Policy Review Policy Review - Tiebreaking Procedure

[DougJustDoug]

If you are not an experienced member of the CAP community, it is strongly recommended that you do not post in this thread.

This thread is intended to contain intelligent discussion and commentary by experienced members of the CAP project regarding CAP policy, process, and rules. As such, the content of this thread will be moderated more strictly than other threads on the forum. The posting rules for Policy Review threads are contained here.

In CAP 5, we had a lot of close polls. They were so close, that many of them could easily have ended in a tie. If that happened, we have no set procedure for determining the winner.

I propose the following steps to break a tie between options in a poll:

1) The sum total of all votes accumulated in all previous polls
2) The number of votes recieved in the single most recent previous poll
3) Topic leader choice​

Those tiebreakers would be applied in sequential order until a winner is determined. It is highly unlikely that either of the first two steps would result in subsequent ties -- but if it happened, then the TL needs to make the choice and move on.

There was talk about holding new polls to break the tie. I am STRONGLY opposed to that. If you have read my "Project Pace" thread, then you know why. This proposal provides a clear, objective way to determine a winner, without extending the length of the project. This also gives a slight reward to options that accumulate a lot of votes in early polls. Up until now, it is common to hear people remark that "Early voting leads are meaningless". If we adopt this tiebreaking procedure, they won't be meaningless.

 

[gorm]

summing up all previous votes makes sense mathematically but he fact is people change their minds over the course of a vote. i agree that tiebreaker votes make things drag on, so i'd actually move for the TL to decide this on his own. if the TL feels they are unable to make a fair descision, maybe run a tiebreaker vote (with or without bold voting+reasonins) along side the next poll...


edit: sorry.

if the final poll is exactly deadlocked i suggest opening another poll to break it alongside the next poll. if that poll cannot advance without a decision, then i believe it should be left up to the TL for the reasons above. I'm opposed to looking at previous votes since people can change their minds.

 

[tennisace]

I agree they shouldn't be meaningless, but they shouldn't have that much weight. In baseball, division winners don't go to the head to head record first, the do so after they check if they won the division. I propose that we switch 1 and 2. If you win the final poll, you win. However, if the final poll is tied, THEN you should go to the other polls. It makes more sense that way. Finally, if the Topic Leader would act as a final tiebreaker, he or she should not be able to vote on ANY poll, lest there be a tiebreaker. That way the vote isn't counted twice. If a topic leader's vote would tie a poll, then it would keep things much simpler, since the TL could easily vote, then presumably break the tie if it works out that way. Its a long shot, but its how the Vice President works in the Senate.

 

[DougJustDoug]

Maybe the OP isn't clear: This isn't changing how winners are determined in every poll.

This is the tiebreaking procedure if the final poll is a tie.

Considering that this has never happened in the history of the CAP project, I would say the impact of this rule is almost irrelevant. But, if this ever happens in the future, I would like to have a defined process for handling it. This rule doesn't change anything we do today.

 

[zarator]

Quite agree with what tennis said. Even if I find pretty difficult that we have the 2 main options tie on every poll from the beginning to the end.
I mean, lets suppose we did 3 pools, one with 10 choices, another one with 5 and the final one with 2. If there should be a tie in the final poll, we can go to the previous thread and see which of the 2 choices got more votes. If they tied even in the second poll, we can get back to the first.
This way we could be warded a bit off those who, in the final poll, vote B because they hate A. It happens, and EVO project underlined it. But should a tie take place, this system would help those who really choiced A because they want A in spite of those who choice B because they dont want A...

 

[DougJustDoug]

I'm fine with switching steps 1 and 2. I don't really care which comes first, just so long as the procedure is clear and objective. I can see the argument that a "recent win" should be more important than the "aggregate lead".

 

[zarator]

Just to say, but, if we resort to:

1) The sum total of all votes accumulated in all previous polls

After

2) The number of votes recieved in the single most recent previous poll

We dont arrive to a solution. I mean, if the first step I said dont break the tie, it arguably means that - incredibly - said A and B final contenders have tied on every poll before. So, why bother counting the votes? We will find themselves to see that "A"votes=B"votes", which is pretty obvious...

 

[Elevator Music]

Uhh... zarator that is incorrect, because there were more options in the previous polls which means not all the votes will be for only those two choices, plus people that change their minds also mess that up quite a bit.

I agree that 1 and 2 should be switched though.

 

[zarator]

Sorry but I cant understand. Lets make a concrete example.

