I have posted an issue to the tracker MDL-12096 and a patch concerning the way that the new Gradebook on Moodle 1.9 handles the computation of grades and averages.
Since this is such a critical matter, I would like to open a discussion on the matter.
For most teachers and students, if they have a 100 point assignment and a 10 point assignment, they expect the 100 point assignment to be worth 10 times as much. But when using the default "Mean of Grades" for computing summary statistics in 1.9 Beta 2, it normalizes each assignment to a scale between 0 and 1 before computing the average. Using weighted grades does not solve this problem unless one manually assigns weight to each category and item.
I very much doubt that any teacher computing grades with a manual gradebook would do this type of computation. Moodle 1.6 did not either. Other gradebook software that I have does not do this by default. Hence it will be very confusing and not the desired behavior for most users.
For example if a 100 point assignment received 90 points, and a 10 point assignment received 8 points, the an expect grade would be 98/110 or 89.1%. The 1.9 Beta 2 gradebook returns 85%.
If normalizing grades before computation is to be retained, IMO, it should be an advanced option and the default in computing the mean should use point values.
In Petr's response, he said that this new method of computation was by design. He said that it would allow grading of scales or items with non-traditional ranges. While I respect his view, I strongly disagree that this is the proper solution.
Normally, if I give a scaled grade, I would not weight it in a linear way. For example, if I wanted to give a scale of Bad, Good, and Excellent, I would not assign Bad as 0%, Good as 50%, and Excellent as 100%. So, scaled grades will nearly always be over written.
Even more exotic numeric scales like the 200 to 800 range for the SAT (normal distribution, mean approx 500, SD 100) are clearly not intended to be linear, nor will they be able to be averaged in with typical grading scales.
The suggested work-around of manually assigning weights to each activity does not seem like a good alternative to me, nor does coming up with yet another averaging method to choose from.
If a teacher wishes to grade by normalizing grades, they would simply give each item a percentage grade and base it out of 100.
So, I urge that grades be computed as is recommended in my bug report, and if not, I would hope we can have a thorough discussion on the matter. It is easy for my to apply my patch for my school, but I think that this computation will cause Moodle much difficulty for those thinking of moving version 1.9 so this is a worthy topic for discussion.
--Gary
Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Gary Anderson -
Number of replies: 10
In reply to Gary Anderson
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Martin Dougiamas -
Hi, Gary.
I posted in MDL-12096 before I saw this. Yes, the current method was by design, since all the items by default have an equal weight (not an implicit weight governed by their maximum value).
I think a new "Sum of grades" aggregation method is a very clean solution to this, providing exactly what you want in a clear simple way. I've asked Petr to implement that.
As well as weights, you could also do this in a less-easy way by creating a manual calculation for the final column, something like:
However, I'm also keen to hear discussion on all this. Throughout the whole gradebook design I've really come to appreciate that there are often a lot of unique points of view about how grading should be done! We really need feedback like Gary's to find out what you want.
Is the way we are normalising grades something that no-one wants?
I posted in MDL-12096 before I saw this. Yes, the current method was by design, since all the items by default have an equal weight (not an implicit weight governed by their maximum value).
I think a new "Sum of grades" aggregation method is a very clean solution to this, providing exactly what you want in a clear simple way. I've asked Petr to implement that.
As well as weights, you could also do this in a less-easy way by creating a manual calculation for the final column, something like:
=SUM( A, B, C )(Then you can easily display the result as a percentage/value/letter as you want).
However, I'm also keen to hear discussion on all this. Throughout the whole gradebook design I've really come to appreciate that there are often a lot of unique points of view about how grading should be done! We really need feedback like Gary's to find out what you want.
Is the way we are normalising grades something that no-one wants?
In reply to Martin Dougiamas
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Gary Anderson -
OK. I must concede that there was less of an outcry of the new method of averaging than I would have expected. Still, I think that you will see it, as we did, when it comes to serious user testing.
Summing does not solve the problem (one still needs to know the number of points graded to compute performance), and I am a math teacher, and I would not do the manual calculation suggested above.
At Seattle Academy -- and it looks like we will go 1.9 for all classes ready ready or not in a couple of weeks -- I will add a "normalized mean of grades" as a computation option. But I don't expect it to be any more popular than mode
Still, I agree with Anthony, the former way of finding averages (both mean and weighted mean) needs to be restored and be at least a site option for administrators to set. I still would argue that this should be the out-of-the-box setting for Moodle, but the need for that should be confirmed by usability studies with actual teachers and students. But clearly we are not seeing the passion for this (either way) in this forum at this time.
Summing does not solve the problem (one still needs to know the number of points graded to compute performance), and I am a math teacher, and I would not do the manual calculation suggested above.
At Seattle Academy -- and it looks like we will go 1.9 for all classes ready ready or not in a couple of weeks -- I will add a "normalized mean of grades" as a computation option. But I don't expect it to be any more popular than mode
Still, I agree with Anthony, the former way of finding averages (both mean and weighted mean) needs to be restored and be at least a site option for administrators to set. I still would argue that this should be the out-of-the-box setting for Moodle, but the need for that should be confirmed by usability studies with actual teachers and students. But clearly we are not seeing the passion for this (either way) in this forum at this time.
