Panda Crazy Script for Panda's

[turkinthailand]

Oh, I have so many questions now, lol, but this is a quite important one. If job is search job, I see 4th line output is something like this:

[Acc 84/5068 | Srch: 16211 ]

What does that tell me exactly? I think 'srch' is total number of searches? And the '84' is grabbed jobs. Is '84/5064' therefore the fraction of jobs acquired? What I really want to do is increase the 84 number to something higher

 

[turkinthailand]

The other immediate question is about leaving jobs running. From what you say, it looks like a really bad idea for me to leave 'collect' on anything that isn't very likely to get a hit? is that right? every job I have 'collect' on decreases chance of other jobs grabbing hits?

 

[Johnnyrs]

Oh, I have so many questions now, lol, but this is a quite important one. If job is search job, I see 4th line output is something like this:

[Acc 84/5068 | Srch: 16211 ]

What does that tell me exactly? I think 'srch' is total number of searches? And the '84' is grabbed jobs. Is '84/5064' therefore the fraction of jobs acquired? What I really want to do is increase the 84 number to something higher

Yes the srch is the total number of searches it has made.
The 84 is the total number of hits actually collected.
The 5068 was added afterwards from a suggestion from someone. It basically means the total number of hits it found. It only looks at the 3 highest paid hits in the search. It does try to only count unique hits but it might count hits that were sent back depending on how many hits there are for a requester. 5068 seems like a lot so I'm guessing you are trying to get hits from a requester that puts out a lot of hits all at once. A search mode might not be the best way of doing that because it only tries the first 3. It works best for batches. So for some examples from my PC today: ipsoft: 7/9, tax engine: 31/5, c-sats 0/17. It's not totally exact because I believe I saw only 1 tax engine hit today but the script says there was 5. It's counting hits that either was sent back or it's because of how mturk throttle hits. The 0/17 for c-sats doesn't look very good. I might have too many hits collecting, missed them because of bad luck or I might have already did those hits yesterday. So it's something you can use to find out if a search is useful but doesn't give the best information to figure what is really happening.

 

[catnapped]

Is there any feasible way (in the future) to add a setting (either individual pandas or collectively) so that it gives up after 'x' number of searches?

 

[Johnnyrs]

The other immediate question is about leaving jobs running. From what you say, it looks like a really bad idea for me to leave 'collect' on anything that isn't very likely to get a hit? is that right? every job I have 'collect' on decreases chance of other jobs grabbing hits?

Yes that is totally correct. I recommend keeping it under 10 so the chance of grabbing a hit is higher. A really good place to look is the status log tab at the bottom. It will show all jobs collecting and the most important stat is the Elapsed time at the right. That shows you how many seconds it takes to get back to that hit to try to collect it. If the Elapsed cycle timer is set at 1 second then if you had 5 jobs collecting it should say 5 seconds for all the jobs. The more jobs collecting the higher the elapsed time will be.

There is a way to make a job skip a cycle which can make 1 job only try at maybe once a minute. The hit weight in the details window is for that. The larger the weight number the more the job skips and it won't affect the other jobs as much. It's useful for rare hits that have big batches that could last for a long time so grabbing them first before others isn't as important. I'm testing it with noblis right now at 50 weight so it only checks it at 1 minute. It only adds 1 second elapsed time to the other jobs every minute. It's a bit complex but I tried to explain it.

 

[Johnnyrs]

Is there any feasible way (in the future) to add a setting (either individual pandas or collectively) so that it gives up after 'x' number of searches?

I could add something like that in the future. I could see it work well for surveys so if you forget about it then it can just turn itself off after an hour. But some surveys do get returned so it could be a problem. I'll think about where I can put something like that and how a user could easily tell it to give up after a certain amount of time or amount of searches.

 

[catnapped]

I could add something like that in the future. I could see it work well for surveys so if you forget about it then it can just turn itself off after an hour. But some surveys do get returned so it could be a problem. I'll think about where I can put something like that and how a user could easily tell it to give up after a certain amount of time or amount of searches.

It's also good for things like UNU or remesh where if you haven't caught it in 'x' amount of time, there's no point anyway.

 

[turkinthailand]

Just as a slight aside, if you return a job, are you still eligible to pick it up again, or does Amazon ban you from jobs you've already dropped?

 

[Johnnyrs]

Just as a slight aside, if you return a job, are you still eligible to pick it up again, or does Amazon ban you from jobs you've already dropped?

Well banning isn't the best word for it but yes if you returned a hit then you won't be able to pick that hit again unless the hit ID gets changed. Not sure the process on how mturk assigns the hit ID. Hit ID is different than the group ID. So for example if a batch drops and only has 100 hits for everyone to do and you return 5 or have 5 expired then you can only do 95 of those hits in the batch.

 

[Deleted member 4195]

I've tried using this and just don't think I'm setting everything up right. Could someone PLEASE screenshot their panda crazy pages so I know the best way to do this?

Would be a big help!!

 

[turkinthailand]

Is there any guide to running PC at the old site and new site together? I have firefox running now, so I thought I would run one version on Chrome and one on Firefox, just for a bit of experimentation.