There had been a multiple choice poll with the following 10 choices: A,B,C,D,E,F,G,H,I,J

They have narrowed down to 5 in a single choice poll: A,B,C,D,E

And finally we have the last poll: A versus B

Lets suppose that A and B tie with 100 votes each. We go to look on the previous thread and we see:
A- 65
B- 65
C- 45
D- 27
E- 13

Damn, tie again. Then we go to the first, multiple choice poll:
A- 57
B- 57
C- 61
D- 53
E- 41
F- 11
G- 39
H- 30
I- 31
J- 16

In this case we have a tie, but if we count the vote A and B got we'll find that both received 222 votes across the polls. It is quite mathemathical:
1) A - x votes, B- x votes ----> tie
2) A- y votes, B-y votes and others which we dont care about -----> tie
3) A- z votes, B-z votes and others which we dont care about -----> tie

Of course, both A and B will have x+y+z votes in total, no matter which values can assume x, y and z.

Hope this was clear

 

[gorm]

I'm opposed to looking at previous votes since people can change their minds.

i guess I'm alone in this?

 

[Elevator Music]

Both A and B don't have to have the same number of votes in each poll though. A could have gotten 56 and B could have gotten 58 and they would still make it onto the next poll.

 

[Gothic Togekiss]

Before the PR:

When my stat spread submission was running against X-Act, Doug and Tennis was talking about using baseball logic to decide the winner of the final poll if it became a tie. All I had to say about that was "you got to be kidding me, how's that fair?!? D:" I mean, using rulings from a randomly chosen sport seem arbitrary at best. It basically gave added pressure on one of the competitors at the time by reducing his/her chances of winning by half.

Right now:

Tennis' suggestion strike a cord in my mind. I don't find anything wrong with it at the moment and I personally agree with his idea on switching #1 and #2 around.

i guess I'm alone in this?

No really but I'm still not sure how one would handle a tie if someone was a TL.

 

[zarator]

@ Elevator Music: should it be, we dont need to count. A= 56? B= 58? B wins. And, before anyone ask, I'm assuming most recent post overshadow previous ones. If the 1st poll was won by A, the 2nd was won by B and the 3rd one was a tie, then B should win the entire poll.

@ Gorm: Sincerely, you have been an exception. Most people simply voted houndoom in the final poll because they hated farfetch'd. The same in the 2nd one. Most people change their choice just because they dont want a particular choice to win. For example, I didnt vote Houndoom on the 1st poll, but I voted it on the 2nd poll because it was the only realistic farfetch'd contender votes-wise.

 

[Elevator Music]

zarator, I still don't see your logic.

Using a different example, A and B have tied with an inconsequential number of votes.

A and B both also got the same number of votes in the previous poll as well, meaning the totals need to be added (which means check the total of the first poll unless there were four polls, which rarely if ever happens [I think?])

Of course if A got more votes than B in the first poll, it would have a higher total and therefore win. There's nothing wrong with that... I think you've just looked at it differently than everyone else or something (though your way is still correct).

 

[zarator]

Differently, maybe, but with same and simplier results i think.

Basically, you say: we proceed with counting all votes who A and B got across the poll if method 1 dont tell us a winner. Methos 1, if I have not mistaken was: supposing last poll was a tie, i look at the previous poll. If A>B, A wins, if A<B, B wins, if Z=B we examine the previous other poll and repeat the same points, until either we found a winner or we see that all polls were a tie.
Now, if all polls are tied, how could you ever find different numbers comparing the total votes of with the ones of B?

EDIT TO ANSWER GORM: Just because we know of this issue, and we know that only a narrowed minority effectively has a true, uninfluenced conversion, I cant see why we should make an already difficult project even more complex just to meet the needs of this little, always fluctuating minority (fluctuating= those who change their mind are not always the same ones)

 

[RBG]

Honestly, I think the nail has been hit on the head already. The most surefire way to break ties have been listed in the proper order, that being, 2, then 1, and finally 3, using doug's original numbers. Each of those is fair. Number 2 because that poll is most important of the recent polls. Number 1 then because resorting to the Topic Leader should always be last.

One question though, what if it is a poll like the ability poll where we can use both options (such as the Ability Poll)? Would it be easy enough to let both options be used?

 

[jagged_angel]

zarator, in the example you proposed, A + B tied every time, so under DJD's proposed rules, the TL would choose.

If in poll one

A 56
B 78
C 22
D 30
E 50
F 10
G 9
H 11
I 8
J 12

and poll two

A 66
B 66
C 45
D 10
E 55

and poll three

A 98
B 98

Then you go back to the most recent poll, poll two. (this is tennis' suggestion)

That is also a tie, so you aggregate the total votes from all 3 polls.
A = 56 + 66 + 98
B = 78 + 66 + 98

so B wins.


Gormenghast I share you concerns to an extent but would be much more concerned if the TL got to choose the result straight away. Take for example CAP4. If for the primary typing, Poison and Electric had tied, the TL would have been incredibly unpopular choosing either Type since there was masses of support for each. This would be the case to some extent in every tied poll. Your proposal gives the TL a massive responsibility.


I agree with RBGolbat the nail has already been hit on the head with 2 1 3.

