
Now I'm really off to bed. See you tomorrow.

Theoretically, that in my example above B wins even though A wins every head to head. Every way of counting such votes has a problem of this sort. For what we're doing here I think average rank is the fairest way, but since we've done head to head the last two cups people are primed to look at it as if each film has 3 head to head matches where the voters contribute to each match, rather than what is actually happening, which is each voter's ranking is really its own match.Mario Gaborović on Dec 19 2017, 04:21:01 PM wrote:![]()
![]()
What's wrong with average rank?
Cool, so I guess Joachim would then assign the points (#1 gets 3 wins, #2 gets 2, #3 gets 1).PeacefulAnarchy on Dec 19 2017, 05:44:47 PM wrote:We watch all 4 films and rank them 1,2,3,4 no ties allowed.
This is how I imagine the scoring should go if there are no draws; average rank. It's only logical.PeacefulAnarchy on wrote:So for the example above
10 people vote A>B>C>D
9 people vote B>C>D>A
A is 30W-27L Avg 2.42
B is 47W-10L Avg 1.52
C is 28W-29L Avg 2.52
D is 9W-48L Avg 3.52
Head-to-head records against other three opponents; 2 wins, 1 loss means more people voted for your film than for the other two, but the fourth one was more preferred by others than your film.Cynical Cinephile on Dec 19 2017, 06:01:13 PM wrote:How does this "2 wins, 1 loss" vs "1 win, 2 draws" scenario occur?
To continue the soccer/football analogy (which is where we've been getting most of our inspiration for this cup). head-to-head (i.e. 3W-0L) would equate wins/losses/ties, while total votes (i.e. 30W-27L) would equate goals. Total vote scores should produce the same result as average rank, I think.PeacefulAnarchy on Dec 19 2017, 04:30:59 PM wrote:Theoretically, that in my example above B wins even though A wins every head to head. Every way of counting such votes has a problem of this sort. For what we're doing here I think average rank is the fairest way, but since we've done head to head the last two cups people are primed to look at it as if each film has 3 head to head matches where the voters contribute to each match, rather than what is actually happening, which is each voter's ranking is really its own match.Mario Gaborović on Dec 19 2017, 04:21:01 PM wrote:![]()
![]()
What's wrong with average rank?
So for the example above
10 people vote A>B>C>D
9 people vote B>C>D>A
The way we've seen it is:
A is 3W-0L Avg 2.42
B is 2W-1L Avg 1.52
C is 1W-2L Avg 2.52
D is 0W-3L Avg 3.52
but I'd say a more accurate representation is
A is 30W-27L Avg 2.42
B is 47W-10L Avg 1.52
C is 28W-29L Avg 2.52
D is 9W-48L Avg 3.52
This points system also makes sense to me. Using PA's sample data from above, the points would beCynical Cinephile on Dec 19 2017, 05:33:36 PM wrote:How does the scoring works actually? 4 films in a group, let's name them A, B, C and D respectively.
3?Wouldn't it be easier if we simply ranked the four films by order of preference and score
4 points for #1
3 for #2
2 for #3
1 for #4?
Of course, I'm only talking about group phase.
Anyway, scoring system used in most sports should work for this as well...win 2 points, loss 1 point (let's forget about football for now). The winner is obviously the film that gets the most points. Here's my proposition for tie-breaking procedures:
1. head to head winner
2. most wins
3. average rating
So to answer your question, 2 wins and a loss are better than 1 win and 2 draws, in case that head to head, the two films are tied.
I like this.joachimt on Dec 20 2017, 01:08:15 AM wrote:I've slept on it. I think the easiest, most fair way to do this is:
1. Average ranking (lowest two ranks proceed)
2. Ties will be broken head-to-head.
3. Still a tie? Most #1 votes.
4. .........
5. Still a tie? 3 day extension.
Should there be another tie-breaker on 4th place?
Mario might think of best vote difference, but I think that's too complicated.maxwelldeux on Dec 20 2017, 02:43:22 AM wrote:Only #4 I could think of is "Fewest #4 votes."
I think I remember how we solved ties between two or three countries in previous seasons. But if the tie involves all the four countries we're back to the situation of your initial post. I'm still not sure what we're doing in that case. Using the 3/1/0 score system? Proceed to step 3?maxwelldeux on Dec 20 2017, 02:43:22 AM wrote:2. Ties will be broken head-to-head.
