*The New Voting Method is Better

*The New Voting Method is Better - Printable Version

+- International Simulation Football League (https://forums.sim-football.com)
+-- Forum: Community (https://forums.sim-football.com/forumdisplay.php?fid=5)
+--- Forum: Media (https://forums.sim-football.com/forumdisplay.php?fid=37)
+---- Forum: Graded Articles (https://forums.sim-football.com/forumdisplay.php?fid=38)
+---- Thread: *The New Voting Method is Better (/showthread.php?tid=32121)

*The New Voting Method is Better - slate - 05-13-2021

Code:
5,525 words, ready for grading.

Awards season, huh? I can barely taste food I eat for the few days after awards because of how much salt is flowing around the ISFL. It seems that a lot of people who are upset that their preferred candidates didn't win awards cast blame on a lot of issues, ranging all the way through the process from how the initial ballots are formed through nominations to how the final votes are tallied and winners are decided. While awards are ultimately a fun way to pass some offseason time and a nice way to recognize users and players that didn't win their respective league championships, they are also something that a lot of people get emotionally invested in and I think it's important to strive to improve the system to minimize the negative feelings that arise. While some of these may be inevitable, general confidence in the process used to arrive on award selections is important so that arguments can ideally be limited to sports banter rehashing stats and arguing about the relative importance of this stat vs. that one.

However, one piece of the process has been receiving some ire from the community as of late which I will adamantly defend, and this is the Instant-Runoff Voting (IRV) system implemented starting with the S26 postseason awards. While I will definitely acknowledge that this is a relatively confusing way to go about counting ballots and that this opacity can exacerbate some of the frustration at the awards process, I strongly believe that this method is far better than the previous Heisman style voting method, which some are hoping to return to as a rules proposal to revert the S26 decision to change systems is once again on the ballot this season. IRV leads to more fair outcomes that better reflect the preferences of the voters who submit ballots, while being less gameable by biased voters who are trying to ensure that their preferred candidate wins at all costs.

Of course, as I just said that this is a complex system and these advantages are not at all obvious on its face. So I decided to write a media piece to come to the defense of the current IRV system and help equip the general userbase with better knowledge about how it works and its advantages, and more trust that the actual results reflect the preferences of the submitted ballots even when it can sometimes seem counterintuitive. I'll first discuss how each method works in detail, then go over advantages of the IRV system and give examples where these came up in recent awards, then maybe close with some general thoughts on the awards system in general.

Let's get into it!

The Previous Method

Prior to S26 (and potentially after S28 if the rules proposal to repeal IRV passes this offseason), a Heisman style voting system was used to decide awards. Each voter would rank their top 3 preferences, and they would each be assigned a point value. I am almost positive that first place would get 3 points, second place 2 points, and third place 1 point, but I didn't try to find a source that stated exactly that so it's possible I'm wrong. Then, quite simply, the total points for each nominee would be added up and the winner would be the candidate with the most total points across all ballots.

I learned while making this article that this is apparently a type of positional voting method. So that's cool that it has a name, I guess! Not really much else to explain about the mechanics of this one I think, so let's move onto the more complex one.

How IRV Works

Instant runoff voting is often called ranked-choice voting when implemented in US political systems, so my fellow American readers may recognize it better with that name as it's been a bit of a fad recently to attempt to move local and state elections over to this system.

In its most general terms, an instant-runoff system is one where voters rank a subset of candidates in a preference order. Once all the ballots are collected, the number of 1st place rankings for each candidate is tallied to determine how many people ranked each candidate as their most preferred. If no candidate is the most preferred on a majority of ballots, then the following procedure repeats iteratively:

Starting at the bottom, the candidates with the fewest 1st place votes are eliminated from the pool of candidates. Any ballots who ranked the eliminated candidate as their most preferred, rather than being discarded as in a normal first-past-the-post system or giving partial credit to a secondary candidate as in the older positional voting system, "pass on" their 1st place preference to their 2nd place candidate, adding 1 to their vote tally. This repeats until the vote tally of any candidate reaches the majority threshold needed to win outright.

