A Crowd Conundrum
“We finally get a big crowd in and look at what we've served up... AGAIN!"
Much has been made of the inability of Melbourne City to make the most of previous momentum and/or big opportunities. In this season alone, we’ve choked in the FFA Cup final, failed to show up against Perth and Victory at home then not capitalised on being a man up over Sydney for over an hour.
Anecdotally, it sometimes feels like the best performances we have as a team are those when we have a small crowd in, but when we’ve got a relatively large crowd in the performance or results drop.
Simply stating that we underperform in front of large crowds isn’t enough – is it something that we can categorically prove throughout our history?
We took a *very* deep dive into our attendances from the last ten years to figure out if there really is a relationship between our performances and the size of the home crowd we're playing in front of.
Just as a forewarning, there's some pretty intense statistical analysis coming up, but we'll do our best to summarise it in simple terms as well.
The method to our madness
Simply looking back at a handful of games and saying we dropped points when we shouldn’t have isn’t enough to statistically prove that we don’t perform well in front of large crowds. To perform a true statistical analysis of Heart/City’s performances, it’s important to have a structured methodology.
This is a retrospective cohort analysis of Melbourne Heart/City’s crowds and match outcomes from August 5th, 2010, until the final home match of 2019, the December 22nd Melbourne Derby.
Data collated included opponent, date and time of match, result (win/loss/draw, including goal difference) and who the coach was. Some extenuating circumstances (i.e. Del Piero or David Villa playing) were accounted for.
The crowd numbers were assessed as derby or non-derby crowds and compared to our season average. Our average crowds were also compared to the A-League average crowds to assess whether our average trend matched that of the league. Differences in crowds were also assessed compared to the non-derby crowd average for the season and the crowd from the home game prior.
Particular note was taken of the team’s current win/loss/draw streak, to see if that either affected crowd numbers or match outcome.
For those stats nerds out there, we’ve used a linear regression model to assess the trend of outcomes based on crowd figures, as well as a chi-square analysis to see if there is a statistically significant trend of outcomes based on crowds in three categories: <6499, 6500-8999, >9000.
We’ll leave the raw data at the bottom of this article as a download to encourage peer review and allow people to fact check our math.
Do the basics right, lads
Since our inception as Melbourne Heart, our average non-derby home crowd is 7485 patrons. The peak of this was during the Tim Cahill era in 2016/17, with a home non-derby crowd average of 9396. The lowest average was during our first season in 2010/11, with an average non-derby crowd of 5828.
We’ve played a total of 116 games at home (table 1) that aren’t derbies, with 57 wins, 30 draws and 29 losses, giving us a winning percentage of 49% and only losing 1 in every 4 games. This is on the higher side of average compared to the rest of the A-League sides (range 39-54%) with only Sydney and Victory posting a higher winning percentage at home. In the 75 games of the City era, we’ve won 41, giving us a winning percentage of 55%, which would be a league-high.
The vast majority of our games have been played on a Friday night (35, 30%). Our most successful match slot is Sunday afternoon (17W, 7D, 4L, 61% win percentage) and the session where we’ve suffered most of our losses is Friday evening (14 losses, 40% matches in this time slot lost).
Amazingly, we’ve only had 37 home matches that have followed on immediately after a win in the previous game. Of these 37 games, only 16 of them follow on from multiple wins (8 games on a 2-win streak, 5 games on a 3-win streak, 2 games on a 4-win streak and 1 game on a 5-win streak). In the games that follow on from wins, we’ve won 19 (51%), drawn 8 (22%) and lost 10 (27%), which is similar to our normal win/loss/draw percentages, suggesting that we don’t improve our home performances based on our previous form.
Our season average crowd (including derbies) has never been higher than the league average (table 2), but the trend we’ve had over our existence has been similar to that of the league. There doesn’t appear to be a significant crowd trend based on the teams we’ve been playing, but our highest non-Derby crowds generally tend to be against Sydney FC (table 3).
Lies, damned lies & statistics
When we break our crowds down into poor (<6500), average (6500-9000) and good (>9000), somewhat of a trend starts to appear.
We’ve played 43 games with a crowd of <6500. Of these, we’ve won 25, drawn 11 and only lost 7, giving us a winning percentage of 58% and a 16% losing percentage.