That would only be a problem if it is a tie between more than two movies. I didn't consider that option, but of course it is possible.sortile9io on Dec 20 2017, 09:58:19 AM wrote:I think I remember how we solved ties between two or three countries in previous seasons. But if the tie involves all the four countries we're back to the situation of your initial post. I'm still not sure what we're doing in that case. Using the 3/1/0 score system? Proceed to step 3?maxwelldeux on Dec 20 2017, 02:43:22 AM wrote:2. Ties will be broken head-to-head.
I think the chances we get a 3 day extension are very small.Rules on wrote:1. Winner = lowest average ranking
2. Tie between two movies? Head-to-head decides.
3. Still a tie after #2 or tie between three or four movies? Most #1 votes wins.
4. Still a tie? Fewest #4 votes wins.
5. Still a tie? 3 day extension.
Pure coincidence, but a funny one! I only just saw this thread and the nominations.monty on Dec 17 2017, 07:29:45 AM wrote:Quietly pleased to note that burneyfan watched my nom a few hours ago and then promptly favorited it.
I'm afraid I'm late to the party, but isn't this the most plausible option? Looks ten times as simple to me. Which would be nice, because also non-beta guys could understand how the scoring works. ;)Fergenaprido on Dec 19 2017, 11:38:07 PM wrote:To continue the soccer/football analogy (which is where we've been getting most of our inspiration for this cup). head-to-head (i.e. 3W-0L) would equate wins/losses/ties, while total votes (i.e. 30W-27L) would equate goals. Total vote scores should produce the same result as average rank, I think.PeacefulAnarchy on Dec 19 2017, 04:30:59 PM wrote:Theoretically, that in my example above B wins even though A wins every head to head. Every way of counting such votes has a problem of this sort. For what we're doing here I think average rank is the fairest way, but since we've done head to head the last two cups people are primed to look at it as if each film has 3 head to head matches where the voters contribute to each match, rather than what is actually happening, which is each voter's ranking is really its own match.Mario Gaboroviæ on Dec 19 2017, 04:21:01 PM wrote:![]()
![]()
What's wrong with average rank?
So for the example above
10 people vote A>B>C>D
9 people vote B>C>D>A
The way we've seen it is:
A is 3W-0L Avg 2.42
B is 2W-1L Avg 1.52
C is 1W-2L Avg 2.52
D is 0W-3L Avg 3.52
but I'd say a more accurate representation is
A is 30W-27L Avg 2.42
B is 47W-10L Avg 1.52
C is 28W-29L Avg 2.52
D is 9W-48L Avg 3.52
In this example, A is the winner for me, even if it's divisive.
This points system also makes sense to me. Using PA's sample data from above, the points would beCynical Cinephile on Dec 19 2017, 05:33:36 PM wrote:How does the scoring works actually? 4 films in a group, let's name them A, B, C and D respectively.
3?Wouldn't it be easier if we simply ranked the four films by order of preference and score
4 points for #1
3 for #2
2 for #3
1 for #4?
Of course, I'm only talking about group phase.
Anyway, scoring system used in most sports should work for this as well...win 2 points, loss 1 point (let's forget about football for now). The winner is obviously the film that gets the most points. Here's my proposition for tie-breaking procedures:
1. head to head winner
2. most wins
3. average rating
So to answer your question, 2 wins and a loss are better than 1 win and 2 draws, in case that head to head, the two films are tied.
A - 49
B - 66
C - 47
D - 28
which turns out to give the same results as average rank and votes... I should have seen that coming.
So now I'm not sure which method I prefer. I also need to think about it more.
You're suggesting expressing the lowest average ranking as a sum of points instead? I'm ok either way, they're mathematically equivalent rankings. And yeah, the flow of this whole conversation has been a bit confusing, but it seems to have worked out ok.joachimt on Dec 20 2017, 01:01:50 PM wrote:What if we only use #2 for two-way-ties? Three-way-ties can get tricky very quickly. So let's try this.
I think the chances we get a 3 day extension are very small.Rules on wrote:1. Winner = lowest average ranking
2. Tie between two movies? Head-to-head decides.
3. Still a tie after #2 or tie between three or four movies? Most #1 votes wins.
4. Still a tie? Fewest #4 votes wins.
5. Still a tie? 3 day extension.
Thanks. And yes, indeed, because it's much easier to read natural numbers than numbers with a fractional component, when only having a quick glance (like I just did in the thread, earlier tonight). Makes it more accessible for people who are bad at math, as well. And it's the most accessible when the winner simply gets the most points, so everyone gives 4 points to the film he likes most, etcetera.PeacefulAnarchy on Dec 20 2017, 08:13:44 PM wrote:This is what joachim settled on.You're suggesting expressing the lowest average ranking as a sum of points instead? I'm ok either way, they're mathematically equivalent rankings. And yeah, the flow of this whole conversation has been a bit confusing, but it seems to have worked out ok.joachimt on Dec 20 2017, 01:01:50 PM wrote:What if we only use #2 for two-way-ties? Three-way-ties can get tricky very quickly. So let's try this.
