A profile of 100 ranked ballots, and the majority in every pair
ballot is one function. Everything below came out of it during this
build, at parameters taken from the essays rather than invented for this page — so a figure
here is the same figure a reader meets in an essay, and if the generator changes, this page
changes with it.
With nothing chosen
Everything that beats one proposal
On a line the median cannot be beaten
A balance that one small shift destroys
A chain of majorities that leads anywhere
The smallest circle that meets every median line
What it checks while it draws
Collected by running the family and recording what it asserted, not written here. The count is how many separate times the claim was put to the test while these drawings were made.
- voter 1 ranks A and B the same way in both profiles ×5
- the number of candidates is a whole number between 2 and 5 ×4
- the number of judges is a whole number between 3 and 6 ×4
- the total for 000 is the minorities plus the margins it overrides ×4
- each ballot ranks all 3 of them exactly once ×3
- the 3-voter profiles fall into one class per pattern of the pair ×3
- the space of 3-voter profiles was built entire ×3
- B wins all 4 of its pairs ×2
- each candidate of the pair is a whole number between 0 and 2 ×2
- every one of the 64 profiles is examined ×2
- the largest jury is a whole number between 21 and 1001 ×2
- the manipulating voter's true ranking ranks all 3 of them exactly once ×2
- there are 12 majority patterns on 5 candidates up to relabelling ×2
- A ≻ B ≻ D ≻ C is not an improvement and is not marked as one ×1
- A ≻ C ≻ B ≻ D is not an improvement and is not marked as one ×1
- A ≻ C ≻ B is not an improvement and is not marked as one ×1
- A ≻ C ≻ D ≻ B is not an improvement and is not marked as one ×1
- A ≻ D ≻ B ≻ C is not an improvement and is not marked as one ×1
- A ≻ D ≻ C ≻ B is not an improvement and is not marked as one ×1
- A beats B by exactly two ×1
- a cyclic electorate has its first two margins of one sign ×1
- A does not beat everybody ×1
- A has more Borda points than B ×1
- A has more Borda points than C ×1
- A has more Borda points than D ×1
- A has more Borda points than E ×1
- A has more first places than B ×1
- A has more first places than C ×1
- A has more first places than D ×1
- A has more first places than E ×1
- a mixed jury is no worse than a uniform one at the same average ×1
- a profile is between one and eight voter groups ×1
- a rational is a whole numerator over a non-zero whole denominator ×1
- A really beats B ×1
- a score is never smaller than the votes still to be won ×1
- a single judge finds for the conclusion in a quarter of profiles ×1
- a single vote moves at most twice the yolk's radius further out ×1
- a single voter is right with the average competence ×1
- A wins all 4 of its pairs ×1
- above one half, every two voters added make the majority likelier to be right ×1
- after the swaps the Dodgson winner beats every rival ×1
- an average competence above one half ×1
- an odd electorate ×1
- an odd electorate of 21 to 101 ×1
- an odd jury, so a majority always exists ×1
- an odd number of judges ×1
- an odd number of judges, so majorities are decisive ×1
- an odd number of judges, three to seven, each with a yes or no on three questions ×1
- an odd number of voters ×1
- an odd number of voters leaves no tie ×1
- an odd panel of three to nine voters, each better than a coin ×1
- and 2 of the 56 ways of counting the ballots ×1
- and becomes reliable ×1
- and every one is a dictator or an oligarchy ×1
- and it is silent on more profiles than the majority quota is ×1
- and it is unanimity on all three ×1
- and it rises towards that ceiling with every two voters added ×1
- and most do not ×1
- and none is both, which is the impossibility this agenda exhibits ×1
- and on some that are not aligned, so alignment is sufficient and not necessary ×1
- and some are consistent ×1
- and some do not, so neither procedure is always wrong ×1
- and stays below Guilbaud's limit ×1
- and the aligned rule gives up universal domain ×1
- and the drawn profile really is one of them ×1
- and the more voters share it, the rarer cycles are ×1
- and the premise vote better ×1
- at 51 voters each rule is within three points of its limit ×1
- at 51 voters runoff beats Borda beats plurality beats anti-plurality ×1
