Why are the passing scores on OCPJP low?

Tina Smith
Ranch Hand
Posts: 208
9
Not to ignite a can of worms here...

I've written a few other certification exams in the last two years and their pass percentages are at or above 80%. Which makes sense to me because in order to get a certification you should be able to prove that you know more than 1/4 of what you're talking about (assume 4 choices, 25% right by knowledge and 25% right by guessing = 50% ...and yes I know that doesn't add up in real world math). So why are the passing scores on the OCJP and OCJA in the 50-60% range? Is it because the questions are just that difficult?

And will that change when the OCJA 7 and OCJP 7 exams are out of beta? They've stepped up the exam, will they also step up the marks you need to pass?

Any thoughts?

Henry Wong
author
Marshal
Posts: 21003
77
Tina Smith wrote:
I've written a few other certification exams in the last two years and their pass percentages are at or above 80%. Which makes sense to me because in order to get a certification you should be able to prove that you know more than 1/4 of what you're talking about (assume 4 choices, 25% right by knowledge and 25% right by guessing = 50% ...and yes I know that doesn't add up in real world math). So why are the passing scores on the OCJP and OCJA in the 50-60% range? Is it because the questions are just that difficult?

And will that change when the OCJA 7 and OCJP 7 exams are out of beta? They've stepped up the exam, will they also step up the marks you need to pass?

I was under the impression that it was done by statistics. The beta exams are use to flag questions that don't work, and to get a feel of the right / wrong ratio for the questions that do work -- from that a passing score is calculated.... Basically, it "is grading on curve". Of course, if the beta exams are attended by too many smart people, then it kinda messes up the bell curve....

Henry

Tina Smith
Ranch Hand
Posts: 208
9
I'm sure someone at Oracle has done a study of the % in the beta exams and the % in the real exams so they can compenstae for the people taking the betas.I'd (randomly, I can't substantiate this) guess that those who take the betas usually either have a certification and want to upgrade to the latest piece of digital paper (and improve their skillset), work in the field and see an opportunity, or have been studying for the previous version of the exam, so pass rates would be higher. Of course there are always those who say: for \$50, I can try and I just might succeed (and if you pass, kudos to you).

Even if they're bell-curving though, to curve at 50% either means that very few people pass (hard to get > 50%), or the bar is set quite low (20% fail rate). Any thoughts on this and its impact to the value of the certification?

Or do they, when you say grading on curve, curve the results before you get them? ie if you get 90/100, but 50% do better than you, then you're awarded a 50%? This would seem almost impractical to me since you could re-grade the same exam and get a different score depending on the current pool of grades.

Henry Wong
author
Marshal
Posts: 21003
77
Keep in mind that this is all speculation at this point -- so this whole discussion may be moot.

Tina Smith wrote:I'm sure someone at Oracle has done a study of the % in the beta exams and the % in the real exams so they can compenstae for the people taking the betas.I'd (randomly, I can't substantiate this) guess that those who take the betas usually either have a certification and want to upgrade to the latest piece of digital paper (and improve their skillset), work in the field and see an opportunity, or have been studying for the previous version of the exam, so pass rates would be higher. Of course there are always those who say: for \$50, I can try and I just might succeed (and if you pass, kudos to you).

Too complicated. It is just easier to keep the beta going until the statistics stabilizes. As data sets get larger, the outliers become less significant.

Tina Smith wrote:Even if they're bell-curving though, to curve at 50% either means that very few people pass (hard to get > 50%), or the bar is set quite low (20% fail rate). Any thoughts on this and its impact to the value of the certification?

Or do they, when you say grading on curve, curve the results before you get them? ie if you get 90/100, but 50% do better than you, then you're awarded a 50%? This would seem almost impractical to me since you could re-grade the same exam and get a different score depending on the current pool of grades.

I think you are missing the point of grading on the curve. With grading on the curve, you don't care what the pass score is, you pass a certain point on the bell curve. The passing score is then determined from this point -- not the other way around. The purpose of the beta test would be to detemine that point, which you would then use for the general access test.

Henry

Bert Bates
author
Sheriff
Posts: 8898
5
Well the Java exams are meant to be hard. I think that Henry's basically got it right - it's kind of "grading on a curve".

Also, the typical question has 5 options and 2 are correct, so if you're guessing you have a 1/10 chance of getting it correct.

Javin Paul
Ranch Hand
Posts: 295
Another reason possibly is coverage , coverage has been kept increasing from Java 5 and since there are more topic to covered and some of them like Generics, RegEx are real complex combined with multiple answers may result in this drops.