Thirty-six years ago, George Lewis and Jonathan Lewis wrote about negative evidence in social research.
In the introduction to their paper, they wrote:
In social research, there is an overwhelming emphasis upon collecting positive data, whether it be in the form of statistically significant attitudes, important documents, or observer descriptions of unique settings. This emphasis, while responsible for shedding much light in previously dark areas, none the less has had the important and dangerous side-effect of minimizing the worth of negative evidence, that is, the significance of a thing’s absence. (1980:544)
Two games in last weekend’s AFL Round 8 sent me back to the paper and this concluding remark:
The reasons why researchers fail to come to grips with negative evidence are many. They range from overlooking the obvious to dogmatic adherence to ideological positions
The AFL web site provided these scoring signatures of the nine games played in Round 8.
My Google Doc copy of this image can be found here.
Pre-game, the data about current ladder position, last season’s regular season ranking and betting odds indicated wins for Sydney and Port Adelaide. Their defeats raise the fascinating question about how to report counter-predicted outcomes.
Both games were, in my taxonomy, Contested 4Q games.
In the Richmond v Sydney game:
- Richmond led at half time by 5 points
- Sydney led at the end of 3Q by 18 points
- Sam Lloyd kicked Richmond’s winning goal 31m 39s in 4Q with the last kick of the game.
In the Carlton v Port Adelaide game:
- Port Adelaide led at half time by 11 points
- Port Adelaide led at the end of 3Q by 5 points
- Carlton scored 20 unanswered points in the final 13 minutes of the game
From my data, of the two defeats, Port Adelaide’s was the more likely. I have not developed a sufficiently robust index of winning probability but my very basic measure of current form had Port Adelaide at -7 (a measure of the difference between the two teams’ current ladder position in relation to the 2015 regular season position).
The Sydney defeat was at the other end of the scale. I had Sydney at +11 for this game. (Only Hawthorn v Fremantle at +14 was a more ‘certain’ game.)
For this Round, I was not able to determine who might win the Adelaide v Geelong and Brisbane v Collingwood games. I excluded them from my predictions.
The significance of a thing’s absence
My relationship with AFL performance rests solely upon macro data. The two games that provided a ‘surprise’ for me in Round 8 were:
This game had four performance episodes that culminated in the 4Q when Sydney scored 19 points from 14.58 to 23.01. Richmond scored the last two goals of the game in the final 7 minutes.
Port Adelaide led at the end of the first, second and third quarters. Port Adelaide were controlling the first half of the final quarter. The AFL match feed suggested that at 18.33 in the final quarter “The Blues hanging in with a chance”.
I started this post with a mention of George and Jonathan’s paper. I conclude with a later paper by Mira Crouch and Heather McKenzie (2006) in which they observe:
Justification of small-sample studies hinges most frequently on phenomenological assumptions (broadly speaking) which underwrite investigations of personal experience in a largely subjectivist framework
It has to be borne in mind that exploratory research has little value if it is restricted to stand-alone acts – ‘‘owned’’, as it were – by individual researchers, rather than embedded in fields of relevance that are tended by communal knowledge-building labour.