involved in result in scoring shots; ‘wasting’ those that do is significant.

VERDICT: Creating a scoring shot is valuable. Scoring one point rather than six points with that shot is generally a MISSED OPPORTUNITY.