Conversion Rates in Finals versus Home-and-Away Games

Subsequent to the previous post, which looked broadly at the differences in aggregate scoring metrics for Home-and-Away games compared to Finals from the same era, Friend of MoS, Liam, made a great suggestion for a follow up analysis.

By way of context for his suggestion we should note that some potentially confounding factors in the previous analysis are that:

• only some teams play Finals, but all play Home-and-Away games
• Finals see only particular pairs of teams face one other, while the Home-and-Away season includes some games between these same pairs, but most between completely different pairs

Liam's suggestion then was to restrict the analysis of games in any single season just to those involving the same pairings as appeared in the Finals. So, for example, in 2014, we saw Port Adelaide play Richmond, so we should include from the 2014 Home-and-Away season their sole meeting from Round 17. If a pairing played multiple Finals contests in the same season or met multiple times in the Home-and-Away season of a single year, we'll combine the scoring data for all such multiple contests.

View fullsize

Today I'll just be using that data to revisit the question of whether or not teams convert at a lower rate in Finals compared to Home-and-Away season games. In the first analysis I'll look, separately, at every pairing of teams seen in the Finals series and compare the aggregate Conversion rate in those games with the Conversion rate for the same pair of teams in games played during the Home-and-Away season.

The results appear in the table at right and show that, for example, of the 40 team pairings witnessed during the 1897-1909 era (recognising that some pairings might have repeated in different individual seasons if the same teams played Finals in different years), 16 pairings recorded lower Conversion rates in their Finals encounters than they did in their Home-and-Away encounters during the same season. The figures for other eras can be interpreted similarly.

What we see then is a pattern of comparative Conversion rates by era similar to that which we saw in the previous blog, with the exception that the 1970-1984 era here shows higher Conversion rates in Finals than in Home-and-Away matches between the same teams. In the previous analysis this era showed a very marginally lower Conversion rate in all Finals compared with all Home-and-Away season games.

Most notably, the last two eras show a preponderance of pairings with lower Conversion rates in Finals than in Home-and-Away contests. Neither, however, is statistically significantly above 50%. (Incidentally, of the 548 pairings across the entire period 1897-2014, the difference between their Conversion rates in Finals versus Home-and-Away contests attains statistical significance at the 5% level in only 24 cases. This is largely because the sample size of Scoring shots for any pair of teams in the Home-and-Away season or in Finals, tends to be small with the result that differences even as large as 20% points do not attain significance).

We might also, instead of analysing the data on a pair-by-pair basis, pool the Home-and-Away and Finals data within a season for all games involving the pairings witnessed in the Finals. This methodology is similar to that used in the previous post but excludes from the Home-and-Away season any games that were not between pairs of teams seen in a Finals contest in that year.

View fullsize

The results for this analysis appear in the table at left and are broadly similar with those above and in the previous blog (which I've repeated in the right-most column of the table). The difference between the Conversion rates in Finals versus Home-and-Away season contests for the two most-recent eras is, as before, negative, but less so in this latest analysis.

As well, this later analysis, consistent with what we saw earlier in this post, suggests that Conversion rates in the 1970-1984 era might have been slightly higher in Finals than in Home-and-Away contests, which is a reversal of what we found in the previous post.

Overall then, this new evidence seems to support the view that teams from the modern eras have tended to record lower Conversion rates in Finals than they did in Home-and-Away season contests. The best estimate of the effect size, however, remains an extremely small one - of the order of 0.5% to 1%. That's equivalent to suggesting that about 1 Scoring shot in 100 or 200 that would have been a Goal in the Home-and-Away season is, instead, "converted" to a Behind when it comes in a Final between the same teams.