Of the 46 games we’ve played with crowds of 6500-9000, we’ve won 20, drawn 13 and lost 13, giving us a winning percentage of 43% and a losing percentage of 28%.
On the 27 occasions we’ve play in front of a crowd of over nine thousand people, we’ve won 12, drawn 6 and lost 9, giving us a 44% winning percentage and 33% losing percentage.
The trend amongst those numbers is that as our crowds go up, our winning percentage drops and losing percentage increases. The sample size is, however, relatively small, so a few results here and there can change those percentages reasonably quickly.
When assessing this statistically using a linear regression analysis, there's a slight trend but it certainly isn’t statistically significant (P = 0.1296).
When further assessing the association between crowd size (poor, average, good) and outcome using a Chi-square test, there was no statistically significant link between crowd and outcome (P = 0.2509).
The biggest shock comes when assessing our results when we have “above average” crowds for the season, defined in this case as a crowd which is more than 1000 people higher than the season average. We’ve had 25 games in our existence with crowds more than 1000 greater than our season average. In these games, we’ve only won 7, drawn 6 and lost 12. This gives us a losing percentage of 48% and a winning percentage of only 28%, the complete opposite of our usual winning percentage.
The statement “we lose important games in front of big crowds” is difficult to statistically prove, particularly given the limited number of matches we have to assess from. From a purely emotional perspective, we’ve lost a number of games in front of large crowds that have led to significant fan frustration:
- 10/11 Rd1 vs CCM. Crowd 11,050, 0-1 loss. First match in Heart era.
- 13/14 Rd26 vs WSW. Crowd 10,003. 2-3 loss. Last match of Heart era. #KTRAW campaign.
- 14/15 Rd4 vs AUFC. Crowd 13,083. 1-2 loss. 2nd home David Villa match.
- 15/16 Rd27 vs AUFC. Crowd 11,454. 0-2 loss. After being top of the table 2 weeks prior to the end of the regular season, we choke. AUFC go on to beat us in a semi and win the league. End of the Mooy era.
- 16/17 Rd21 vs SFC. Crowd 10,706. 1-3 loss. Previously undefeated in 5 home games.
- 16/17 EF vs PG. Crowd 9444. 0-2 loss. Had lost away to Perth the week before. Still going in as favourite, but put in an abysmal performance and were bundled out in the first week of finals.
- 17/18 Rd5 vs SFC. Crowd 11,034. 0-1 loss. 4 game winning streak to start the season, top of the table clash and sitting in first place. All downhill from there.
- 18/19 Rd3 vs SFC. Crowd 12,086. 0-3 loss. A miserable season where crowd figures were decimated from this point onwards.
The counterpoint to this is that we've still had a reasonable number of games with large crowds where we’ve had great results; remember the time we had signed Lucas Neill on a contract while we were Heart, only for him to break the contract and sign with Sydney FC? We played them that week in front of 13,752 fans and flogged them 3-1. It’s not all doom and gloom.
What have we learned today?
The question we raised at the start of this is whether City chokes every time we have a large crowd in. Statistically, no, they don’t. There are certainly signs of a trend towards better performances in front of smaller crowds, but nothing decisive.
Part of the reason for this may be that the teams that pull larger crowds are generally higher on the ladder (ie. Sydney FC) and thus, harder matches to win. The other factor to consider for crowds is our ladder position at the time of the games. Our peak crowds were around the Mooy/Fornaroli/Novillo era when we were in the top four and scoring for fun.
One of the limitations of this assessment is that we weren’t able to source the ladder positions of ourselves and our opponents at the time of the match, which may have added additional context as to whether our success leads to higher crowds, or whether we’re more likely to lose against teams ranked higher than us.
This analysis also wasn’t able to take into account a number of other extenuating circumstances, such as matches being scheduled at the same time as Big Bash League games across the road at the MCG, or games being scheduled at the exact time as the racing carnival’s Derby Day.
So what do we take home from this? Do we adopt the fairly pessimistic mindset that we're better off with lower crowds anyway and shouldn't be too concerned by this year's fairly low average attendance?
The most enjoyable era in the club’s history was that 'MFN' season where we had just one match with a crowd smaller than 7200. We lost just 3 home games out of 15 that year, winning 10 in the process.
We’ve done it before and someday we're going to get back to those levels consistently and do it again.
Raw data available for download here: