Some Songfight statistics

[j$]

My head hurts now!

Thanks, Deshead, for working out my nearly 6% of the fight. Actually that's roughly what I figured, and if you say (for the sake of sakiness) that 100 people with fairly wide musical tastes vote in a fight, 6 out of 100 extrapolated up means I have a reasonable success rate, and that I'm not just wasting my time!



j$

[fluffy]

we need a "most entries without a win" ranking. call it the "king arthur" category... or did he win one?
-bill

I'd like to see where I stand on that list as well.

[Leaf]

I think, since you are all so into statistical goop (like, don't you get enough of that at your job????) that a really awesome spreadsheet ...or better yet, report would have to include some kind provable stat for every songfighter... you know... most common collab name, , biggest avatar, most average use of the world "speckle", most frequent attempts at hitting on Bjam, , most obtuse posts, songfighter with the average minimum songlength,... stuff like that. Oh... did ya'll take the nerd test?

[Caravan Ray]

...and if you split all the statistics into Northern Hemisphere/Southern Hemisphere groupings - I should at least make it into the top 3 of the Southern Division.

[JonPorobil]

Some of these can be misleading because of various artist names and pseudonyms. I also met the Octothorpe 26-weeks-in-a-row record, from "So Aggravating" through "Driving." And many of those weeks saw me in multiple songs.

[fluffy]

Tough.

[deshead]

The best measure of consistency is standard deviation, if you have the means to calculate that.

Indeed I do. The table's hard to format in PhpBB, though, so I'll link to it:

[table] Average Percentage of Vote by Songfighter

It appears, then, that Silent E is the King of Consistency, what with the consistently getting no votes, and such.

I'd like to see where I stand on that list as well.

From the linked table: fluffy porcupine, 9 fights, 4.3% of the vote on average (with a standard deviation of 2.7)

What if you do that stat up in percentages?

Fight with highest percentage of one-time-only fighters:(full list)
Flood Victim - 60.0% (3/5)
Golfpunk Drives A Cadillac - 60.0% (3/5)
Postcard - 50.0% (11/22)
Secrets - 50.0% (8/16)
Failure and Regret - 50.0% (5/10)
Eternity Without You - 50.0% (4/8)
Do It For Captain - 50.0% (4/8)
In The Ditch - 50.0% (3/6)
Adam & Steve - 50.0% (2/4)
Asassinate Queen Genius - 50.0% (2/4)
Standing In The Death Car - 50.0% (2/4)
The Last Words of Cotton Wool - 50.0% (2/4)
Glass Eye - 48.0% (12/25)


The other end of the table is interesting, too

Fight with lowest percentage of one-time-only fighters:
Acid Mouth - 0.00% (0/13)
Black Hole - 0.00% (0/11)
Cancer - 0.00% (0/11)
Fear is Free - 0.00% (0/11)
WTF - 0.00% (0/11)
Hero Points - 0.00% (0/10)
Chaos vs. Order - 0.00% (0/10)
Baba Ganoush - 0.00% (0/10)

Those are all fights in which none of the artists were one-time-only. I'm not sure if the titles have anything in common that might contribute to this.



Here are a few more stats I thought of. First, one we can all aspire to:

Most fights without a vote:
8 - Andre Was Here At Midnight
7 - Baker
6 - Jimmy Richards
6 - Jonny Juice and the Good Time Boys
6 - Poor June
5 - Charcoal
5 - Wages
4 - Abecedarian
4 - Feldspar
4 - Kamakura


Next: I'm trying to think of a rank based on how consistently an artist places high in the voting. One idea: a "run-off" type score, where the artist gets points based on standing, and the points are averaged over all the artist's fights. I used the following breakdown:
1st place - 10 points
2nd place - 5 points
3rd place - 3 points
4th place - 1 point
5th and under - 0 points

Ya, it's a little arbitrary, but the results actually seem to capture what I was after:

Average points per fight for artists with at least 3 fights:(full list)
10.00 - MC Frontalot
8.33 - Southwest Statistic
8.00 - Nil
8.00 - Pipe Fist
8.00 - ADD
7.50 - Kompressor
7.00 - A Kid Named Ace
6.67 - Everybody Likes Hats
6.67 - Revel Yth
6.66 - Josh Woodward
6.60 - Frankie Big Face
6.40 - Redcar
6.00 - Milwaukee Youth Center Choir
6.00 - The Def Author
...
0.00 - Masters of Grip
0.00 - Lightning Ear Fart
0.00 - Kokiri Warriors


Excluding the one- and two-fighters, Frontalot is the only artist who's never lost. Moving down the list: SoWestStat, Nil, and Pipefist always rank highly ... Josh and FBF usually do ... Kokiri Warriors, not so much.


One last stat: I've always noticed a correlation between the number of submissions in a fight, and the total number of votes that fight receives. I wondered if this is actually the case, and if so, has it always been the case: (apologies for the logarithmic y-axis.)



The correlation is pretty direct! I know we probably all knew this instinctively anyway, but seeing it laid out, it's shocking how closely the two lines follow eachother. Conclusion: the best way to get more votes is to have more submissions. (Heh, it's either that or "the best way to get more submissions is to have more people vote"... I've always struggled with cause and effect.)

Some of these can be misleading because of various artist names and pseudonyms.

Ya, it's just a matter of how you interpret the stats. For the purpose at hand, don't think of 'artist' and 'person' as synonyms. (In other words, Blind Faith, Cream, and Eric Clapton are all different artists.)

and that I'm not just wasting my time!

No sir, you most certainly are not.

[fluffy]

Instead of having time on the X axis and two separate submissions/votes lines on the Y axis, how about doing a scatter plot with submissions on the X axis and number of votes on the Y axis? That'd do a much better job of showing a correlation.

[bortwein]

I feel so left out... I've been in 2 fights..... I want to play too.


Yeah yeah Yeah I can hear you all now... "then submit more songs... blah blah blah"


I'm just curious how I stack up, too.

[jack]

hey bort, you can always pretend you're edward tufte

[deshead]

I feel so left out... I've been in 2 fights..... I want to play too... I'm just curious how I stack up, too.

Bortwein, 2 fights, average percentage of vote = 6% (with a standard deviation of 4.55). So on the first table above, you'd be just behind the JBB.

how about doing a scatter plot with submissions on the X axis and number of votes on the Y axis? That'd do a much better job of showing a correlation.

Yep:

[fluffy]

Yeah, that shows a much lower correlation than what you'd think there was just by looking at the temporal plot, which really just shows that the vote ebb and flow happens at roughly the same time as the submission ebb and flow. Though, as a nitpick, you swapped the axes from how they should be. (The number of submissions is the input - the independent variable - while the number of votes is the output. Which only makes that badly-correlated linear regression all the more misleading.)

How about plotting ds/dt against dv/dt? That might be pretty interesting. (You can approximate ds/dt and dv/dt by subtracting the previous week's value from the current week's value.)

[fluffy]

I'd like to see where I stand on that list as well.

From the linked table: fluffy porcupine, 9 fights, 4.3% of the vote on average (with a standard deviation of 2.7)

Uh, what were you measuring there? I've been in <a href="http://trikuare.cx/audio/fp/free/">way more than 9 fights</a>. Like, 25 without a win (24 if you don't count the 'entropic pincushion' alias, 27 if you do count my multiple 'so aggravating' submissions), and that doesn't even have all of the aliases I've entered under. Or were you only counting fights you had vote data for?

[deshead]

Uh, what were you measuring there? I've been in way more than 9 fights.

Huge caveat with all these numbers: there are no vote counts for the first 80-odd fights (Let's get Naked is the first one where the full results were saved, though the numbers for So Aggravating are missing too.) So more accurately, you've been in 9 fights for which statistics can be calculated.

[fluffy]

Okay, I think that people just wanted to see (for the King Arthur award) who had been in the most fights without ever winning, <em>period</em>. I know I'm interested in that number, at least, and it doesn't need vote counts. :)

I think 15-16 puzzle and King Arthur both have me beat, but just barely.

[deshead]

Okay, I think that people just wanted to see (for the King Arthur award) who had been in the most fights without ever winning, <em>period</em>.

OK, that one's easy:

The King Arthur Award (Number of fights without a win): (full list)
47 - WreckdoM
36 - King Arthur
36 - Johnny Cashpoint
34 - The Idiot Kings
32 - Blind Mime Ensemble
31 - 15-16 puzzle
25 - Eddiebangs
24 - fluffy porcupine
24 - Level Nivelo
23 - Poor June
21 - The Landfills
21 - Jon Eric
21 - Caravan Ray

[fluffy]

Wow, I'm really out of touch. At least I'm finally on a list though!

[jack]

there was this period of songfight, back when there was only one fight per week, when certain folks liked to change their name each week. this probably amounted to skewing the stats down the road but it was still fun to do.

i did some of my own math too. not that anyone cares but.....

my many names and fights i've been involved in.

jack shite (26)
brody (7)
alfred e. shrub (1)
yokels only (1)
the mockingbirds (1)
the IRC all stars (2)
captain emo (1)
biggie shite and the songfight boy band (1)
la playa caballeros (1)
insects ahead (1)
the krispy kremes (1)
jack's super duper band n' stuff (1)
mojack (1)
a rich king's fat jester (1)
kc and the sunshine boys (1)
w2 and the temps (1)
jack daniels (1)
yoko oh no! (1)
johnny in the corner (3)
the bloody hams (2)
the glatex love (1)
the iron clef (1)
hedafine (1)
the songfight orchestra (1)

total songfights: 59 (i think)

[Leaf]

I care.

[fluffy]

Of course, anyone can just go into word salad mode, record some random noodling on some instruments, and shout out the title over and over again until they pass out from hypoxia and submit it under a new name, but only Wesley Willis could do it <em>right</em>.

[LMNOP]

Okay, des, I've got one for you if you're up for it -- stemming from the anamolies in your entries-to-votes ratio. The question is: who consistently attracts the most "friends"?

I don't mean to imply that this is necessarily negative. I suspect that some folks (like maybe Josh) attract an outside fan base to particular fights. Others (like maybe State Shirt) may generate greater-than-usual interest within the songfight community itself. Then there are the little snot-nosed punks...

So I did a little data regression on a partial sample and I propose that the "expected" number of votes in a particular fight is equal to 92 plus 1.15 times the number of entries in the fight minus 1.00 times the number of entries in the other fights that week. Or so.

Let's call any fight with at least 10 more votes than expected a "friendly fight" (over the last four weeks, "Ninja Gang" and "When It Snowed" would qualify). And lets call any entry in a friendly fight with at least 10 votes a "potentially friend-attracting entry" or PFAE.

So divide a songfighter's number of PFAE's by his/her total number of entries and -- voila! -- you have his/her, uh, "F factor". Yeah, that's it.

[fluffy]

Lies, damn lies, and statistics.

Consider that maybe certain bands are so much better than the others that they happen to attract more of the vote, or mobilize more people to actually vote. Like, I don't vote every week, I only vote when I hear one particularly outstanding song.