Vote Thread for Ack's Omake Corner

[X] Meet the Heberts - 6
[X] NSW - 5
[X] Really Bad End - 4
[X] Confrontation II - 3
[X] Hostage Situation - 2
[X] Trump Card - 1

 

Okay. The spreadsheet itself has been updated (see my sig), but a full in-depth analysis will have to wait until I get home later tonight.

 

[X] Confrontation II - 6
[X] Hostage Situation - 5
[X] NSW - 4
[X] Bait & Switch - 3
[X] Meet the Heberts - 2
[X] All Alone - 1

 

[X] NSW - 6
[X] Meet the Heberts - 5
[X] Alea Iacta Est - 4
[X] One More Trigger - 3
[X] Bait & Switch - 2
[X] Security! - 1

 

Done, don't forget to like.

 

So are you just extending the average number of votes in the projections infinitely from the point where you starting counting? That doesn't quite sit right with me. I would have cut it off at a 5- or 10-vote average.

 

[X] Recoil - 6
[X] Alea Iacta Est - 5
[X] I, Panacea - 4
[X] Wyvern - 3
[X] All Alone - 2
[X] Security! - 1

 

[X] All Alone - 6
[X] Confrontation II - 5
[X] Wyvern - 4
[X] Trump Card - 3
[X] Alea Iacta Est - 2
[X] One More Trigger - 1

 

[X] Confrontation II - 6
[X] I, Panacea - 5
[X] Security! - 4
[X] One More Trigger - 3
[X] War Games - 2
[X] All Alone - 1

 

[X] MirrorVerse - 6
[X] All Alone - 5
[X] One More Trigger - 4
[X] Alea Iacta Est - 3
[X] Wyvern - 2
[X] Confrontation II - 1

All Alone?

 

[X] Nemesis - 6
[X] Junior Hero - 5
[X] Recoil - 4
[X] One More Trigger - 3
[X] Wyvern - 2
[X] Hostage Situation - 1

 

[X] Confrontation II - 6
[X] Slippery Slope - 5
[X] Junior Hero - 4
[X] Recoil - 2
[X] Alea Iacta Est - 3
[X] Trump Card - 1

 

I thought about stopping at a 10-vote average, as that's as far into the future as I'm predicting, but I figured the more votes averaged in the better.

 

Done.

 

As I see it, the issue with extending the average is that even if a story increases or decreases in popularity, it barely shows because you're averaging so many rounds. Of course, at the end of the day, it's your spreadsheet, so you can average it however you like. It's really just a matter of my personal taste.

 

[X] Recoil - 6
[X] Wyvern - 5
[X] Trump Card - 4
[X] One More Trigger - 3
[X] Bait & Switch - 2
[X] NSW - 1

 

Suggestion: Weight the results so that more recent votes are counted more highly than earlier ones.

 

You could also try a regression equation, which should get more accurate the more data points you have. Basically it predicts what the value should be, if the trend in voting remains.

I've gone ahead and done it for the currently active stories testing the results for predicting the results of Vote 68 (using all available data up to vote 67).

According to your spreadsheet for Vote 68, your accuracy (for the projected vote total) was:

Most: 100%
Average: 96.25%
Least: 57.14%

My method:

Most: 99.77
Average: 97.74%
Least: 91.71%

If you'd be willing to share your spreadsheet with me, I could add in my formulae for you and/or explain how to use the Trend function on Google Docs.

Tomorrow, when it's not 2 AM I might try making my own version just to see long term differences in projected winners.

 

A man after my own heart.

Why not just do both? If you have it set up like I'm guessing you do, you just have to enter the raw vote data each vote, and the spreadsheet does the rest for you. In which case you just need to copy/duplicate the prediction page, edit a few lines, and the job's done.

this is actually very hard to do with spreadsheets, and would require multiple cells doing background calculations instead of just one or two cells using simple algorithms to calculate things out.

 

Dammit. 3 AM. Anyway, have my own version of Jim Starluck's spreadsheet.

Spoiler: Predicitons

Vote 69

1. Security: 996
2. All Alone: 897
3. Meet the Heberts: 771

Vote 70

1. All Alone: 967
2. Trump Card: 877
3. Meet the Heberts: 835

Vote 71

1. Trump Card: 988
2. Alea Iacta Est: 979
3. Meet the Heberts: 900

Vote 72 (After this, it looks like all votes will have winners in the 1000+ area)

1. Alea Iacta Est: 1137
2. Meet the Heberts: 964
3. I, Panacea: 879

Vote 73

1. Meet the Heberts: 1029
2. I, Panacea: 970
3. Slippery Slope: 851

Due to recent trends in the stock exchange, the gains of some stories are substantially inaccurate. Mainly for those stories which have been available for voting for a long time.

So tomorrow, I'll check to see if accuracy improves by moving the initial start of calculations to vote 59. Which should be when all the current stories became available for voting.

These errors currently amount to ~256 votes last round.

For reference, the errors for Jim Starluck's spreadsheet was ~248 votes last round.

My overall accuracy was greater because I had more votes slightly off than his predictions.

 

Meet the Heberts?

 

Hmm. Looked over your stuff, and I think I have the hang of it... but... hmm. Incorporating it directly is difficult. I can't just track the trend in vote totals, because they get reset to 0 periodically. I need to track the trend in votes per round... but the trend value only predicts one round at a time. I'll need to completely redesign the spreadsheet to invoke it multiple times, once for each upcoming round. And it'll be tricky to handle the reset-to-zero mechanism.

I'll look into using it for future predictions. For now I'll stick with the tried-and-true method.

Previously...

As expected, fan-favorite Wyvern took home the prize and shattered the 1k-vote ceiling. Security! looks like it's going to try and pull the same off next round, but it might fall short if votes shift; current predictions have it hitting exactly 1,000 votes, so we'll see. No real chance of upset; the next story is too far behind.

Vote 70 goes to All Alone in its first win, climbing clear into the 900s. Trump Card flirts with the 1k line in Vote 71 but may fall a handful of votes short; but don't worry, the line won't get blue-balled because Alea Iacta Est will drive deep into it in vote 72, hard and stro -- err, I'm sorry, I'm jumping the gun here. I should save the sex metaphors for Vote 73, when Meet the Heberts wins its first vote and joins the 1k club at the same time. Perhaps the 1k line is down for a threesome?

Not to be left out, I, Panacea also passes the line in Vote 74 (and blushingly declines Meet the Heberts' teasing invitation to join it and AIE in the boudoir), while Wyvern comes back around for another pass in Vote 75, and may pass 1,100. One More Trigger and Recoil settle for hitting 1,000 in votes 76 and 77, while newbie Slippery Slope wins its first round in the upper-900s for Vote 78.


In terms of accuracy, we had several stories come in well ahead of predictions this round, none moreso than NSW, which... ah-heh... streaked back up to 91 votes, nearly 40 more than the expected value of 52 -- a measly accuracy of only 57.14%. Most accurate this round was Nemesis, coming in exactly on-target at 100% accuracy. Overall votes were 96.25% accurate for the round as a whole.


Current votes-per-round rankings:

1: Wyvern - 160
2: Alea Iacta Est - 124
3: Recoil - 115
4: Trump Card - 113
5: I, Panacea - 96
6: Security! - 79
7: Meet the Heberts - 64
8: One More Trigger - 62
9: All Alone - 54
10: NSW - 51
11: Slippery Slope - 47
12: Confrontation II - 46
13: Hostage Situation - 44
14: War Games - 40
15: MirrorVerse - 32
16: Nemesis - 26
17: Danny & Taylor - 22
18: TIED! - 20
- Junior Hero
- Bait & Switch




But what's this?!

A Challenger appears!

Well, let it not be said that I am unsporting. May the best numbers win!



As for my own vote...

[X] Meet the Heberts - 6
[X] NSW - 5
[X] One More Trigger - 4
[X] Security! - 3
[X] I, Panacea - 2
[X] Recoil - 1

 

[X] Wyvern - 6
[X] Trump Card - 5
[X] Recoil - 4
[X] Alea Iacta Est - 3
[X] War Games - 2
[X] Hostage Situation - 1

 

[] Hypnosis Idea - 13
[X] NSW - 6
[X] Meet the Heberts - 5
[X] All Alone - 4
[X] I, Panacea - 3
[X] Bait & Switch - 2
[X] Danny & Taylor - 1

Hypnosis idea has stayed the same, the smut scenes have been drifting in and out of my mind lately and I have almost worked up the motivation to actually type them up several times now....but not quite.

 

[X] Recoil - 6
[X] Security! - 5
[X] I, Panacea - 4
[X] Alea Iacta Est - 3
[X] Trump Card - 2
[X] Junior Hero - 1

 

[X] Hostage Situation - 6
[X] Trump Card - 5
[X] One More Trigger - 4
[X] Wyvern - 3
[X] Confrontation II - 2
[X] Nemesis - 1

 

*poke*

 

[X] MirrorVerse - 6
[X] Recoil - 5
[X] Trump Card - 4
[X] Security! - 3
[X] One More Trigger - 2
[X] Alea Iacta Est - 1

 

[X] Wyvern - 6
[X] I, Panacea - 5
[X] Trump Card - 4
[X] One More Trigger - 3
[X] Recoil - 2
[X] Security! - 1

 

I've redone the calculations for my spreadsheet.

New accuracy numbers for the prediction of last vote:

Most: 99.75% (Bait & Switch)
Average: 98.14%
Least: 89.51% (NSW)

Total Votes Off: 173 (down from 256)

But here's a problem measuring accuracy by the predicted total rather than the predicted gain.

Let's say Story A already had 750 votes and it was predicted to gain another 50, but actually gained 100 (twice the prediction).

The accuracy measured by total would be (800/850=) 94.12%.

But we know it was off by 50% because it was 50 points off on a 100 point vote.

So accuracy numbers measured this way:

Most: 98.33% (Trump Card)
Average: 87.17%
Least: 67.88% (Hostage Situation)

Anyway, on to comparing with Jim.

1. Wyvern 155 (my predictions have Wyvern getting less votes each round, likely due to people knowing it'll get more points anyway)
2. Alea Iacta Est 130
3. Recoil 129
4. Trump Card 110
5. Security! 101
6. I, Panacea 97
7. One More Trigger 71
8. All Alone 69
9. Meet the Heberts 65
10. Slippery Slope 62
11. Confrontation II 58
12. NSW 55
13. Hostage Situation 50
14. MirrorVerse 49
15. War Games 47
16. Nemesis 28
17. Really Bad End 27
18. Danny & Taylor 23
19. Junior Hero 23
20. Bait & Switch 19

That said, our predictions still agree on the winners so far.

Next Predicted Close Vote: Vote 73 Meet the Heberts (1,029) vs I, Panacea (1,009)