HiRez Poll About This Section - 2007

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
Maybe what I should do is something simple like:

If a title 20 total [10]s and [9]s, then this would remove a single [1] from the chart calculation. 40 would remove two [1]s, and so on. But, that's not real scientific.
 
How about something simple like removing a 1 if there there are no other 1s and no 2s or 3s. Though I don't see any instances of it, the opposite could also be done - remove a 10 if there are no 9s or 8s. This would be similar to Cai's graphs without doing the work, so therefore less scientific as well - but easy! I guess you would have to have some minimum number of votes as well so that you know the vote is really an abnormality.

I would tend to say leave the votes as they are, though I do see the point of removing them as well.
 
I've given this some thought (no, really?) and I think I have an approach that is both simple and statistically valid.

1. If there are at least 15 votes total and a gap of at least two "has no votes" scores exists between a "has at least one vote" score and a "has only one vote" score, then the outlying one vote score shall be ignored.

2. If there are at least 30 votes total and a gap of at least one "has no votes" scores exists between a "has at least one vote" score and a "has only one vote" score, then the outlying one vote score shall be ignored.


Example for rule 1:

PollRule1.jpg


In this case the "3" score would be ignored since there are more than 15 votes, less than 30 votes and a gap of three "has no votes" exists between the distribution and the outlying score with only one vote.


Example for rule 2:

PollRule2.jpg


In this case the "1" score would be ignored since there are more than 30 votes and a gap of one "has no votes" exists between the distribution and the outlying score with only one vote.

ASSMPTIONS:

- Outlying scores would not actually be deleted, just ignored for purposes of calculating the weighted average (used on Jon's master ranking page). This would eliminate the error of deleting outliers that might converge with the distribution over time as more and more votes for a title are received.

- Outlying scores are only considered for exclusion if there is only one vote. Apparent outliers with two or more votes are retained as valid.
 
I applied my proposed rules to the "top 50" of the current rankings list. There were 12 scores that changed, summarized below:

Code:
NEW	OLD	DIFF	ARTIST		TITLE
9.56	9.21	3.75%	Bob Dylan	BLOOD ON THE TRACKS
9.54	9.38	1.67%	Deep Purple	MACHINE HEAD
9.53	9.13	4.47%	Bryan Ferry	BOYS AND GIRLS
9.31	9.24	0.81%	Elton John	GOODBYE YELLOW BRICK ROAD
9.27	8.96	3.42%	Elton John	TUMBLEWEED CONNECTION
9.18	9.10	0.85%	ELP		BRAIN SALAD SURGERY
9.02	8.93	1.07%	Elton John	CAPTAIN FANTASTIC
9.00	8.53	5.52%	O'Jays		SHIP AHOY
8.93	8.80	1.50%	Porcupine Tree	IN ABSENTIA
8.84	8.73	1.20%	Fleetwood Mac	RUMOURS
8.75	8.54	2.45%	Carpenters	THE SINGLES
8.69	8.51	2.10%	Beck		SEA CHANGE
 
Nice work. I like that no votes are actually deleted.

These polls are great and I think they've added a lot to the site - I often read the reviews to find new discs to buy and wind up with music that's new to me that I really enjoy. I think many of the 'outliers' were cast early on before we all got a feel for the system. Obviously we will all vote in a manner that makes sense to us, but your system corrects for some of that.
 
JonUrban said:
A NOTE ON VOTING

Please, do our membership a favor.

Do not vote a low score if you do not like the artist
Do not vote a low score if you do not like a format
Do not vote a low score just to "screw things up"
Do not vote if you do not own or have not heard a title, just because you like or dislike the artist/format/recording.

Let me again suggest a separate listing for musical content and for audio production/presentation. These are independant considerations for me and no more so than in the vast majority of the recordings posted here.

That's why I rarely vote.

Kal
 
Kal Rubinson said:
Let me again suggest a separate listing for musical content and for audio production/presentation. These are independant considerations for me and no more so than in the vast majority of the recordings posted here.

That's why I rarely vote.

Kal

Kal,

If there were an easy way to do it, I would. I understand how there could be a disc that SOUNDS really good and has an excellent surround mix, but the music or the type of music is something that you REALLY dislike - be it country, hip-hop, pop, jazz, classical, etc.

This would not only disuade someone from getting the disc, but should they get it, voting on it in a positive light.

With this forum structure, there would have to be TWO polls per title. If I did this, there would be no way to keep the two polls linked together, thus the "musical content" poll would migrate to page 4, while the production poll might fade to page 6. In the case of a popular title, people would end up voting for the music even if they did not own the surround disc, just because they liked or disliked the music.