In developing an intuitive understanding of this system, I think it's important to deemphasize the idea of 1st place vs. 2nd place vs. 3rd place votes, and instead think in terms of votes as expressions of preferences. If a voter ranks a candidate 1st, that means they would prefer that candidate to win over any other. If they rank a candidate second, that means that they would prefer that candidate wins over any other aside from their 1st place vote. And so on. The general idea behind this method is to whittle down nominees with weaker support until a majority of voters prefer the winner to any other candidates remaining in the pool, where the ones who are remaining are likely to be those that rank highly in peoples' preferences (since those without many 1st place votes will be eliminated early).

IRV Example: ISFL Defensive Rookie of the Year

[Image: 0wc7CgE.png]

First, each individual ballot is counted and 1 vote is allocated from each ballot to the player ranked 1st on that ballot. This gives us the initial count:

[Image: 1J8YfVL.png]

As no candidate has reached the majority of ballots needed to win the award, we then proceed to the iterative process of eliminating candidates and transferring their votes to lower preferences. In this specific example there's a round where Kevin Morrison is eliminated due to 0 1st place votes and thus no ballots are passed on, but after that since no one will have reached a majority (as nothing changed), this process repeats and Evan Jones is eliminated. The ballot with Evan Jones in 1st place was Arizona, whose 2nd highest preference was Jonathan Shuffleboard. So Evan Jones's 1 vote is given to Shuffleboard:

[Image: DX4Srua.png]

And since we still have not reached a majority winner, the lowest vote total candidate Bean Beanman is eliminated in this round. Bean Beanman's votes came from Honolulu, who ranked Marlon Brando 2nd, Orange County, who ranked Marlon Brando 2nd, and Yellowknife, who once again ranked Marlon Brando 2nd. So all 3 of Beanman's votes are transferred to Brando, who is now the most preferred candidate among those remaining for 10 of the 14 voters and thus wins through a clear majority:

[Image: 9nxJSiR.png]

This was a relatively straightforward example in which neither the initial tally nor the final result were particularly close, but what if all of Beanman's voters had ranked Shuffleboard as their second preference? In that case the voting process would have come down to 7 voters who preferred Marlon Brando to Jonathan Shuffleboard (all of whom ranked Brando as their top preference), and 7 voters who preferred Jonathan Shuffleboard to Marlon Brando (of which 3 expressed that he was their 1st choice, and 4 had him as their 2nd choice after a candidate who had already been eliminated). In that case, it would come down to a tiebreaker.

Sidebar: Tiebreaker Explanation

The tiebreaker that the ISFL has decided to use is as follows: the candidate who receives more 2nd place votes from the supporters of the opposing candidate is the one who wins. This can be counterintuitive, but I believe the idea behind it is motivated by trying to reach the greatest consensus. If it's come down to a tiebreaker, the process has found that there's an equal number of voters who prefer candidate A to candidate B and who prefer candidate B to candidate A. So we look at how many of A>B voters have B ranked closely behind A as a measure of how acceptable B is to the first group, and how many B>A voters have A ranked closely behind B as a measure of how acceptable A is to the second. One of those numbers being higher means that that candidate is more acceptable to those who prefer the other, and so that candidate has the greatest consensus behind them and thus deserves the win.

To further clarify this point, when I say "2nd place votes" in the above paragraph I really mean "immediately subsequent preference", or basically the person who is placed right under the candidate in question on voters' ballots. An example of this is the DSFL safety of the year vote, where 'Captain' John Price wins a tiebreaker over Kelvin Harris because Minnesota and Norfolk ranked Price 3rd after Kelvin Harris 2nd. The fact that Price immediately follows Harris (2 -> 3) is the important fact here, not where the votes were cast on the original ballots.

IRV Tiebreaker Example: Altered ISFL DRotY

Back to our DRotY example, let's doctor some votes to create this scenario by swapping some votes between Brando and Shuffleboard (marked in yellow), as well as making some minor alterations by removing Shuffleboard from a few Brando voters' ballots (marked in red):

[Image: OUTTyaW.png]

In this case, the 3 votes for Bean Beanman from HON, OCO, and YKW all pass on to Jonathan Shuffleboard, who is preferred by 7 teams to Brando while 7 teams prefer Brando to Shuffleboard. Neither has reached a majority, we proceed to a tiebreaker. In this tiebreaker we find that of voters who prefer Brando, three have Shuffleboard immediately following him in 2nd place (AUS, BER, and PHI). Meanwhile, only two teams who prefer Shuffleboard to Brando have Brando immediately following (NOLA, who ranks JS 1st and Brando 2nd; and OCO, who ranks JS 2nd and Brando 3rd). Shuffleboard wins in this purely hypothetical example!

