There have been many discussions in the forums about the way that points are awarded in BOINC. We have been reviewing them and in order to validate what the members have been observing and to collect better data, we sent out the same workunit 800 times about two weeks ago.

The purpose of points is to award an equal number of points to different members for accomplishing the same amount of scientific work. By sending out 800 copies of the same workkunit we are able compare the points claimed by different computers for the same work.

We have been evaluating this data. There are two key observations that match what the members have found.

1) Linux earns fewer points per workunit then the other platforms.
2) About 5% of results claim points significantly in excess of the average for their platform.

In order to address issue #1 we have sent our findings to Dr David Anderson at BOINC. He said that they will be compiling the Linux agent in the 5.8 release with optimizations that are comparable to the optimizations used on the other platforms. This should allow the credit claimed by Linux to match the other platforms. This new version should be available with 6 weeks or so.

In order to address issue #2 we are going to implementing a new algorithim for awarding credit granted. In the new algorithim we will no longer automatically discard the high and low values during validation in order to compute the credit granted. Instead we will exclude any claimed credits that are statistical outliers and then average the remaining values. Additionally, in cases where the claimed credit is significantly higher then the granted credit, the member will only be awarded half of the granted credit. Finally, invalid results will only be granted half credit as well.

We are about to launch beta tests of new graphics for FightAIDS@Home and Help Defeat Cancer. During these beta tests we will be running the new algorithim for points awarded to ensure that they work correctly.