With Lazy Scholar I anonymously collect data about the abstracts that are queried for full texts and if they are successfully found in Google Scholar. With this data we can do some basic analysis of the proportion of papers that are either open access or freely available on the web. Edit 8.8.2013: there is too much data for google spreadsheet to handle so I am planning to launch a comprehensive data page in October for Open Access week. Subscribe to the blog here to receive updates.
Here is some preliminary data auto-updated daily in a Google Spreadsheet. When I have more time (and when there is more data to use) I will do more interesting analyses. The following values reflect: % of pages that have a full text available | Total searches | Date updated