Now at its face this result might seem absurd. One player got 7 1st place votes out of 14 but lost to someone who got only 3. By removing Shuffleboard from NYS and SAR's ballots, we've also arrived at a scenario where Brando would win narrowly under the old voting method, 25 to 24, even though these changes didn't significantly change any measure of voter's preferences between Brando and Shuffleboard. However, I would argue that Shuffleboard has a strong claim to be the consensus winner under this hypothetical voting pattern, placing in the top 2 of 10 ballots compared to only 8 for Brando. Furthermore, Brando's only top 2 placing that isn't a 1st place vote comes from a team who prefers Shuffleboard to Brando (NOLA). The fact that it came down to a tie means that we can definitively say that the same number of voters preferred Shuffleboard to Brando as preferred Brando to Shuffleboard, so having a greater consensus seems like a good way to resolve that tie. Once again, the fact that Brando has more 1st place votes isn't weighed as heavily under the IRV system as understanding how many ballots prefer each of the remaining candidates to each other.

Of course, the difficulty of voting theory and deciding on voting systems is that it's really impossible to say who "deserves" to win. However, I think that the weight that IRV gives to consensus building makes it suitable for our purposes as we are unlikely to see a candidate who is highly divisive among voters win despite only a few voters favoring them strongly, which would definitely lead to more drama and complaints than anyone wants to deal with. It's also the case that IRV systems either certainly or with high likelihood lead to several desirable outcomes in terms of voter behavior and the properties of winners. Now that we hopefully have a stronger understanding of the mechanics of the system itself, let's get into some of these examples!

Majority Criterion

One method of comparing different voting systems is to establish criteria that are desirable to have in a voting system, and testing whether each system meets these criteria. One such criteria that has been established is the majority criterion, which states:

Quote:if one candidate is ranked first by a majority (more than 50%) of voters, then that candidate must win.

While I was just espousing the value of having a voting system that emphasizes consensus building and finding common ground among ballots in our tiebreaker example above, I also believe that violation of this criterion may take that a bit too far. When two candidates are equally preferred among voters, it makes sense to try and find whether one has a stronger consensus. But if one is the outright favorite of a majority of voters, it seems cut and dry that they should win if we're looking for the majority opinion.

It's easy to see that IRV satisfies this criterion, as checking if any candidate has a majority of 1st place votes and letting them win if so is literally the first step in the process I described above. However, the Heisman-style voting method fails it. While it's not the most likely thing in the world, it's possible that the votes fall in such a way that a candidate with a majority of first place votes loses.

As an example, say that 8 of the 14 ballots rank A first, the remaining 6 rank B first, and then every ballot with A first ranks B second while every ballot with B first doesn't rank A in the top 3. In this scenario we have A with 24 points from its 1st place votes, and B with 18 points from 1st place votes plus 16 from 2nd place votes for a total of 34 points. In the end this is a pretty lopsided win from a candidate that would lose 6-8 in a straight vote between it and the other frontrunner. Again, switching to IRV helps avoid outcomes that make little sense like this.

And speaking of comparing two candidates in a head-to-head vote...

Finding Condorcet Winners

A Condorcet winner is a fancy math term that means a candidate who beats every other candidate if they are matched up in head-to-head votes with each other. It is generally thought that successfully "finding" a Condorcet winner if one exists is a sign of a good voting system, as otherwise what can happen is picking a winner that would lose in a head-to-head vote against one of the other candidates even if that vote was held amongst the same pool of voters! So the criterion could be expressed as:

Quote:If any candidate is a Condorcet winner among the voters, then that candidate must win.

To give an example of what it means to be a Condorcet winner, we can look to our previous example where A has a majority of 1st place votes (8). This automatically makes them the Condorcet winner as they will win any head-to-head vote at a minimum of 8-6 regardless of who the other 6 ballots prefer. However, a Condorcet winner may still exist in less obvious scenarios.