- at every size the majority's accuracy is under the chance that the shared competence exceeds one half ×1
- averaged over the worlds, enough correlation turns the comparison round ×1
- B ≻ A ≻ C ≻ D elects B, which the voter ranks above D ×1
- B ≻ A ≻ C elects B, which the voter ranks above C ×1
- B ≻ A ≻ C is not an improvement and is not marked as one ×1
- B ≻ A ≻ D ≻ C elects B, which the voter ranks above D ×1
- B ≻ C ≻ A ≻ D elects B, which the voter ranks above D ×1
- B ≻ C ≻ A elects B, which the voter ranks above C ×1
- B ≻ C ≻ A is not an improvement and is not marked as one ×1
- B ≻ C ≻ D ≻ A elects B, which the voter ranks above D ×1
- B ≻ D ≻ A ≻ C elects B, which the voter ranks above D ×1
- B ≻ D ≻ C ≻ A elects B, which the voter ranks above D ×1
- B beats C by exactly two ×1
- B beats D by exactly two ×1
- B beats every other candidate head to head ×1
- B does not beat everybody ×1
- B has more Borda points than A ×1
- B has more Borda points than C ×1
- B has more Borda points than D ×1
- B has more Borda points than E ×1
- B really beats C ×1
- below one half, every two added make it likelier to be wrong ×1
- Borda can be flipped by moving a third candidate somewhere in this range ×1
- Borda puts A above B in the first profile ×1
- Borda puts B above A in the second ×1
- Borda's agreement is a share of the electorates counted ×1
- C ≻ A ≻ B ≻ D is not an improvement and is not marked as one ×1
- C ≻ A ≻ B is not an improvement and is not marked as one ×1
- C ≻ A ≻ D ≻ B is not an improvement and is not marked as one ×1
- C ≻ B ≻ A ≻ D elects B, which the voter ranks above D ×1
- C ≻ B ≻ A is not an improvement and is not marked as one ×1
- C ≻ B ≻ D ≻ A elects B, which the voter ranks above D ×1
- C ≻ D ≻ A ≻ B is not an improvement and is not marked as one ×1
- C ≻ D ≻ B ≻ A is not an improvement and is not marked as one ×1
- C beats A by exactly two ×1
- C beats D by exactly two ×1
- C does not beat everybody ×1
- C really beats A ×1
- C really beats D ×1
- chain: the pruned search finds exactly what the plain one does ×1
- conj: the pruned search finds exactly what the plain one does ×1
- Coombs can be flipped by moving a third candidate somewhere in this range ×1
- Coombs eliminated exactly one candidate in every round but the last ×1
- Coombs puts A above B in the first profile ×1
- Coombs puts B above A in the second ×1
- Coombs reaches a winner without a level elimination ×1
- Copeland's agreement is a share of the electorates counted ×1
- correlated errors help the verdict vote ×1
- counting the three-way cycles gives the same answer as looking for a winner ×1
- cycles are commoner among close contests than among lopsided ones ×1
- D ≻ A ≻ B ≻ C is not an improvement and is not marked as one ×1
- D ≻ A ≻ C ≻ B is not an improvement and is not marked as one ×1
- D ≻ B ≻ A ≻ C is not an improvement and is not marked as one ×1
- D ≻ B ≻ C ≻ A is not an improvement and is not marked as one ×1
- D ≻ C ≻ A ≻ B is not an improvement and is not marked as one ×1
- D ≻ C ≻ B ≻ A is not an improvement and is not marked as one ×1
- D beats A by exactly two ×1
- D beats every other candidate head to head ×1
- D does not beat everybody ×1
- D really beats A ×1
- D wins all 4 of its pairs ×1
- dictatorship of judge 1 gives up at least one condition ×1
- dictatorship of judge 1 is consistent wherever it is defined ×1
- dictatorship of judge 1 is defined somewhere ×1
- E does not beat everybody ×1
- each candidate's Borda score is the number of head-to-head votes it gets ×1
- each covariance entry is the exact covariance of one random ranking's verdicts ×1
- each extra candidate makes a missing winner likelier ×1
- each judge is consistent ×1
- each jury size is the smallest odd one that reaches the target ×1
- each size has a crossover, and larger courts cross earlier ×1
- each voter group carries a count and an order ×1
- each voter group's size is a whole number between 1 and 500 ×1
- electorates at each size is a whole number between 50 and 800 ×1
- enough electorates without a Condorcet winner at every size ×1
- enough electorates without a Condorcet winner to count ×1
- every assignment of positions to judges is considered ×1
- every ballot the voter could submit produces a single winner ×1
- every class holds the same number of profiles ×1
