Moving the goalposts

Grain of Salt is about how a claim is tested, so every example here is drawn from work, projects and inspections, where the standard is usually written down and the move is therefore easy to see.

What moving the goalposts is

Moving the goalposts replaces a criterion of success with a stricter one, after the first criterion has been satisfied and without conceding that it was. The person who set the bar does not deny the evidence. They accept it, absorb it and produce a new requirement in the same breath, so that the concession never has to be made out loud.

Two neighboring logical fallacies are often given this name and should not be. The no true Scotsman fallacy edits the definition of the subject rather than the standard of proof. The argument from ignorance, sometimes called the appeal to ignorance, demands that a claim be disproved rather than supported. Moving the goalposts leaves both the subject and the burden where they were, and changes only the height of the bar. All three are species of non sequitur, since in each case the conclusion stops following from what was actually shown.

The move in one sentence, and the two parts it always has

Every instance has two parts: a standard that was set, explicitly or by clear implication, and a second standard introduced only after the first was met. Without the first part there is no fallacy at all, which is why unstated expectations cause so many arguments that feel like this one and are not.

A gardener is asked to prove that a new hedge will screen a neighboring car park. He plants a trial section and photographs it through the year. The neighbor then says that photographs are not enough and asks for a written statement from an arborist. When the statement arrives, the neighbor says the arborist should have visited in summer. Three standards, one hedge, and at no point has the neighbor said what evidence would end the conversation.

The four forms the move takes

The move takes 4 forms, and the last of them is the hardest to name in the room because it sounds like diligence.

• Raising the threshold. The same measure is kept and the required value rises: two seconds becomes one second.
• Adding a requirement. The original measure is met, and a second, previously unmentioned measure appears beside it.
• Restricting the evidence type. The finding stands, but now only a certain kind of source will count.
• Extending the scope. The claim was proved for the case at hand, so the demand is quietly enlarged to a wider set of cases.

All four produce the same signature, and it is worth learning: after the exchange, nobody can state what would settle the question. A standard that cannot be written down in advance is not a standard. It is a mood.

A worked example, and the repair

Here is the raw version, from a software project. The client's brief says the report page must load in under 2 seconds on the office network. The developer delivers a page that loads in 1.5 seconds. The client replies that it should also work with the network disconnected, and that until it does, the performance work is not accepted.

The repair is procedural rather than rhetorical, and it is the reason acceptance criteria exist. Write the standard down before the work starts, in a form that can be measured. Close each criterion out loud when it is met. Log the new requirement as new, with the schedule change it implies. None of this stops a standard from rising, and it is not meant to. It stops a standard from rising invisibly.

What moving the goalposts is confused with

4 things resemble the move, and two of them are legitimate.

• Scope creep is the project management name for the same shape, applied to deliverables rather than to evidence. It is often nobody's fault and it still needs the same paperwork.
• A revised standard is legitimate whenever new information shows the old one was measuring the wrong thing, provided the revision is announced as a revision.
• The sunk cost fallacy pulls in the opposite direction: it keeps a failing project alive because of what has already been spent, where moving the goalposts keeps a succeeding one from ever being finished.
• Ad hominem appears when the goalposts stop moving and the objection turns to the person supplying the evidence, which is usually the next stage of the same conversation.

When raising the bar is not a fault

Raising a standard is legitimate under 3 conditions, and all three are about disclosure rather than about the height of the bar.

1. Announce the change as a change, and say plainly that the previous standard was met before the new one is introduced.
2. Give the reason, if new information has arrived: a measurement error found in the original test, a case the criterion never anticipated, a consequence nobody had priced.
3. Fix the new standard in advance of the next attempt, so the person doing the work knows what finishing looks like.

There is also an honest limit worth stating. Extraordinary claims do warrant more evidence than ordinary ones, and asking for more when a result is surprising is good practice, not a fallacy. The difference is that a genuine skeptic will say what the additional evidence would be, and will accept it when it comes. Someone moving the goalposts will not tell you, because telling you would end the game.

The test to run on the next moving standard

Ask 4 questions the moment a requirement changes after evidence has been supplied.

1. Ask what the original standard was, and whether it can be found written down anywhere from before the evidence arrived.
2. Ask whether the original standard has been met, and require a yes or a no before any new requirement is discussed.
3. Ask what evidence would settle the matter, and treat the absence of an answer as the finding.
4. Ask what changed to justify the new standard, and whether the reason would have applied just as well last week.

The phrase outside argument

The metaphor comes from field sports, where the goal is fixed for the whole match and the absurdity of carrying it about is the joke. It circulates now mainly as an image, a short reaction clip or gif of posts being dragged across a pitch, which is a fair summary of the fallacy in two seconds and no substitute for naming the standard that actually moved.

The phrase has a second life in economics and management writing, where it describes the changing objectives of long running programs: a scheme launched to raise output, reported later against employment, and reviewed later still against exports. That usage is descriptive rather than accusatory, and it points at the same underlying problem. When the measure of success is chosen after the results are in, no result can fail, and a program that cannot fail has stopped telling anyone anything.

Where to go next

FALLACIESLogical fallaciesWhat they are, with examplesFALLACIESAd hominemAttacking the person, not the argumentFALLACIESNon sequiturThe conclusion that does not followBIASESThe sunk cost fallacyThrowing good money after bad