The ISFL GM of the Year voting from this past season is a good example of this. Here are the ballots (simplified to team name as opposed to GM names for better readability):

[Image: T5mDg0q.png]

If we conduct a head-to-head vote between Berlin's GMs and those of every other nominated team using these ballots, we find the following:

Berlin vs. Austin:
BER 11 (Bex, Darkness, Steg, AUS, ARI, COL, NYS, PHI, SAR, SJS, YKW)
AUS 8 (Mooty, Swanty, BAL, BER, CHI, HON, NOLA, OCO)

Berlin vs. Arizona:
BER 10 (Bex, Darkness, BAL, COL, HON, NOLA, OCO, PHI, SAR, SJS)
ARI 9 (Mooty, Steg, Swanty, AUS, ARI, BER, CHI, NYS, YKW)

Berlin vs. New York:
BER 15 (Bex, Darkness, Steg, Swanty, AUS, ARI, COL, HON, NYS, NOLA, OCO, PHI, SAR, SJS, YKW)
NYS 2 (BAL, CHI)
Undecided 2 (Neither Mooty nor BER ranked either Berlin or New York in their top 3, so we can't know which they preferred between the two)

Berlin vs. Yellowknife:
BER 12 (Bex, Steg, Swanty, ARI, BAL, HON, NYS, NOLA, OCO, PHI, SAR, SJS)
YKW 6 (Darkness, Mooty, AUS, BER, COL, YKW)
Undecided 1 (CHI)

Berlin is the winner in every matchup here, although as might be expected in a season with so many great options for the vote there are some close calls with Austin and Arizona. Although the voters on each side are different each time (BAL voted for Austin over Berlin, but for Berlin over Arizona, for instance), the result of the vote would be the same. In this case, Berlin is the Condorcet winner, and if teams and GMs had been asked to vote between every pair of candidates the final result would be identical to that found by IRV. Good job, IRV!

With the Heisman style voting system, the same result would occur actually:

BER 32
AUS 28
ARI 25
YKW 22
NYS 7

So once again I've been able to find a case where Heisman voting would yield the same result as IRV. However, if we again alter the scenario just slightly by changing a handful votes that don't affect anyone's preferences between Berlin and other teams we may find a different outcome.

[Image: bpoGqHz.png]

Here I've added 5 points to Austin's Heisman total by changing only votes that ranked Austin's GMs under Berlin's. So all of the head-to-head results from earlier are preserved, but now Austin wins 33 to 32 under the Heisman system. However, because every voter whose vote I changed had Berlin preferred to Austin, the IRV result would be identical.

Sadly, unlike the majority criterion which is satisfied by IRV, this system is not guaranteed to successfully find the Condorcet winner. Obviously through the example I just demonstrated, it should be clear that the Heisman system is not either. However, although it is not guaranteed, IRV should have a relatively high success rate at finding the Condorcet winner if one exists, thanks to its focus on binary preferences. The winner of the final runoff round when the vote is narrowed to 2 candidates should always be preferred in a head-to-head vote among voters compared to the loser of that round unless there are a large number of ballots which rank neither in the top 3, so the risk is mainly just that a Condorcet winner may be eliminated earlier if it does not have a high number of 1st place votes.

Meanwhile, the previously used Heisman system does not consider preferences at all, which makes it ill-suited for meeting this criterion. We can reuse our previous example in the discussion of the Majority criterion to demonstrate this. It may be the case that voters who prefer one candidate to another but still give the lower candidate a high rank may assist their less preferred candidate in accumulating enough secondary and tertiary points to win despite losing a straight head-to-head comparison with the winner.

Strategic Voting

One final thing to talk about isn't exactly a voting system criterion like the other two, but is still a goal when deciding upon a voting system to use. Strategic voting is the practice of using your vote not as an expression of your true preferences among the nominees, but instead tactically deciding upon how to vote to maximize the chances of your most preferred candidate winning even if doing so is at odds with your true preferences. Some examples which I can confidently say happen frequently in the ISFL are GMs putting one of their own players 1st even if they don't believe they truly deserve to win the award because it gives them the best chance of pulling off a big upset (although this is arguably a reflection of preferences... more on this later), or a GM leaving the strongest competitor to one of their players off the ballot entirely to avoid giving any assistance in defeating their preferred candidate.

