I think it was a reference to Malcom Knowles Quinella booklet but it wasn't exactly 9 runners. The percentage of quins that were an overlay decreased as field size increased but it happens in a gradual way ( as you'd expect ) and there was no clear cut-off point.
I did a science course at uni and had 4 years of Stats as a result. One of the fundamental questions you must ask when evaluating the outcome of an experiment is "is there a SIGNIFICANT difference in the outcomes". There are protocols and calculations which can answer this question but these are mostly ignored by many punters who can be misled by a shear weight of data such as "10years of races and 33000 races"
One of the things I love about punting is that at the end of the day all theory is tested in very rigorous manner i.e your bank is up or down and you can't argue that outcome!