- every consistent rule leaves some proposition undecided ×1
- every inconsistent majority on this agenda has the same shape ×1
- every judge's own three answers hang together ×1
- every member of the family shows Anscombe's paradox ×1
- every pattern is produced by one or three voters ×1
- every profile in the space was decided one way or the other ×1
- every ranking the voter could submit was built ×1
- every rule misses the Condorcet winner in some electorate of at most 21 ×1
- every survivor gives a consistent verdict here ×1
- every survivor is consistent on every profile ×1
- every systematic anonymous rule on this agenda is considered ×1
- every troubled profile lands in one of the four outcomes ×1
- every voter is counted once ×1
- every voter prefers the start to the target ×1
- exactly one pattern — the transitive order — needs a single voter ×1
- for a large court the conclusion vote switches at √½ ×1
- for three judges and two premises the crossover is the square root of a half ×1
- for three judges the average difference is (3/2)p²(1 − p)²(2p² − 1) ×1
- for three judges the crossover is exactly √½, where (3/2)p²(1 − p)²(2p² − 1) vanishes ×1
- four of the six rankings are single-peaked on the axis A, B, C ×1
- free: the pruned search finds exactly what the plain one does ×1
- in the liable world the premise vote is never the worse ×1
- instant runoff can be flipped by moving a third candidate somewhere in this range ×1
- instant runoff eliminated exactly one candidate in every round but the last ×1
- instant runoff puts A above B in the first profile ×1
- instant runoff puts B above A in the second ×1
- instant runoff reaches a winner without a level elimination ×1
- its average majority is 3m/(4m + 1), below three quarters ×1
- larger sets find for the conclusion no more often ×1
- majority is consistent on every aligned profile ×1
- majority is defined somewhere ×1
- majority never fails exactly when no minimally inconsistent set has more than two members ×1
- majority survives on the chain and the free pair and not on the conjunction ×1
- majority, aligned profiles only gives up at least one condition ×1
- majority, aligned profiles only is consistent wherever it is defined ×1
- majority, aligned profiles only is defined somewhere ×1
- more correlation, a less reliable large majority ×1
- more premises never need a smaller court ×1
- no candidate beats every other, which is what having no Condorcet winner means ×1
- no electorate showing Anscombe's paradox has an average majority of three quarters or more ×1
- no misreport helps the voter once only two candidates remain ×1
- not every profile is aligned, which is what makes the restriction a restriction ×1
- of the 216 profiles of three voters, 204 have a Condorcet winner ×1
- on the chain a pair of quotas is safe exactly when the stricter threshold has the stricter quota ×1
- on the conjunction exactly one triple of quotas is safe ×1
- on the conjunction, one survivor for each non-empty set of judges ×1
- on the ranking, the survivors are the dictators and nothing else ×1
- on this profile the majority corner is not a consistent one ×1
- one candidate has strictly the most Borda points ×1
- one candidate has strictly the most first places ×1
- one disagreeing profile is drawn ×1
- one for each judge ×1
- one to eight thousand sampled elections for each number of candidates ×1
- one to six competences between 0.3 and 0.9, none exactly one half ×1
- one to three odd electorates up to 45 ×1
- one to three target accuracies ×1
- party X wins three groups of five and loses every issue three groups to two ×1
- plain majority on every proposition is the rule the impossibility rules out ×1
- plurality can be flipped by moving a third candidate somewhere in this range ×1
- plurality loses ground as candidates are added and Borda stays ahead of it ×1
- plurality puts A above B in the first profile ×1
- plurality puts B above A in the second ×1
- Plurality's agreement is a share of the electorates counted ×1
- premise-based gives up at least one condition ×1
- premise-based is consistent wherever it is defined ×1
- premise-based is defined somewhere ×1
- raising the quota never reduces how often the rule says nothing ×1
- rank: the pruned search finds exactly what the plain one does ×1
- reachability is computed ×1
