One-Synth-Challenge: General discussion thread

[wagtunes]

The change in the voting system is not to counter troll voting. Nothing will do that. The change in the voting system is to get a truer measure of the "quality" of each song for those who are not trolls and vote honestly.

In short, some 5s will end up as 9s with the new system and some 4s will end up as 9s with the new system whereas with the old system you had no choice but to give someone a 4 or a 5. Now you have a vote in between, or the equivalent of a 4.5.

This will absolutely shake up the results in the top 10 especially if you're talking about 40 to 50 tracks.

Just as a quick example, here are 10 tracks with their total scores on a 5 point system with only these 10 people voting to keep this simple.

Track 1 - 5,4,5,5,4,3,4,5,3,4 = 42 points
Track 2 - 3,4,5,5,4,5,4,4,3,3 = 40 points
Track 3 - 4,3,3,3,5,4,5,5,4,4 = 40 points
Track 4 = 5,3,3,4,4,5,3,4,5,5 = 41 points
Track 5 = 4,4,4,4,4,4,4,4,4,4 = 40 points
Track 6 = 5,4,3,3,4,5,5,3,5,4 = 41 points
Track 7 = 4,3,4,4,5,5,5,3,4,4 = 41 points
Track 8 = 4,5,5,4,3,3,4,3,5,5 = 41 points
Track 9 = 4,4,5,4,4,5,3,5,3,3 = 40 points
Track 10 = 3,4,4,4,4,4,4,4,4,4 = 39 points

As you can see, it's an absolute log jam. Now, same votes with a greater leeway, especially for the 4s and 3s and look at how drastically different a result we can get and that's for only 10 voters.

Track 1 - 9,8,10,9,7,6,8,9,6,9 = 81 points
Track 2 - 7,9,10,10,9,10,9,9,7,7 = 87 points
Track 3 - 8,7,6,7,9,9,9,10,8,9 = 82 points
Track 4 = 10,7,7,9,9,10,7,9,9,9 = 86 points
Track 5 = 9,9,9,9,9,9,9,9,9,9 = 90 points
Track 6 = 10,8,7,7,7,9,9,7,9,9 = 82 points
Track 7 = 7,7,7,7,9,9,9,6,8,8 = 77 points
Track 8 = 8,9,9,9,6,7,8,6,10,10 = 82 points
Track 9 = 8,8,10,8,8,10,6,10,6,6 = 80 points
Track 10 = 7,9,9,9,9,9,9,9,8,8 = 86 points

I won't go through every rise and fall but here are the two MOST significant.

With the 10 point system, the track that finished 1st with the 5 point system didn't even finish in the top 5 and ended up 8th.

The track that ended up LAST with the 4 point system finished 3rd with the 10 point system.

The troll voter, whoever he is, giving everybody a 1, does not alter the above.

So by all means, give your 1s to everybody. With a 10 point system, the "potential" to drastically alter the results (all the above scores make absolute mathematical sense relative to the original scores) is significant. And I stress the word "potential"

There is no question a 10 point system is better and the above scenario proves this beyond any shadow of a doubt.

[574X]

"As you can see, it's an absolute log jam... beyond any shadow of a doubt."

"I didn't find an easy way to set up an alternative voting, all methods I found require massive editing or scripting."

To loosely quote Steve Duda, "It's been done before, so we can copy and paste the code."
It's a relieving reminder that if you set it up right once, it only takes a single input to generate the results automatically, as opposed to manual data entry. Steve is a master of this.

As far as simplistic synths which emulate chips used for classic consoles, I much prefer ones with adorable GUIs. I'm very fond of this ProtoPSG

http://www.g200kg.com/jp/software/protopsg.html
I wish I had used it more when I found it, but I considered it too cpu hungry for my poor old computer, though it's actually completely reasonable at a maxium of 4.1% (on my decade-old laptop) with everything engaged, playing its 8 voices. (Don't try to turn the capacitors )

(Note the chorus and noise oscillator)

Here's a list of similar classic emulations, where I originally found it.
http://woolyss.com/chipmusic-plugins.php?s=sega

But we just spent a month doing something like that, albeit it's much more limited and not invigorating to look at

Steven's "de la Mancha" synths must be added here, because they are really excellent.
Though monophonic, it has a diverse pallete of easily accessible sounds with its intuitive controls.

https://delamanchavst.wordpress.com/201 ... -now-free/
and I've mentioned his great FM synth before, which reminds me of Live's Operator, even though it's very different.
FMMF won 3rd place at KVR DC '09

https://delamanchavst.wordpress.com/2013/10/18/fmmf/

If only FM-Four was worth mentioning, but the horror stories about his original installers becoming sentient and destroying computers , coupled with how it's still in version 0.9 after half a decade and half his website download links are dead makes it less apetizing than his nub of a release time, which is such a disappointment in an otherwise excellent synth.
Would anyone care to find a way of improving FM-Four's release?
http://www.vst4free.com/free_vst.php?pl ... ur&id=1456

[RuediRena]

Coming to a cinema in your neighbourhood!