In reply to Gary Anderson
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Anthony Borrow -
Gary - I suspect you are right - the outcry will come when it is released and gets life-tested. I think the default behavior of 1.9 should be what we saw in 1.8. It is great that there are more options with 1.9; however, most users are going to expect that things function the same unless they change it and I do not find that to be a necessarily unreasonable expectation. Peace - Anthony
In reply to Gary Anderson
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Martin Dougiamas -
There might be some misconception here, Gary ... with the "sum of grades" option the maxgrade of the column WILL be set to the sum of the other maxgrades (automatically).
So it should be exactly what you want. Petr's finished it actually, it should be in CVS soon.
You can choose to display the result as a percentage if you like.
As for following 1.8 gradebook ... I totally agree we should copy that behaviour in our default defaults so there are no surprises for casual upgraders, even if the older gradebook behaviour was a bit strange ... (for example, it would always average all numbers including zero values, whereas this is now an option in 1.9).
So it should be exactly what you want. Petr's finished it actually, it should be in CVS soon.
You can choose to display the result as a percentage if you like.
As for following 1.8 gradebook ... I totally agree we should copy that behaviour in our default defaults so there are no surprises for casual upgraders, even if the older gradebook behaviour was a bit strange ... (for example, it would always average all numbers including zero values, whereas this is now an option in 1.9).
In reply to Martin Dougiamas
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Matt Campbell -
I do have a question on this - when you say zero values, do you mean null, or do you mean null or 0? We may not want to include an unset value in the average, because it hasn't been completed yet, but we may want to include a grade of 0, since that would have been assigned by the instructor.
Thanks,
Matt
Thanks,
Matt
In reply to Matt Campbell
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Martin Dougiamas -
Sorry, I was not clear. I'm referring to empty/null grades. The new gradebook 1.9 setting allows these to be included as an option.
I'm not actually sure if the old gradebook distinguished between zero and NULL ... can anyone confirm this? (I didn't write or use the gradebook much in 1.8!)
I'm not actually sure if the old gradebook distinguished between zero and NULL ... can anyone confirm this? (I didn't write or use the gradebook much in 1.8!)
In reply to Martin Dougiamas
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Anthony Borrow -
Martin - 1.8 does not appear to distinguish between NULL and zero. If a teacher does not want a grade counted that they have not yet graded, it needs to be in a separate category. I had always used a Not graded category that had a weight of zero as a workaround. The new functionality in 1.9 will be very helpful. Some folks may want to continue to do this depending on the policies at their schools. Where this becomes an issue is for eligibility to participate in sporting events. If a student is absent or does not submit work and the grade is blank, some schools require that it be counted as a zero until the work is completed. Since this was the policy at my school, it meant a teacher would have to exclude the grade for that particular student which required a little extra effort on the teacher's part to get around the policy if there were extenuating circumstances.That is precisely what we wanted. So each school should be mindful of its policy about incomplete work. I liked the fact that it was a zero and the only extra work for me was having to re-categorize the assignment once I had done the grading. Peace - Anthony
In reply to Anthony Borrow
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Gary Anderson -
Anthony - I agree with you that the old behavior of counting ungraded assignments as zero should be an option. But I think that you would also agree with the case that I have been making in the tracker, that in most cases and by default, it is best to simply not factor ungraded assignments into the grade at all. This is especially the case for activities due in the future and when the teacher has not yet marked them. And it can avoid the whole uncategorized category with zero weight that we have had to hack with, and the lock hide unlock cycle in most cases, etc. Exempting processes can also be avoided-- you just put "exempt" in the comment until, and if, you decide to put in a score.
I can imagine that courses taught with the topic format (rather than the default weekly format) and where all activities were set in advance of the start of the course may want a "count up to completion" scoring system where activities count as zero until graded. But for courses taught over the time domain (like in schools and hybrid classes) or where activities are added during the progress of the course, it is best to not factor unmarked assignments into the overall grade.
Now I need to get back to preparing my final exams so that I can bring my fall trimester courses to a close this week
--Gary
Gary Anderson
Math Department Chair
Seattle Academy
I can imagine that courses taught with the topic format (rather than the default weekly format) and where all activities were set in advance of the start of the course may want a "count up to completion" scoring system where activities count as zero until graded. But for courses taught over the time domain (like in schools and hybrid classes) or where activities are added during the progress of the course, it is best to not factor unmarked assignments into the overall grade.
Now I need to get back to preparing my final exams so that I can bring my fall trimester courses to a close this week
--Gary
Gary Anderson
Math Department Chair
Seattle Academy
In reply to Gary Anderson
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by A. T. Wyatt -
Greetings, Gary!
I am glad you brought this up. I would also vote to retain 1.8 gradebook behavior or add the options discussed above. At my institution, we seem to have as many ways to calculate grades as there are faculty calculating!
For myself, I went to straight cumulative total long ago.
Thank you all for working on this. When a large number of people start using it, the extra effort will (hopefully) pay off in far fewer complaints!
atw
I am glad you brought this up. I would also vote to retain 1.8 gradebook behavior or add the options discussed above. At my institution, we seem to have as many ways to calculate grades as there are faculty calculating!

For myself, I went to straight cumulative total long ago.
Thank you all for working on this. When a large number of people start using it, the extra effort will (hopefully) pay off in far fewer complaints!
atw
In reply to Gary Anderson
Re: Discussion: The new way that Gradebook 1.9 Beta 2 is handling averages
by Anthony Borrow -
Gary - I think you are right that most teachers (at least in the U.S.) would not expect the grades to be normalized. I think as long as there is an option for handling this and preferably a way for a site administrator to set whether grades should be normalized or not by default then it should work just fine. Thanks for raising this issue. Peace - Anthony