- rules that need not respond monotonically are searched for three judges only ×1
- so it is at least as accurate as the simple majority and the best voter ×1
- some profiles make the two procedures disagree ×1
- some profiles produce an inconsistent majority ×1
- some rules are complete ×1
- some set of swaps always makes the candidate a Condorcet winner, with enough voters ×1
- spreading the same product unevenly lowers the premise vote ×1
- the agenda is one of chain, conj, rank ×1
- the agenda is one this family knows ×1
- the average voter is better than a coin ×1
- the conclusion vote never convicts wrongly more often than the premise vote ×1
- the Coombs winner ends with a strict majority ×1
- the correlation is one the two competences allow ×1
- the counting floor sits below Stearns's n + 2, which sits below McGarvey's n(n − 1) ×1
- the court has an odd number of judges from 3 to 31 ×1
- the crossover falls as the court grows ×1
- the dictatorship gives up anonymity ×1
- the distinct winners counted by walking the list agree with the set ×1
- the Dodgson winner is unique here ×1
- the electorate is a whole number between 2 and 2000 ×1
- the electorate is odd, so the two-candidate control cannot end level ×1
- the enumerated profiles' probabilities add to one ×1
- the exact arithmetic stays inside the safe integer range ×1
- the exhausted profile space is small enough to give every profile its own cell ×1
- the exhausted profile space stays small enough to walk ×1
- the honest ballot appears exactly once among the submissions ×1
- the honest winner does not depend on the order the ballots are listed in ×1
- the instant-runoff winner ends with a strict majority ×1
- the largest electorate is a whole number between 5 and 51 ×1
- the largest electorate is odd, so no majority is level ×1
- the largest electorate searched is a whole number between 3 and 6 ×1
- the largest number of candidates is a whole number between 50 and 2000 ×1
- the log-odds weighted majority achieves the best accuracy any rule reading the votes can ×1
- the majority between A and B is decided rather than level ×1
- the majority between A and C is decided rather than level ×1
- the majority between A and D is decided rather than level ×1
- the majority between B and C is decided rather than level ×1
- the majority between B and D is decided rather than level ×1
- the majority between C and D is decided rather than level ×1
- the majority on this court contradicts itself ×1
- the majority quota is inconsistent on some profiles ×1
- the majority tournament runs in a cycle through every candidate ×1
- the majority's three answers do not hang together, which is the whole point ×1
- the majority's verdicts are the median judge's ×1
- the margin between A and B is antisymmetric ×1
- the margin between A and C is antisymmetric ×1
- the margin between A and D is antisymmetric ×1
- the margin between A and E is antisymmetric ×1
- the margin between B and C is antisymmetric ×1
- the margin between B and D is antisymmetric ×1
- the margin between B and E is antisymmetric ×1
- the margin between C and D is antisymmetric ×1
- the margin between C and E is antisymmetric ×1
- the margin between D and E is antisymmetric ×1
- the margin covariance is positive definite ×1
- the mixed jury improves as it grows ×1
- the mode is one of profile, cycle, rules, iia, manipulate, judge, agenda, escape, doctrinal, impossible, escapes, quota, aligned, distance, cube, distancesweep, kemeny, agendas, mis, quotak, median, cycleodds, cyclescatter, cyclecands, cyclemodel, jury, jurycorr, juryweights, jurysize, juryhetero, mcgarvey, mcgarveymin, dodgson, dodgsonfreq, mcgarveybound, dodgsonmargins, dodgsontideman, survivors, blocked, oligarchy, vetoprice, quotagrid, truthworlds, truthcurve, truthliable, truthcross, truthprior, trutherrors, premthreshold, premliable, premaverage, premdisagree, premdisjunct, premsize, uchero, ucworlds, ucmap, ucspread, uccorr, ucsize, oshero, osfreq, osanscombe, oswagner, osspread, effhero, effminimal, effthree, effcandidates, effborda ×1
- the most candidates is a whole number between 4 and 12 ×1
- the number of distinct winners lies between one and the number of rules ×1
- the number of electorates is a whole number between 200 and 6000 ×1
- the number of judges is odd, so no question ties ×1