Voting Wars Part 8


Picture by rpb1001 under Creative Commons CC-by-nc

[rghvdberg]

Coming to a cinema in your neighbourhood!

Voting Wars Part 8


Picture by rpb1001 under Creative Commons CC-by-nc

LOL

Also
"you won't believe what she voted next"

[Frostline]

There is no question a 10 point system is better and the above scenario proves this beyond any shadow of a doubt.

Eh, I think the former winner in the 5 point sample might disagree that is it "better" that they now no longer are a winner in the 10 point system.
It also seems peculiar that a track that no one deemed the best (as in getting the top possible individual score in either system) is awarded the win in the second system. Though track 5 could have gotten all 8's and then not won, so the relationship between results might be skewed to "prove" your point so to speak. Basically a 10 point system could end in different results than the current system but it does not necessitate different results.

But, I am not against adding additional rankings in potential scoring. I often find myself wishing for 3.5 and 4.5 slots to drop tracks into.

*ending direct reply to Wags post

Since I am here I would also like to ask if Spaceship Delay https://www.kvraudio.com/product/spaces ... al-entropy
is allowed to use for OSC tracks.
If one agrees not to use the phaser effect section is there anything else in the plug-in that would bar it from use in OSC? Admittedly I still don't quite grasp the "spirit" behind some of the effect rules and since I saw the question brought up in another thread but not responded to I thought I would ask here in the general section in hopes of getting an answer. Personally I don't see a reason to bar it from competition but I am also willing to admit the purely selfish reasoning of I want to play with it in my OSC tracks in the future.

Also since we are making suggestions for future OSC instruments I have found Dead Duck synths https://www.kvraudio.com/product/dead-d ... k-software to be stable and fun (at least to me) to use. But sadly no AU/AXX.

[wagtunes]

There is no question a 10 point system is better and the above scenario proves this beyond any shadow of a doubt.

Eh, I think the former winner in the 5 point sample might disagree that is it "better" that they now no longer are a winner in the 10 point system.
It also seems peculiar that a track that no one deemed the best (as in getting the top possible individual score in either system) is awarded the win in the second system. Though track 5 could have gotten all 8's and then not won, so the relationship between results might be skewed to "prove" your point so to speak. Basically a 10 point system could end in different results than the current system but it does not necessitate different results.

But, I am not against adding additional rankings in potential scoring. I often find myself wishing for 3.5 and 4.5 slots to drop tracks into.

*ending direct reply to Wags post

Since I am here I would also like to ask if Spaceship Delay https://www.kvraudio.com/product/spaces ... al-entropy
is allowed to use for OSC tracks.
If one agrees not to use the phaser effect section is there anything else in the plug-in that would bar it from use in OSC? Admittedly I still don't quite grasp the "spirit" behind some of the effect rules and since I saw the question brought up in another thread but not responded to I thought I would ask here in the general section in hopes of getting an answer. Personally I don't see a reason to bar it from competition but I am also willing to admit the purely selfish reasoning of I want to play with it in my OSC tracks in the future.

Also since we are making suggestions for future OSC instruments I have found Dead Duck synths https://www.kvraudio.com/product/dead-d ... k-software to be stable and fun (at least to me) to use. But sadly no AU/AXX.

The illustration was used to show that you COULD (COULD, COULD COULD) have different results and not that you WILL have different results.

By giving people more options instead of forcing them into a narrow voting system, I illustrated how things can end up completely different. Borderline 5s that are better than 4s don't have to be "forced" into the higher ranking out of frustration and can instead be put where they belong, in the 4.5 slot.

I can't be the only one here who listens to 3 tracks and wants to start throwing things because track 1 is better than track 2 but track 3 is in between the two of them and there's no place to put it.

If it'll make you feel any better, I know the voting system is never going to change, which is a shame. I'd be more inclined to vote in these contests (even though I am done entering them) with a more flexible voting system.

[aciddose]

The usual solution is to force votes to include every entry listed from best to worst. The worst gets a score of 1, the best gets a score of N (number of items on the ballot.)