If someone could come up with a way to do this, or an alternate method to do it, I'd be open to suggestions.

I also have to get a few more titles up there. Folks have PM'd me some titles, and I will get to them shortly.
 
Jeez, didn't we have this discussion already? Kal, the vBulletin software does not support the voting methodology you would like us to pursue. It's like complaining that your Ford Taurus can't fly. There's just no point in bemoaning something that is simply not supported by the system.
 
JonUrban said:
It would be a royal pain. The "list" comes from a raw database dump from the poll database. Once it's exported, I have to move it into excel and try and format it for the list. At that point, I would have to hotlink each entry into it's respective slot. Holy smokes.

And it would have to be redone everytime I do the poll, as the chart is redone each time.

If you click the link to the polls on the "list" page, it renders that section in alphabetic order, so at least you can easily find a disc!

KHMMMM.... Not the best solution, really;)
For my site I wrote useful script in php. For example, list of all audio discs sorted by average.
It can separate DVDAs/SACDs/DTSes, sort them by votes summary, average, count of votes and popularity. And of course, any item is clickable.
I can help to do the same for QQ, but I am unsure about vBulletin formats and data handling.
 
Wow! That looks very professional! I'm impressed!

Cai is working on something now, but your implementation is fantastic. Very nice!
 
Cai Campbell said:
Jeez, didn't we have this discussion already? Kal, the vBulletin software does not support the voting methodology you would like us to pursue. It's like complaining that your Ford Taurus can't fly. There's just no point in bemoaning something that is simply not supported by the system.
OK. OK. Of course, I would never have considered buying a Taurus.

Kal
 
Kal Rubinson said:
OK. OK. Of course, I would never have considered buying a Taurus.

Kal
Hey...My 2000 Taurus just hit 100k miles and nothing has fallen of of it...YET:D
Still going strong getting me to work and back at 50miles each way 5 days a week!
Tim
 
timw said:
Hey...My 2000 Taurus just hit 100k miles and nothing has fallen of of it...YET:D
Still going strong getting me to work and back at 50miles each way 5 days a week!
Tim

OT but I drive only on weekends and it has to be fun as well as functional. Currently, a MB wagon with only 35K miles on it but I once hit 220K with one of my previous MBs. Of course, I got rid of that one out of boredom and I believe the buyer is still using it.

Kal
 
2006 UPDATE - THE HI-REZ POLLS


There has been much talk about the HiRez Poll Chart and the effects of one bizarre negative vote affecting the rating of an exceptional disc. In the past, I have been reluctant to just trash the low vote, as I want to respect each and every members vote in these polls.

However, in a matter of fairness because many people (not just QQ members) use the chart for reference when selecting discs, it seemed unfair to have a disc with all high votes and one "1" vote be rated lower than it should because of the odd vote.

So, I am going to change the way the chart is calculated. This calcualtion WILL NOT CHANGE the actual vote as seen in each thread. The change will only be reflected in the chart.

Cai Campbell has created a SQL script that will take the raw forum poll data and process it for the monthly chart. The original rule that a disc must have at least 10 votes in the forum thread to make the chart stands firm, however, the new calculation of the score will include the algorithms that Cai explained above in factoring out the odd votes.

Cai can elaborate if he wants to, but this will make the chart a better representation of the actual feelings for the disc, yet keep the poll data solid and true.

Expect to see the new, updated chart on February 1st.

Please, if you have the time, be sure to register your thoughts on the discs that you own and have listened to. The more data, the better the data! :mad:@: :smokin :mad:@:
 
What about "Newbie Ecstasy" effect, when any first surround sound album is voted as 10 ? It is very distinctive in my polls, where only mix quality is evaluated. In QQ it can be not notably at all - mix is only 1/3 part of mark (content/sound/mix). Most voters, I think, are expirienced in some degree in "content/sound" par ;)
JonUrban, I think that you as Board Operator watched for many members evolution in telling good from bad ))

May be it has sence to add "coefficients" to votes in determining average, depending on total number of polls participated... ?

Drawback is in some kind of separating...
But it can be stimulus for active voting too.
"The more data, the better the data!" :)
 
JonUrban said:
.. I have been reluctant to just trash the low vote, as I want to respect each and every members vote in these polls.
If you want to dim the effects of some extreme votes in a simple way, you could leave (one of) the lowest and (one of) the highest scores out of the calculation.
 
narcopolo said:
How about adding Depeche Mode's 4 (so far) surround sound SACDs?

Which 4 are out? I'm not up on these.............................yet! :D

I'll add them when someone lists the titles.
 
Back
Top