Some (or many) of you might have noted or suspected recent site anomalies besides the failure of the top-ten scores to update without refreshing the screen. E.G.
- A familiar calibration you were sure was flowing claimed to be a stall, but showed the stall location far outside the target oval
- Annotating a movie you thought you recognized as a calibration only gave you 116 or fewer points
- What you thought was a "real" movie from the dataset was scored as a calibration
- Sometimes refreshing the screen, especially with a calibration movie, results in a new movie being displayed - other times not
Anyway, Pietro graciously provided an update on what was going on, and said I should share it with other catchers who might have similar questions. Following is a slightly edited version of what he provided.
One possible culprit for these mysterious changes is that we had a large segment of our code that was modified when (our former coder) left the project, but had never been properly compiled. There were also changes in preparation of a new, improved interface design. Recently, we improved automation in our process for updating code on the servers. We did not realize that this suddenly compiled all the changes the former coder had made before she left, many of which introduced (or reverted to) odd behaviors. We are now systematically trying to fix these.
Additionally, our new developers wish to rewrite the entire code base from scratch to make it easier to understand and maintain the code, and to create new features. (Any of you who've ever done any coding know how hard it can be to understand another coder's work.) This means that they are dividing their time between addressing issues such as the ones you encountered on the legacy code, and building something newer and "shinier" from scratch. So we have a bit of a balancing act in terms of allocating dev resources. On top of that we are supporting an experiment in human/AI cooperation that will be run separately from Stall Catchers, but using Stall Catchers movies and open to any catchers who want to participate. This will probably run in early to mid December. Of course this is pulling some dev resources aside as well, but supporting this effort helps us continue to keep Stall Catchers running now and in the future.
Much of this may not be directly pertinent to your specific observations, but (Pietro) wanted to share the broader context so you know what's going on.