Ideally we should strive to have a voting system which discourages these types of strategic votes, or at the very least minimizes their impact in deciding the final winner. There are separate criticisms of the voting process that could be directed at having GMs with vested interests in certain players winning being the ones doing the voting, but the truth is that regardless of who votes there will always be reasons for people to strategically vote and a good system will be resilient to that.

Let's take a look individually at each of the types of strategic voting I described earlier that are almost certainly practiced by ISFL GMs, and how each voting system handles them:

1. Ranking an undeserving candidate first

In this situation, the concern is that a GM who has a player up for an award might choose to put that player higher than an unbiased voter would otherwise rank them (for example, 1st), pushing down the true contenders for that award to lower places on the ballot which might impact their voting. In particular, the concern here is that this player isn't really a contender for the award at all and has no real chance of winning. Especially if some GMs practice this and others don't, it might lead to a situation where a preferred candidate loses even if the GMs that ranked their pet candidate 1st would otherwise prefer that candidate.

Depending on the specifics of the situation under the Heisman system, this may or may not have an impact. If A and B are the two frontrunners for the award and I prefer A > B, whether I rank A 1st and B 2nd or C (my pet candidate) 1st, A 2nd, and B 3rd would make no difference to the ultimate result as I am simply subtracting a point from each. However, this can come up if a voter has B nowhere close to A, but changes from a rank of A 1st C 2nd D 3rd to a rank of C 1st A 2nd D 3rd. This costs A one point, but doesn't change B's total at all, which has the potential to cost A a close race even though this voter vastly prefers A to B and both ballots reflect this. Alternatively, it could matter if a voter is not quite bold enough to rank C 1st, but instead bumps them up from 3rd place to 2nd place (A 1st B 2nd C 3rd -> A 1st C 2nd B 3rd), costing B a point in this situation despite the relative preference of A vs. B not changing.

Under IRV, it is actually quite difficult for this practice to matter. Let's take the 2 examples above. In the first, where I don't have B on my ballot at all, C will be eliminated and my vote will pass on to A. Thus my ballot will end up counting as a vote for A even if I do exhibit slight bias by ranking C 1st over A. Similarly in the 2nd, unless A is otherwise eliminated (at which point it is likely B has already won), my ballot will end up counting only in A's favor regardless of where exactly I rank B underneath him. That is, unless it comes down to a tiebreaker, at which point the fact that I had someone besides B as my 2nd place vote does matter, but this requires a pretty specific set of circumstances. It is clear that IRV is more resilient to this type of strategic voting than the Heisman method.

2. Avoiding votes for the strongest competitor to a favored candidate

Let's say A and B are in a tight race for an award in the ISFL. I am a voter/GM and B is on my team, and it would be really cool if he won it, so I decide to come up with some questionable reasons for why A is worse than C and D (maybe something along the lines of deciding to focus only on 1 stat for a position and ignoring someone who had a great year in some other stats I choose not to care about), and submit a ballot with B ranked first, C second, and D third. Meanwhile, the rest of the ballots are as follows:

[Image: UxnBlV4.png]

Under the Heisman system, my subterfuge pays off. A has 5 1st place votes (15 points), 7 2nd place votes (14 points), and 1 3rd place vote (1 point) for a total of 30 points, whereas B has 7 1st place votes (21 points), 4 2nd place votes (8 points), and 2 3rd place votes (2 points) for a total of 31 points. If I had ranked B's main competitor A even in 3rd place, they would have tied, and if I had ranked them in 2nd place I would have cost my favorite candidate the award, despite none of those rankings changing my ballot's expressed preference between B and A. Note also the interplay with the point above - the fact that Teams 13 and 14 ranked their favorite candidates C and D 1st cost A points and therefore the win even though they both prefer A to B.

In the current IRV system however, my subterfuge has no impact on the final result. Candidates C and D are eliminated for having fewer 1st place votes than A and B, and Teams 13 and 14 have their votes transferred to candidate A. Candidates A and B both have 7 votes, so it proceeds to a tiebreaker, where the fact that Teams 12 and 13 have a candidate other than B as their 2nd favorite behind A gives A the win.