- the number of voters in the exhausted space is a whole number between 2 and 6 ×1
- the number of voters is a whole number between 3 and 15 ×1
- the option is one this view reads ×1
- the ordered pairs inside a class counted two ways agree ×1
- the pair is two different candidates ×1
- the premise-based answer is consistent ×1
- the premise-based rule gives up systematicity ×1
- the profile space has one entry for every assignment of ballots ×1
- the quota is safe exactly when it exceeds n(k − 1)/k ×1
- the ranking is totally blocked and the conjunction is not ×1
- the rule being manipulated is one of plurality, borda, irv, coombs ×1
- the rule to violate is one of plurality, borda, irv, coombs ×1
- the seed is a whole number between 1 and 999 ×1
- the shortcut does not fall behind as electorates grow ×1
- the simulated electorates is a whole number between 200 and 4000 ×1
- the simulated electorates per count is a whole number between 5000 and 100000 ×1
- the simulated share of cycles is within four standard errors of the exact limit ×1
- the two opposed counts for A against B sum to the electorate ×1
- the two opposed counts for A against C sum to the electorate ×1
- the two opposed counts for A against D sum to the electorate ×1
- the two opposed counts for A against E sum to the electorate ×1
- the two opposed counts for B against C sum to the electorate ×1
- the two opposed counts for B against D sum to the electorate ×1
- the two opposed counts for B against E sum to the electorate ×1
- the two opposed counts for C against D sum to the electorate ×1
- the two opposed counts for C against E sum to the electorate ×1
- the two opposed counts for D against E sum to the electorate ×1
- the two profiles really differ on some ballot ×1
- the two rules disagree on this profile, which is what makes it a choice ×1
- the two-candidate contest has a winner whichever ballot the voter submits ×1
- the unanimity quota is inconsistent on none ×1
- the verdict vote depends only on the product of the competences ×1
- the voters in each electorate is a whole number between 11 and 401 ×1
- the walk reaches the target ×1
- the weighted vote is never level on this panel ×1
- the yolk touches or crosses every median line ×1
- there is a share of liable cases at which the two procedures tie ×1
- this profile has no Condorcet winner, so every candidate needs some swaps ×1
- three candidates reproduce Guilbaud's exact limit within four standard errors ×1
- three judges with three yes-or-no answers each ×1
- three to five judges, each holding a consistent judgement set ×1
- three voters at 0.6: 0.648, by hand ×1
- three voters can never produce a party that loses every issue and wins; five can ×1
- three voters cannot produce Anscombe's paradox on three issues; five can ×1
- three voters: 12 of the 216 profiles cycle ×1
- two groups whose shares add to one ×1
- two judges overrule least and all three most ×1
- two to five correlations below 0.9 ×1
- two voters for each of the n(n − 1)/2 contests ×1
- unanimity gives up at least one condition ×1
- unanimity gives up completeness ×1
- unanimity is consistent wherever it is defined ×1
- unanimity is defined somewhere ×1
- under the first model the chance rises with every two voters added ×1
- under the second it stays below one in sixteen ×1
- when every voter shares the axis, no electorate cycles at all ×1
- when only the weak element fails the verdict vote is the more reliable ×1
- where the majority's set is consistent the rule returns exactly it ×1
- while the premise vote is already reliable just above a half ×1
- with five alternatives the verdict vote convicts the innocent and the premise vote does not ×1
- with five premises the conclusion vote almost never finds liability and the premise vote almost always does ×1
- with many premises the procedures disagree on most liable defendants ×1
- with one premise the two procedures are the same ×1
- with perfectly correlated errors the two procedures coincide ×1
- with these competences a larger court makes the verdict vote worse in the liable world ×1
- with three issues the election and the issue majorities disagree in between a tenth and three tenths of electorates ×1
Where it is called
Every figure on this list is drawn by the same rule, so a change to the rule changes all of them at once. That is why the list is published.