In order to reduce the likelihood of ties irrational scores may be used in place of integers. The square root of any integer is guaranteed irrational except in the case of a perfect square. A perfect square is the square of an integer which is another integer: 2*2 = 4. Perfect squares can be skipped.

The likelihood of a tie is then significantly reduced but still not eliminated and so the benefit of irrational scoring is not clearly in all cases significantly better than integer scoring.

(It's also a trade-off as it introduces clearly undesirable biases into the result which can be argued no longer accurately reflects the popular vote.)

[Frostline]

The illustration was used to show that you COULD (COULD, COULD COULD) have different results and not that you WILL have different results.

Umm, isn't that what I also said? I intentionally used the word "could" when I wrote it.

By giving people more options instead of forcing them into a narrow voting system, I illustrated how things can end up completely different.

So then it would be useful to know if completely different results are desired to occur.
To me the desire for an alternative outcome suggests some dissatisfaction with the previous outcomes.
I am not suggesting or implying that any individual other than myself need hold that opinion or attribute that motivation to others.
Personally I have been satisfied overall with previous results so the potential for those results being different under a new system is not a strong selling point to me.

I can't be the only one here who listens to 3 tracks and wants to start throwing things because track 1 is better than track 2 but track 3 is in between the two of them and there's no place to put it.

Questionable. I have had the desire to have an extra place for tracks like track 3. Yet the denial of that desire has not ever been overly upsetting especially to the point of even considering to have a cathartic outburst.

If it'll make you feel any better, I know the voting system is never going to change, which is a shame.

Why would it make me feel better that a system change that I have freely admitted would be an addition of convenience to me is not implemented?

[aciddose]

Code: Select all

irrational score values:
        1.41421, ^2 = 2
        1.73205, ^2 = 3
        2.23607, ^2 = 5
        2.44949, ^2 = 6
        2.64575, ^2 = 7
        2.82843, ^2 = 8
        3.16228, ^2 = 10
        3.31662, ^2 = 11
        3.4641, ^2 = 12
        3.60555, ^2 = 13

Code: Select all

nutty votes:
item 1: 5,4,5,5,4,3,4,5,3,4,
item 2: 3,4,5,5,4,5,4,4,3,3,
item 3: 4,3,3,3,5,4,5,5,4,4,
item 4: 5,3,3,4,4,5,3,4,5,5,
item 5: 4,4,4,4,4,4,4,4,4,4,
item 6: 5,4,3,3,4,5,5,3,5,4,
item 7: 4,3,4,4,5,5,5,3,4,4,
item 8: 4,5,5,4,3,3,4,3,5,5,
item 9: 4,4,5,4,4,5,3,5,3,3,
item 10: 3,4,4,4,4,4,4,4,4,4,

linear sums: 42,40,40,41,40,41,41,41,40,39,

irrational sums:
        lin 24.8531, squared 617.677
        lin 24.4434, squared 597.481
        lin 24.4434, squared 597.481
        lin 24.6397, squared 607.114
        lin 24.4949, squared 600
        lin 24.6397, squared 607.114
        lin 24.6568, squared 607.96
        lin 24.6397, squared 607.114
        lin 24.4434, squared 597.481
        lin 24.2815, squared 589.59

Code: Select all

better votes:
item 1: 1,2,3,4,5,6,7,8,9,10,
item 2: 8,10,4,2,3,5,7,6,9,1,
item 3: 4,1,2,3,7,6,9,8,5,10,
item 4: 10,9,8,7,6,5,4,3,2,1,
item 5: 9,2,8,3,7,4,6,5,10,1,
item 6: 1,9,8,7,6,5,2,3,10,4,
item 7: 5,4,3,2,1,6,7,8,9,10,
item 8: 5,3,4,2,6,1,9,8,10,7,
item 9: 1,4,7,10,2,5,8,9,3,6,
item 10: 8,6,7,5,3,10,9,2,1,4,

linear sums: 10,19,27,34,40,45,49,52,54,55,

irrational sums:
        lin 3.60555, squared 13
        lin 7.06965, squared 49.98
        lin 10.3863, squared 107.875
        lin 13.5486, squared 183.563
        lin 16.377, squared 268.206
        lin 19.0227, squared 361.864
        lin 21.4722, squared 461.056
        lin 23.7083, squared 562.083
        lin 25.4403, squared 647.211
        lin 26.8546, squared 721.167

As you should see the probability of a tie is already much lower using the ranking (best/worst) system and so in this case the irrational scoring doesn't make much difference. It does make a difference though. In order to see the effect you'd need many examples of ranked ballots which result in ties or near ties and a comparison between the results.

source-code: https://pastebin.com/etrs36SU

It should be absolutely clear that a ranked ballot is a far superior method when the goal is to avoid ties while maintaining a good approximation to the popular vote.