Normally I start PC with:
https://www.mturk.com/mturk/welcome?pandacrazy=on

How to start on the old site ?

cheers,

 

[Johnnyrs]

Is there any guide to running PC at the old site and new site together? I have firefox running now, so I thought I would run one version on Chrome and one on Firefox, just for a bit of experimentation.

Normally I start PC with:
https://www.mturk.com/mturk/welcome?pandacrazy=on

How to start on the old site ?

cheers,

I haven't had time to update the Panda Crazy help website with the newest features so there isn't any instructions in there. I did mention how to use it on the new site in the release notes which could be found in this thread somewhere. But I can just tell you the urls to use here.

Old site url: https://www.mturk.com/mturk/welcome?pandacrazy=on

New site utl: https://worker.mturk.com/?filters[search_term]=pandacrazy=on

There is another url for it to be used on the new site but I think these urls are better and makes more sense anyway. It really can help out with those hits that get eaten up quickly.

Also for anyone who didn't know about this feature. I recommend strongly to use separate browsers if you want to use the old site and new site at same time. The script could corrupt the data if trying to save to the same data space. Something I should have thought of before adding this but It didn't occur to me.

 

[Mary L]

Hello,

I'm sure the answer is somewhere but I've looked all over and can't find it. On Panda Crazy, how do you add a specific requester that will grab any HIT they drop without adding multiple Panda's for them?

 

[Kat8mouse]

Hello,

I'm sure the answer is somewhere but I've looked all over and can't find it. On Panda Crazy, how do you add a specific requester that will grab any HIT they drop without adding multiple Panda's for them?

You add a job and put the requester ID in - then hit add panda for group id - then go in and paste that requester ID in the requester ID space and take it out of the group id space. Select search where is says "action mode" . . . it should now search for any hits by that requester.

 

[turkinthailand]

I seem to be using a different dashboard now:
https://worker.mturk.com/dashboard?ref=w_hdr_db
I don't recall changing my setup, so is this a new dash from Amazon? Will it require script changes?

 

[Wgeo]

I haven't had time to update the Panda Crazy help website with the newest features so there isn't any instructions in there. I did mention how to use it on the new site in the release notes which could be found in this thread somewhere. But I can just tell you the urls to use here.

Old site url: https://www.mturk.com/mturk/welcome?pandacrazy=on

New site utl: https://worker.mturk.com/?filters[search_term]=pandacrazy=on

There is another url for it to be used on the new site but I think these urls are better and makes more sense anyway. It really can help out with those hits that get eaten up quickly.

Also for anyone who didn't know about this feature. I recommend strongly to use separate browsers if you want to use the old site and new site at same time. The script could corrupt the data if trying to save to the same data space. Something I should have thought of before adding this but It didn't occur to me.

Thanks for putting this script out there for everyone to use. Really appreciate it. Just switched over to using PC on the new worker website. Is there a way to get the PCHelper to run on the new site? I can only get the boxes to show up on the old site. Thanks!

 

[catnapped]

A few of us have noticed that the queue script seems to bounce between the old site and worker as of late, another Amazon "improvement" ?

 

[Johnnyrs]

Today I released a new update for Panda Crazy Helper (0.2.9) to greasyfork. Be sure to update it if greasemonkey/tampermonkey doesn't update it automatically. Should update it automatically sometime tomorrow. Be sure to backup your data by exporting it from the jobs menu at top so you won't lose anything if something goes wrong.

New Features:

• Will now add buttons on worker.mturk.com which will send panda's to Main Panda Crazy script using the same worker.mturk.com domain. It will not pass panda's to PC on www.mturk.com unless you are still using www for the buttons. So be sure you are running PC on the correct domain when trying to add panda's. PC running on www will not get messages from worker url or vice versa. That's a browser security feature that I can't change.

This update makes sure that the script will run correctly on the new worker site domain. The buttons are in the show details of each hit when showing on the all hit page. Click on each hit to show the buttons or click on the show details at the top to show all buttons on each hit. Everything seems to be working correctly just like it does with the old site. The queue helper script won't work on worker site yet. Will take some time to figure out what I can do with that.

When mturk starts to redirect people from www.mturk.com to worker.mturk.com then you have to use a new url to run on the old site:

old site: https://www.mturk.com/mturk/welcome?pandacrazy=on&doNotRedirect=true
new site: https://worker.mturk.com/?filters[search_term]=pandacrazy=on

Notice the added text at the end of the old site url: &doNotRedirect=true

 

[Turker2013]

What happens on December 15th if the old site doesn't work anymore?

 

[JKS]

This update makes sure that the script will run correctly on the new worker site domain. The buttons are in the show details of each hit when showing on the all hit page. Click on each hit to show the buttons or click on the show details at the top to show all buttons on each hit. Everything seems to be working correctly just like it does with the old site. The queue helper script won't work on worker site yet. Will take some time to figure out what I can do with that.

Thank you for the updates! I have noticed that the buttons do not show on the "no hits available" page on the new worker site after the update ... is there a trick to that or is it still in the works ?