I think the chances we get a 3 day extension are very small.Rules on wrote:1. Winner = lowest average ranking
2. Tie between two movies? Head-to-head decides.
3. Still a tie after #2 or tie between three or four movies? Most #1 votes wins.
4. Still a tie? Fewest #4 votes wins.
5. Still a tie? 3 day extension.
Average rank is easier to calculate, so that was my plan indeed, but I can do total points if non-betas prefer it.PeacefulAnarchy on Dec 20 2017, 08:13:44 PM wrote:This is what joachim settled on.You're suggesting expressing the lowest average ranking as a sum of points instead? I'm ok either way, they're mathematically equivalent rankings. And yeah, the flow of this whole conversation has been a bit confusing, but it seems to have worked out ok.joachimt on Dec 20 2017, 01:01:50 PM wrote:What if we only use #2 for two-way-ties? Three-way-ties can get tricky very quickly. So let's try this.
I think the chances we get a 3 day extension are very small.Rules on wrote:1. Winner = lowest average ranking
2. Tie between two movies? Head-to-head decides.
3. Still a tie after #2 or tie between three or four movies? Most #1 votes wins.
4. Still a tie? Fewest #4 votes wins.
5. Still a tie? 3 day extension.
What if I present the score so far in the right order, something like this for example?Gershwin on Dec 20 2017, 08:37:38 PM wrote:I find the inverted order, where 1.32 for instance is a higher ranking than 3.24 because it represents the most #1 rankings, very confusing, because my instinct always tells me the winner has the most points.
I could look for something, but it takes time to sort it out.mjf314 on Dec 21 2017, 12:51:29 AM wrote:I think most of the time both methods have the same winner, so it won't make a big difference. Does anyone have any examples from past seasons where the winner would have been different if we used average rank?
What annoys me about my example is that while it shows a "flaw" in using average rank it also shows how non-continuous head to head is. The divisive film becomes all or nothing. Average rank is fairer given how we're voting.mjf314 on Dec 21 2017, 12:51:29 AM wrote:One of the purposes of the head-to-head method was to not make people too afraid of picking divisive films, but whether that's a good thing or a bad thing is a matter of opinion.
In PeacefulAnarchy's example, it seems more fair to me if the winner is the film that wins all 4 head-to-heads, rather than the film with the best average rank.
I think most of the time both methods have the same winner, so it won't make a big difference. Does anyone have any examples from past seasons where the winner would have been different if we used average rank?
Depends on the circumstances, see above.mjf314 on Dec 21 2017, 01:00:14 AM wrote:Another advantage of head-to-head is that it lessens the effect of downvoting (although I don't know how much downvoting happens in the world cup).
You need to re-run it from the start for 2nd place (unless you want a film which is everyone's second choice but no one's first to come in last, though a close variation of this can still happen with the right vote split), and makes vote counting quite a bit more cumbersome. It's better in some ways and worse in others, I prefer it for actual elections where there are other factors in consideration, but I don't know that it's really worthwhile in this context.connordenney on Dec 21 2017, 03:11:42 AM wrote:What about preferential voting like they do in the Oscars for Best Picture?
You mean the first rule of all?Lonewolf2003 on Dec 21 2017, 10:17:53 AM wrote:I’m a bit lost, so sorry if this already was dealt with. But how does these counting systems work when someone hasn’t seen all movies. Or does one have to see all in a match? I don’t remember if we had that rule last times.
Rules in the group phase:
- You must watch all 4 films of a match before you vote.
We always had that rule. Otherwise you can't vote.Armoreska on Dec 21 2017, 10:35:33 AM wrote:You mean the first rule of all?Lonewolf2003 on Dec 21 2017, 10:17:53 AM wrote:I’m a bit lost, so sorry if this already was dealt with. But how does these counting systems work when someone hasn’t seen all movies. Or does one have to see all in a match? I don’t remember if we had that rule last times.Rules in the group phase:
- You must watch all 4 films of a match before you vote.
Would you consider presenting average rank like this?joachimt on Dec 22 2017, 07:16:01 AM wrote:Average ranking
Araya: 2.079 (73-41)
Court: 2.079 (73-41)
The Dark Knight: 2.711 (49-65)
Cabascabo: 3.132 (33-81)