[wagtunes]

Code: Select all

irrational score values:
        1.41421, ^2 = 2
        1.73205, ^2 = 3
        2.23607, ^2 = 5
        2.44949, ^2 = 6
        2.64575, ^2 = 7
        2.82843, ^2 = 8
        3.16228, ^2 = 10
        3.31662, ^2 = 11
        3.4641, ^2 = 12
        3.60555, ^2 = 13

Code: Select all

nutty votes:
item 1: 5,4,5,5,4,3,4,5,3,4,
item 2: 3,4,5,5,4,5,4,4,3,3,
item 3: 4,3,3,3,5,4,5,5,4,4,
item 4: 5,3,3,4,4,5,3,4,5,5,
item 5: 4,4,4,4,4,4,4,4,4,4,
item 6: 5,4,3,3,4,5,5,3,5,4,
item 7: 4,3,4,4,5,5,5,3,4,4,
item 8: 4,5,5,4,3,3,4,3,5,5,
item 9: 4,4,5,4,4,5,3,5,3,3,
item 10: 3,4,4,4,4,4,4,4,4,4,

linear sums: 42,40,40,41,40,41,41,41,40,39,

irrational sums:
        lin 24.8531, squared 617.677
        lin 24.4434, squared 597.481
        lin 24.4434, squared 597.481
        lin 24.6397, squared 607.114
        lin 24.4949, squared 600
        lin 24.6397, squared 607.114
        lin 24.6568, squared 607.96
        lin 24.6397, squared 607.114
        lin 24.4434, squared 597.481
        lin 24.2815, squared 589.59

Code: Select all

better votes:
item 1: 1,2,3,4,5,6,7,8,9,10,
item 2: 8,10,4,2,3,5,7,6,9,1,
item 3: 4,1,2,3,7,6,9,8,5,10,
item 4: 10,9,8,7,6,5,4,3,2,1,
item 5: 9,2,8,3,7,4,6,5,10,1,
item 6: 1,9,8,7,6,5,2,3,10,4,
item 7: 5,4,3,2,1,6,7,8,9,10,
item 8: 5,3,4,2,6,1,9,8,10,7,
item 9: 1,4,7,10,2,5,8,9,3,6,
item 10: 8,6,7,5,3,10,9,2,1,4,

linear sums: 10,19,27,34,40,45,49,52,54,55,

irrational sums:
        lin 3.60555, squared 13
        lin 7.06965, squared 49.98
        lin 10.3863, squared 107.875
        lin 13.5486, squared 183.563
        lin 16.377, squared 268.206
        lin 19.0227, squared 361.864
        lin 21.4722, squared 461.056
        lin 23.7083, squared 562.083
        lin 25.4403, squared 647.211
        lin 26.8546, squared 721.167

As you should see the probability of a tie is already much lower using the ranking (best/worst) system and so in this case the irrational scoring doesn't make much difference. It does make a difference though. In order to see the effect you'd need many examples of ranked ballots which result in ties or near ties and a comparison between the results.

source-code: https://pastebin.com/etrs36SU

It should be absolutely clear that a ranked ballot is a far superior method when the goal is to avoid ties while maintaining a good approximation to the popular vote.

The problem with a best to worst system is that it's a logistical nightmare for the human being who now has to listen to each track, try to remember how each one sounds and then rank them in order from best to worst. With 50 tracks in a contest, no way in hell do I want to do that. There comes a point where the system becomes so unwieldy on a "human" level that you simply don't want to do it.

Assigning scores from 1 to 10 is only marginally harder than assigning scores from 1 to 5. Ranking 50 tracks in order is hell.

How do I know?

Back when I was a kid, I used to do my own top 100 of the year, like WABC did. I took all my 45s and took my 100 favorites. I had to listen to each track and then painstakingly move one ahead of another one at a time. The "voting" took me hours. And these were tracks that I had been hearing for months during the year. Try doing that with tracks that you're hearing for the first time.

No thank you.

So while mathematically it may be the best system, the diminishing returns you get because of the time you have to invest to make it work just isn't worth it.

At least not to me.

[aciddose]

You're using a straw-man here.

