OK, seemingly good points. But let's get back to my other point above. Why do so many requesters leave masters and seem to not go back? *shrugs* Like the examples I gave above? Doesn't that tell us something? Aren't they voting with their feet there?
For every good requester that has left, there are plenty that have stayed. Also, newer requesters are likely unaware that the default is Masters at first.
Yeah but almost definitely less money then masters especially in the long run. Say you wanted to post 100k hits so large batch at say 10 cents each. Make 10k of those masters and use those as the control answers this cost 1300 then make a non master batch of 10k and look at how many people matched masters answers take the best of those and assign them the qual this cost another 1100. Now you have 90k hits you post to your qualed people costing you 9900. Total cost for everything 12300 which is still less then the amount paid to have masters do the entire batch which would come out to 13k. The usefulness of this depends on how much you actually trust masters but the created qualed batch should theoretically be on par with the masters.
They've stayed because their probably satisfied. But have they really tested things out and run two batches simultaneously (master vs non-master) to see how big difference things make? I think they stay with masters as much out of inertia as anything lol
How on earth does anyone still think having masters has anything to do with speed, quality, quantity, accuracy, skills, or ability? Full disclosure: I don't have it, I probably never will, I've resigned myself to that. However, people are getting it who are suspended, who haven't been around for months, who have LOST it and got it back, who have 1000 hits, who have 10 dollar a day goals. All of this tells you that accuracy, results, speed, ability, experience, etc are not deciding factors. They won't tell anyone what the magic calculation for masters is because if you quantify exactly what the criteria is then a requester would be able to decide if it's really worth 20% extra or not. As long as it's mysterious they can continue to allude to the fact that it is somehow worth it. The fact that a requester can compile an entire work force of masters qualified workers on their own with a simple small batch of already known answers on a hit qualified for masters workers and then simply post under their own qualification seems to not have caught on yet.
The fact they are staying doesn't actually carry too much weight. It simply means that the quality of masters is good says nothing in relation to the quality of non masters. They might simply be scared to leave masters because Amazon tells them masters are higher quality. An equivalent thing to masters leaving and never coming back is master requesters leaving and then immediately coming back. That would show that they felt non masters weren't good enough and actually had experience with them.
f'n brilliant man. That would be a GREAT way to get good workers going forward (non-master workers, that is)
Not saying that non-masters do not do as well as masters. However, if I was a (new) requester and decided to spend $10K on a batch of hits, I would probably think to myself, I am spending money and Amazon is telling me that masters are the best workers - I guess that means that non-masters do not do a good job. Therefore, even if I could get the work done by non-masters and save some money, I am wasting money if the work is not done well. I know, I will just lower the pay, post it as a masters hit and then split my original budget between the workers and Bezos
There are great non-masters, but if you were to pin the average master against the average non-master, then more often than not: the master likely performs better or is wiser. Like I said, there's exceptions to the rule since there's some people who def deserve masters (Saku bein' one of them for sure; although he is still killin' it without it).
Title: Reading stories + answering questions. Requester: memoryND [A2QC7NFKE3H3DZ] (TO) TO Ratings: ☭☭☭☭☭ 2.14 Communicativity ☭☭☭☭☭ 2.68 Generosity ☭☭☭☭☭ 3.82 Fairness ☭☭☭☭☭ 3.78 Promptness Number of Reviews: 39 (Submit a new TO rating for this requester) Description: Read short stories and answer comprehension questions about them. Time: 60 minutes Hits Available: 1 Reward: $2.50 Qualifications: Masters has been granted
Here's the thing, a lot of those people who post those massive batches have a fair amount of funds at their disposal. You think they'll do ALL that work and waste all that time just to save $700 in the long run? NAH, they'll be like "FUCK IT, let the Masters have it all. Amazon vouches for their 'high quality' work anyaway!"
The average master is probably statistically better then the average non master but I'd hazard a guess it is for the same reason that the average driver is worse then the average person who has had a license for at least one month. The very new people who have just started mturk are new an unlearned so they slightly drag down the average skill of the non masters. I'm willing to guess that the average non master with 10k hits is as good as the average master with 10k hits and this continues average non master with 100k hits probably just as good as the average master with 100k hits. Assuming they also have same approval ratings. So when you control for the other average qualities of masters so the only difference between masters and non masters is that qual and not the small amount of time and hits done that go into putting you in the running for the qual the average quality is probably very similar.
OCMP5 was a bit tricky. They started as all masters. They grabbed the ones they liked and gave them the qual and then posted them as qualled but non-masters to save money having grabbed talent from the masters pool.
I just chose numbers that were easy to work with. Lots of requesters post far more then 100k hits and you could get away with a test of far less then 10k. I mean my theoretical requester was only giving out 10k dollars worth of hits which hardly makes him a high rolling requester given there are requesters who have given probably over hundred times that much, hell I've had a requester that I've done almost 10k for by myself with RnR. You have to look at it proportionally they saved 700 dollars out of 13000 a 5% savings on cost. That is actually a fairly substantial amount of savings.
Ah, the age old argument of Masters vs non-masters. http://awesomegifs.com/wp-content/uploads/dead-horse.gif Let it go!
Yeah but you notice even they opened up the qual to non masters. I mean I have it now. Kind of a little sad OCMP5 literally didn't even think masters was worth the 0 extra dollars they were paying for it.
At the end of the day, qualifications in this earth go a looooooong way. Even in the event that the average non-master would be similar to the average master, the qual sort of separates a person from the pack. Even throughout school, there's always been that one person who arguably performs as good or better than a person in better classes or programs, BUT, since that person has better quals, people will simply assume that the other person is a better performer because of stats. This whole masters/non-masters thing will be never endin'. Masters will back their qual, non-masters will as well. It's like a Chevy guy arguin' with a Ford guy. They'll both back their team no matter what, so they'll always agree to disagree.