I did have to specifically engineer this example so that the tie was broken in a way that didn't depend on my preferences, though. The fact is that this form of strategic voting can have an impact if the vote is otherwise exactly tied. However, this requires not only a tie vote but also that the tiebreaker would be decided by whether or not I rank the opposing candidate immediately after mine. This is quite a specific set of circumstances, however it is possible that it can happen. An example where this could have happened is the DSFL TEotY vote this past season, where BBB ranked Marston 2nd and DAL ranked Bronko 2nd. If either team had decided to strategically vote and push the other candidate down just 1 spot into 3rd it would have decided the award in their favor. So props to them for voting honestly!

I believe this actually corresponds to a voting criterion of "later-no-harm," which states that:

Quote:if a voter alters the order of candidates lower in his/her preference (e.g. swapping the second and third preferences), then that does not affect the chances of the most preferred candidate being elected.

Wikipedia says that this criterion is satisfied by IRV, but I believe that the tiebreaker we use violates it. Another fun fact from Wikipedia is that the later-no-harm criterion is incompatible with the Condorcet winner criterion from earlier; it is impossible for a system which satisfies later-no-harm to also satisfy the Condorcet winner criterion. I believe that preventing strategic voting by satisfying later-no-harm is probably more important to us given the fact that GMs are voters, so I am good with using a voting system which emphasizes this aspect. (And again, Heisman voting satisfies neither.)

All in all, while strategic voting can have minor influences at the margins, its impact on awards is almost sure to be lower under the IRV system than Heisman voting. This is because, with the minor exception of the tiebreaker we use, you never give any points to any other candidate besides your most preferred one as long as your most preferred one is still in contention. In Heisman voting, your 2nd and 3rd place rankings will always affect the final standings regardless of whether your 1st place vote has a chance of winning or not, which can create perverse incentives for voters.

Disadvantages of IRV, and Other Thoughts on Awards Issues

Around awards time there is always a lot of media that comes out and advances one side of an argument without considering the opposition's points at all. I want to try to avoid this trap, however of course it can always be difficult to know all of the other side of an argument before you make it so please forgive me if you have a specific grievance that I don't address. Besides the complexity argument, which I've hopefully addressed some with the explanations provided in this article, I can see a few downsides to using IRV.

One is the tiebreaker that we use. As I said above, the tiebreaker means that we fail to satisfy some voting system criteria that IRV is supposed to satisfy, like later no harm. This opens us up to strategic voting being influential, and also can lead to more counterintuitive results where a candidate with lower 1st-place support can lose via convoluted tiebreakers in a very close race. I would propose that the number of ballots who rank a candidate 1st overall could be a good first tiebreaker to use, before proceeding to the current tiebreakers if that does not break a tie by itself. This would change the result of our altered Rookie of the Year race from early in the article (although I argued in favor of the consensus candidate winning there, I think it's also defensible to give the candidate with more vehement support the win), as well as the ISFL Punter of the Year race this season which has received a lot of scrutiny since the awards show. The downside of this is that it opens the system back up to being influenced by voters placing an undeserving candidate in 1st, so it's not a strict improvement. Maybe we should just be trying to find a method that results in an odd number of voters to reduce the chances of ties altogether?

Another issue that some people might have with it is its emphasis on head-to-head preferences which might lead to differences in results with Heisman voting in close races. A candidate that wins an ISFL award 8-6 among final preferences has a pretty narrow lead, so maybe it does make sense to consider how highly that candidate ranked in everyones' preference lists even if it might lead to a winner who would lose in a head-to-head vote. There are definite tradeoffs, because a candidate who everyone thinks is a frontrunner would be more likely to win, but it could also lead to a candidate who some users favor heavily losing despite getting a majority of 1st place votes. Depending on which side of the argument someone happens to be on one season, they could change their mind the next with a different fact set, so I think it's worth prioritizing things like reducing strategic voting and finding Condorcet winners more often which are pretty obviously good goals to have.