Yes I'm familiar with the arguments against ranked-ballot systems and they are always flawed.

In order to use the ranked-ballot system the voter must only be familiar with any two items at a time. The comparison is then made between those two and they are re-arranged to suit the preference of the voter.

Repeat until you are satisfied with your rankings.

See: https://en.wikipedia.org/wiki/Bubble_sort


For this particular application a sound-cloud-like tool could be constructed where you have all the tracks in a playlist and simply drag one to move it. You then only need to play tracks immediately next to each other to judge whether your ranking is to your taste.

This is the only solution possible, so I would say if you aren't really interested in seeing any improvement what place have you to complain?

If you want to make entirely arbitrary and subjective choices (1-5 or 1-10) or use a first-past-the-post system you should just accept what is already used.

All systems are "flawed" in representing subjective preferences because they're subjective. This means they have no objective realization.

[aciddose]

I understand of course that time is an issue. One could simply pick random numbers without any consideration given an arbitrary scoring system.

The problem is that just as you describe there are diminishing returns the more you increase the range of scores available. 1 - 10 is actually not twice as good at avoiding ties than 1 - 5, it's only marginally better. Even allowing an infinite number of score values won't help much other than to introduce more noise into the system. It is easy to score from 1 - 5, but much more difficult to accurately score 1 - 10 and extremely difficult 1 - 100 or 1 - 1,000,000.

Just as you mentioned the voter must be comparatively aware of every item on the ballot in order to maintain accurate score values for each which for large ballots is simply impossible.

In a ranked-ballot system though you have many other options. In fact humans tend to produce far better and faster results than bubble-sort intuitively.

See: https://en.wikipedia.org/wiki/Quicksort

[wagtunes]

I understand of course that time is an issue. One could simply pick random numbers without any consideration given an arbitrary scoring system.

The problem is that just as you describe there are diminishing returns the more you increase the range of scores available. 1 - 10 is actually not twice as good at avoiding ties than 1 - 5, it's only marginally better. Even allowing an infinite number of score values won't help much other than to introduce more noise into the system. It is easy to score from 1 - 5, but much more difficult to accurately score 1 - 10 and extremely difficult 1 - 100 or 1 - 1,000,000.

Just as you mentioned the voter must be comparatively aware of every item on the ballot in order to maintain accurate score values for each which for large ballots is simply impossible.

In a ranked-ballot system though you have many other options. In fact humans tend to produce far better and faster results than bubble-sort intuitively.

See: https://en.wikipedia.org/wiki/Quicksort

Well, I guess everybody is different. Having done my own top 100s for 20 years until I got tired of it, it's not something I want to go through again. Marginally better results is still better than what we're getting and it's not that much more work.

[aciddose]

You probably just like everything. You're not picky enough and yes that would be an awful problem.

In my experience listening to OSC entries for example I can play only a minute or a few seconds of a track until I know for certain whether I like it or not.

I can categorize almost immediately like so by skipping/paging through the track:

1. Top, absolute best, loved it
2. Good but not best
3. Indifferent
4. Didn't like
5. Hated it

Once I have very quickly organized into these categories (1 - 5 notice?) I can further refine my vote in the areas I care most. I can look at the top ten tracks for example and sort those more carefully, or I can listen through the worst once and say "well, they put some good effort in" and move only the absolute worst to the very bottom.

Without further refining the ranking the result is fairly random and introduces the very same sort of noise into the system that subjectivity is responsible for given an arbitrary scoring system.

The good thing about a ranked-ballot is it's entirely up to the voter. You can spend as much or as little time as you like!

Not the case with a scoring system. You must assign a score to every item on the ballot and this is not so easy to do in a truly random way. Humans introduce very significant biases by picking 5 3x more often than 3 for example.

Did you notice you didn't use 1 or 2 at all in your example votes? A ranked-ballot doesn't allow you to score everything 3, 4 and 5. Each "score" is unique.

In fact in your supposedly "better" 1-10 system you never used anything below 6!

What exactly is the difference between 1-5 and 6-10 ? They're the same thing ...

A common issue with ranked-ballot systems is that it is possible to influence the voter through the order of items on the ballot. This has a very significant effect on first-past-the-post systems too!

The solution is so simple it isn't even worth typing this sentence: each list/ballot is printed with entirely random order.

[rghvdberg]

Shouldn't someone instead of writing lengthy post filled with scary numbers and math, find a way to turn a Google spreadsheet with a list of tracks into a form?
This needs a Google app script.

Shouldn't be that hard but to be honest I've got more important stuff to do.