A chain of majorities leads anywhere
Put voters on a line and the middle one's position beats everything. Put five in a plane and nothing is safe: seven majority votes can walk the group from its own centre to a place every one of them likes less. One small circle, drawn through the lines that split the voters, says exactly how far a single vote can go, and it shrinks as the group grows.
AppliedA lie that pays
A ballot is usually read as a report of a preference. This one reads it as a move, and walks every move one voter has — all six rankings, the winner each produces, and the ones that beat honesty.
AppliedA majority wiser than its members
Condorcet's other theorem turns voting round: the voters no longer have preferences but judgements about a single fact, each a little more likely right than wrong. Then a simple majority of many of them is almost certainly right — 6,763 voters who are each right 51% of the time make a majority right 95% of the time. The theorem survives voters worse than a coin, if the average is better. It does not survive voters who share their mistakes, and when their skills differ the right rule weighs votes rather than counting them.
AppliedA party that loses every issue can win
Two parties take opposite sides on three issues, and every voter backs the party they agree with on more of them. The party that wins the election can be on the losing side of all three issues — in a fifth of random electorates it loses most of them, and in about one in a hundred it loses every one. A majority of voters can even find themselves outvoted on most of the questions decided. Only when issues are settled by three-to-one majorities on average is that ruled out.
AppliedA share of the votes is not a share of the power
Give three members four, four and one vote, with five needed to pass. Every winning coalition needs exactly two of them, so all three have equal power — and one of them holds a ninth of the votes.
AppliedAgendas that cannot contradict themselves
A court voting on two unconnected questions never contradicts itself, and neither does one voting on a chain of thresholds. A court voting on two premises and their conjunction sometimes does. What separates them is the size of the smallest sets of judgements that cannot all be true: pairs are harmless, because two majorities always share a judge, and triples are not. The same count says exactly how large a supermajority has to be to stay consistent on any agenda.
AppliedDeciding the premises or the conclusion
A body that cannot be both decisive and coherent has to choose which. The two live options are to vote on the reasons and let the verdict follow, or to vote on the verdict and let the reasons look after themselves — and they reach opposite answers on exactly the profiles the impossibility identifies.
AppliedEvery premise makes the verdict vote stricter
A court that must find several things true before it finds liability can vote on each of them or on the verdict. With two elements, judges right seven times in ten are enough for either. With five, the verdict vote needs judges right eighty-seven times in a hundred, or a larger court makes it worse — while the premise vote still needs only better than even. And when any one ground is enough instead of all, the two procedures swap their failures.
AppliedFive rules and five winners
Twenty-seven ranked ballots, five entirely reasonable ways of counting them, and five different candidates declared the winner. Every count is correct, every rule is defensible, and the answer turns out to be a property of the rule rather than of the ballots.
AppliedFour conditions, and no rule that has all of them
Five reasonable rules can return five different winners on one set of ballots, which invites the obvious question of which one is right. The answer is that the conditions anybody would write down cannot all hold at once — and here each named rule's own violation is found by search rather than quoted.
AppliedFour ways out, and what each costs
An impossibility theorem lists conditions and says no rule has them all. That leaves exactly as many escapes as there are conditions, each of them a real institution — a dictator, a two-stage procedure, a supermajority, a restricted agenda — and each escape's price can be counted rather than argued about.