Finally there are some assumptions behind my explanations here that may be undermined by the surrounding conditions of the voting process like eligibility. Regardless of how well the voting system is designed, the fact is that having GMs as the voters means that their preferences may be distorted thanks to having vested interest in certain players and more familiarity with certain teams. I've been assuming that the preferences reflected in votes are a good indicator of who is deserving to win an award, but even ignoring the potential for strategic voting the fact is that GMs' preferences of who wins an award may be far out of line from those of the awards watchers who pay careful attention to stats and care a lot about who wins. Perhaps changing who has votes could be a method to address this issue, but the truth is that there are no impartial, ultra-enlightened users in the ISFL and awards will always be a matter of contention no matter what system is used or who is eligible to vote. Another important piece of this is to continue to keep GMs accountable for their votes (such as through Modern Duke's article series breaking down biased voting by team) so that GMs have incentives to try to vote honestly rather than in their own interest.

And on that happy note, I think I'll conclude the article here! I'm really interested to hear from everyone whether this article was useful in helping you understand the system and its advantages/disadvantages, if there are any issues you have that I haven't addressed, or if you have any questions or hypothetical examples you want me to take a stab at (although as my citations of Wikipedia might indicate I'm really not particularly advanced in my understanding on this topic). Thank you so much for reading!

RE: The New Voting Method is Better - Kotasa - 05-13-2021

This method made me lose therefor it is bad thank you

Jk this is good shit I appreciate you

RE: The New Voting Method is Better - mithrandir - 05-13-2021

Well, I'm convinced. Keep IRV!

RE: The New Voting Method is Better - Jimi64 - 05-13-2021

This is the federalist papers of the ISFL awards system.

This was fascinating, specifically the Condorcet winner section.

Great job Slate

RE: The New Voting Method is Better - Memento Mori - 05-13-2021

I was happy when I saw the title because I prefer IRV, but I'm really happy now that I've gotten to read a nerdy voting systems article.

RE: The New Voting Method is Better - infinitempg - 05-13-2021

ho lee fuk

great article! really hits the points of why IRV is advantageous over the heisman style we had before, and i generally agree with all of it

that being said, it's easier to point at the sheet with heisman scoring and say "this is why you lost" and therefore less anger in the league in general. also it's easier to program and check over

i love the instant runoff we have now and definitely wouldn't mind keeping it, but i have proposed going back and will continue to propose it because i think it might be able to reduce some of the tensions with respect to awards season

now of course the best way to get rid of awards tension is to ABOLISH AWARDS but no one wants to hear that

RE: The New Voting Method is Better - 37thchamber - 05-14-2021

This is good content.

Out of curiosity, I put all the ballots through a Schulze (beatpath) method calculator (which I unashamedly stole from somewhere years ago), since that's my preferred voting system. I like it because it shows pairwise preferences (i.e. who preferred A to B and vice-versa) in a matrix, so its usually easier to see that someone was preferred by all voters (condorcet winner), and tbh, that's what matters most to me. That's the clearest indicator of "clear preference" imo.

Anyway, turns out, most results are the same. Only difference I found was in the Kicker of the Year voting.

IRV puts Danny King over McDairmid on the basis of 3rd place votes. But, Schulze ties McDairmid and King (breaking down the pairwise matrix, I can see that each is preferred over the other directly 7 times; basically, if I count the instances in which McDairmid appears higher than King in the ballot order, there would be 7, and vice-versa) so it throws a random ballot out to determine final order. You can make the case that IRV has a more logical tiebreak method in this case, since it's easily quantifiable in a "points" methodology (assuming 3pts for 1st, 2 for 2nd etc, King has more points), though I'd argue that the actual votes show there is no clear preference among the voters, and whatever you use as a tiebreaker will be controversial.

Someone mentioned that seeing the first ranked name on half the ballots not win is a sign of a problem, but I'd argue that's not true. 50% isn't a majority. If the other 50% of voters prefer someone else over that name, then there is no clear preference between the two, surely? If they were first ranked on the majority of ballots, I'd understand the complaint (and also note that this wouldn't happen under a system that meets the condorcet criterion) but that was not the case. Perhaps this is a flaw with the voting system in that it's possible to have exactly 50% of ballots show someone as first choice. A fifteenth ballot would have precluded this possibility entirely.

Anyway, interesting stuff.

RE: *The New Voting Method is Better - mee - 05-19-2021

I should have just received all 14 first-place votes. It would have made everything easier.