 

[zarator]

One question though, what if it is a poll like the ability poll where we can use both options (such as the Ability Poll)? Would it be easy enough to let both options be used?

Absolutely no. Very often, the choice of the first ability directly influcenced the other. Take Pyroak's example. Should Battle Armor find itself to compete among another defensive ability like Magma Armor - its crap, but take it as an example - in the final poll, people would be very upset if these 2 were chosen as Pyroak's abilities, because both the voters of Battle Armor and the voters of Magma Armor would want the second ability to differentiate from what they voted, and would like more something like Rock Head for an attacking set.

EDIT:@ jagged: try to apply my method. Poll 3 is a tie. Go to 2. Poll 2 is a tie. Go to 1. A<B. B wins.
It seems a lot more logical. Why should you count the vote of the last 2 polls. A and B tied both times, so the sums will be equal. Its so hard to explain?

 

[tennisace]

If all 3 polls are ties then it doesn't matter if 2 or 1 goes first, since both results end up in a tie.

 

[Elevator Music]

Differently, maybe, but with same and simplier results i think.

Basically, you say: we proceed with counting all votes who A and B got across the poll if method 1 dont tell us a winner. Methos 1, if I have not mistaken was: supposing last poll was a tie, i look at the previous poll. If A>B, A wins, if A<B, B wins, if Z=B we examine the previous other poll and repeat the same points, until either we found a winner or we see that all polls were a tie.
Now, if all polls are tied, how could you ever find different numbers comparing the total votes of with the ones of B?

OK this is exactly what I thought you thought.

Because from what I understand of Method 1, we only compare the votes of A and B from the second last poll, not the previous ones. Then we move onto Method 2.

There really is no difference if there is 3 polls. The only thing that differs is if we had more than 3 polls. In your view of Method 1, it would come down to the thir last poll, while in my view of Method 1 that is not necessarily the case.

 

[zarator]

OK this is exactly what I thought you thought.

Because from what I understand of Method 1, we only compare the votes of A and B from the second last poll, not the previous ones. Then we move onto Method 2.

There really is no difference if there is 3 polls. The only thing that differs is if we had more than 3 polls. In your view of Method 1, it would come down to the thir last poll, while in my view of Method 1 that is not necessarily the case.

I understand what you mean, but I dont see any case in which we should do more than 3 polls, especially when in the other PR we are talking about speeding up the process. I took the "no more than 3 polls" matter as a given...

 

[jagged_angel]

EDIT:@ jagged: try to apply my method. Poll 3 is a tie. Go to 2. Poll 2 is a tie. Go to 1. A<B. B wins.
It seems a lot more logical. Why should you count the vote of the last 2 polls. A and B tied both times, so the sums will be equal. Its so hard to explain?

Umm, apparently it is, because you've lost me! :-P If we both agree with tennis' suggestion then hurrah.

Ohhh. Right. I get you - yeah, EM has just said it - counting the total votes only matters if there's more than 3 polls, which there shouldnt ever be. So your method makes sense.

 

[Deck Knight]

I support 2, then 1, then 3.

My reason for this is because generally speaking, one option wins among 8 choices in a landslide. For instance, if we wanted a physical sponge with fighting resistance, a lot of fanboys might go for Ghost, whereas the vets, already having experienced the horrors of Revenankh, instead split themselves among Flying, Poison, Bug, and Psychic. Ghost gets 58 votes, and the rest end up something like 30, 25, 18, 15 and the rest get 8 or so. In a showdown between Flying and Ghost, Ghost gets a 28 vote advantage. Now lets say in the next poll, with just Ghost, Flying, and Poison, Flying gets 88 while Ghost gets 80 and Poison 40.

This second poll has narrowed it down to three legitimate options and has a greater grasp on people's preference. If the final poll between Ghost and Flying lands a dead 136-136, Under Rule 1, Ghost would win because it won a poll where support was divided 10 ways. Under Rule 2, Flying would win because when people's options were condensed down to 3, they clearly coalesced around "not Ghost" options they preferred.

Recap:

Poll 1:
Ghost 58
Flying 30
Poison 25
Bug 18
Psychic 15
Electric 8
Fire 8
Water 8
Grass 6
Ice 4

Poll 2

Ghost 80
Flying 88
Poison 40

Poll 3

Ghost: 136
Flying: 136.

Under Rule 2, Flying wins. Under Rule 1, Ghost wins. I believe Rule 2 is the more equitable of the two in expressing the wishes of the community.

 

[billymills]

I support 2, then 1, then 3.

3 should only be done if TLs are not allowed to vote in the poll, If they are allowed to vote in the poll, they should not get a second vote.

 

[Hyra]

I support 2, then 1, then 3.

<awesome reasoning>

I'm gonna support Deck Knight entirely on this one.

although Gorm raises a valid point, how many people change their mind midpolls? I guess enough to cause a tie. We really should try to account for this, even though I don't see a way how.