AppliedHow few voters any majority needs
Any pattern of head-to-head majorities whatever — cycles within cycles, a candidate who beats the winner of every other contest and loses to its loser — can be produced by voters who each rank the candidates sensibly. McGarvey's recipe needs n(n − 1) of them for n candidates. The truth is far fewer: every pattern on five candidates takes three voters at most, a counting argument shows the number must eventually grow, and it grows only like n divided by its logarithm.
AppliedHow often the majority goes in a circle
Three voters and three candidates give 216 profiles, and 12 of them are cycles. Count every electorate up to 41 voters exactly and the share climbs towards 8.77%, a number Guilbaud found in 1952 as the solid angle where three half-spaces at the tetrahedral angle overlap. Add candidates and a winner goes missing half the time; let voters share one axis and cycles vanish. The number is always a property of the model of how ballots are drawn.
AppliedHow often the rules find the majority favourite
When one candidate beats every rival head to head, a voting rule ought to elect it. Plurality does about three times in four when every ranking is equally likely, the runoff between the top two 96 times in a hundred, Borda's count 90. Every profile of up to 51 voters is counted exactly, the limit is sampled, and adding candidates pulls plurality down below one half.
AppliedNo rule escapes the doctrinal paradox
A court whose members each hold a consistent position can reach an inconsistent verdict by majority. One such case is easy to build, which invites the hope that a better rule would avoid it — and every rule that responds to the votes at all fails somewhere.
AppliedNone of the four conditions is spare
Four conditions pick out one sharing rule. The half that is usually shown is that they are enough; the other half is that each is needed — drop any one and a different rule satisfies the rest, so the list cannot be shortened.
AppliedThe court that contradicts itself
Three judges each answer three questions, and each answers them consistently. Take the majority on each question separately and the answers no longer hang together — the body as a whole endorses a combination no member of it holds, and no rearrangement of the procedure removes the problem.
AppliedThe element the judges are worst at
A court can vote element by element or on the verdict, and the comparison so far assumed judges equally good at every element. Real judges are surer about the law than about a contested fact. Make one element hard and both procedures sink towards the majority on that element, but they sink differently: the verdict vote sees only the product of the judges' competences and cannot tell an even court from a lopsided one, while the premise vote is ruled by its weakest majority. And when a judge's errors on two elements come from the same misreading of the facts, the gap between the procedures closes.
AppliedThe fewest swaps to a winner
When no candidate beats every other head to head, Charles Dodgson proposed in 1876 to elect the one that is closest to doing so — the candidate that the fewest swaps of neighbouring names on the ballots would turn into a winner of every contest. The rule is easy to state and hard to compute: the count needs a search, and deciding the winner is provably among the hardest problems of its kind. A much simpler count, the votes still to be won, usually agrees, more often the larger the electorate.
AppliedThe majority that goes in a circle
Every voter hands in a ranking, and a ranking is transitive by construction. Compare the candidates two at a time and let the majority decide each pair, and the verdicts need not fit together into a ranking at all.
AppliedThe nearest consistent verdict
When a court's majorities contradict each other, one repair is to announce the consistent verdict that disagrees with the judges least. It treats the premises and the conclusion alike, which neither of the two standard procedures does. On the classic case it returns a three-way tie; on five judges, with every question weighted equally, it never returns a single answer on a troubled profile at all — and what breaks the tie is a decision about which question matters more.
AppliedWhat the agenda leaves standing
Ask for a rule that settles each question from the votes on that question, follows a unanimous court and never contradicts itself, and search every such rule for three judges. On a ranking of three options, three survive: one dictator per judge. On two premises and their conjunction, seven survive: every rule in which a fixed set of judges must all agree. On a chain of thresholds, a hundred and twenty-nine, majority among them. The difference is not in the rules. It is in which answers force which, and whether that forcing ever runs back.
AppliedWhich vote finds the truth more often
A court that must decide two premises and their conjunction can vote on the premises or on the conclusion, and the two votes sometimes disagree. Suppose there is a right answer and each judge is more often right than wrong. Then the question has an exact answer: for three judges the conclusion vote is more often right when competence is below √½ and the premise vote above it, and in a large court the premise vote wins at any competence — while the conclusion vote turns out to be a built-in standard of proof.