Moving the Goalposts

"Show me one study." You show one study. "One study is never enough. Show me a review." You show a review. "Reviews can be captured. I need a scientist I already trust, saying it in their own words, with the raw data." The raw data arrives and the new requirement is tone, or sample size, or a result that also flatters a second belief. The goalposts moved. Moving the goalposts is changing the standard of proof after the evidence meets the standard you stated, so the original claim never has to budge.

A harder standard can be the right second thought. One story should not settle a medical question, and saying so is fair. What stings is the order of events. You named a test as if it would decide the thing. The test got met. Then, with no "I asked for the wrong thing," a stricter test appears. Owning the change would be an update. "That one doesn't count, try this new one" keeps the score at zero for as long as you like.

You can do this to other people and to yourself. A project is "successful if we hit this number." The number hits, and success is redefined as a feeling you do not have yet. A relationship is "fine if they apologize." They apologize, and fine now requires a different sentence, then a different history. The test was never a test. It was a door with the lock on your side.

• "That's not the kind of evidence I meant, now that you've brought it."
• "We said the launch was a win at this revenue. We hit it. I'm still not calling it a win."
• "Apologize and we're good." They do. "Well, not like that."
• "I'll believe it when I see a randomized trial." The trial appears. "Randomized trials aren't real life."
• "If you finish the course, you've proved you care." The course ends. The next proof is already designed.
• "Any counterexample would change my mind." A counterexample appears. "That one is a special case."
• "I asked for the source. I didn't ask for a source that disagrees with me this politely."

The meeting target: the team was told a feature ships if the error rate drops under two percent. It drops under two percent. The review discovers a new bar, delight, which was not in the document. The team did the job and lost the argument about whether the job was the job. Next quarter they will sandbag the number, because the number is not what gets honored.

The argument with a parent: the condition for trust was a steady job. The job is steady. Trust now waits on a kind of partner, a kind of savings account, a tone of voice. Each condition, alone, could have been a fair hope. Issued one at a time, after compliance, they teach the other person that compliance is not the point.

The self-review: you told yourself the anxious prediction would retire if the presentation went fine. It went fine. The prediction retires nothing. It writes a stricter exam for next time. You call this standards. It is a goal that flees so the worry can keep its job.

People stop bringing evidence, because evidence has been demoted in advance. Trust erodes in both directions: the asker never has to change, and the asked learns that the ask was theater. In public argument, a moved goalpost lets a claim survive contact with exactly the test its holder requested, which is a good way to keep a community fact-proof. In private life, it turns repair into an infinite task. You can apologize, improve, or hit the number, and the finish line is already elsewhere, looking reasonable.

Updating hurts when a belief is doing identity work, status work, or worry work. A stated standard feels fair, so you offer one. When meeting it would force the update, a new distinction appears and feels like rigor rather than retreat. The new standard is often genuinely better, which makes the move easy to defend one step at a time. From outside, the pattern is visible only across steps. From inside, each step feels like raising the quality of the conversation. Self-protection helps. If the goal can move, you never have to feel the drop of "I asked for this and I got it and I was still wrong."

Moving the goalposts is a named informal fallacy, not a laboratory effect with an effect size. The pattern is documented in argument theory as a way of avoiding the burden of a claim: set a condition for acceptance, then replace the condition once it is satisfied. It sits near special pleading, which exempts the present case from a rule, and near the motte-and-bailey move, which retreats to a safer claim. The difference is the target. Here the claim stays put. What moves is the standard of proof.

There is a fair version, and it should stay legal. New evidence can reveal that your old standard was too weak. The honest form is to say that out loud, including that you are changing the test after seeing the result, and to apply the new test to your own side. A standard you announce in advance, write down, and then obey is the opposite of this bug. Scientists raise bars all the time. They get into the fallacy when the bar rises only for the result they dislike, and the reason for the rise is invented after the result is known.

• You stated a test, it was met, and you feel no pull to update.
• The new requirement appeared after the old one succeeded.
• You would not have accepted this new requirement if the other side had imposed it on you.
• The finish line in a relationship or a self-promise keeps its distance by exactly one more request.
• You call the move rigor, and you cannot say what result would actually settle it.

Write the test where someone else could score it, before the evidence shows up. "Shown if X. Still open if Y." When X happens, say what moved, even if it is only a piece. If you want a harder test after that, say the old one was the wrong ask, and use the new one on yourself too. A mood is not a test. You can have the mood and still grant the point they actually earned.

Is it moving the goalposts to ask for better evidence?

Asking for better evidence at the start is ordinary care. Replacing a standard you already set, after it has been met, is the fallacy. You can still raise the bar. Say, "What I asked for wasn't enough, and here is the stronger test, applied to me too." The admission is the part that keeps it from being a foul.

What if I genuinely changed my mind about what would count?

Then say that you changed it, and when. "I thought one review would do it. It didn't, and here's the stronger test, which I'll also apply to my side." People can argue with that. They cannot argue with a test that quietly retires the moment it is passed.

If the evidence meets the test you named, let the test count.

The study

The number

The apology

Practice this pattern in the Reframing App - capture the trigger, label it (like Moving the Goalposts), check evidence, and write a more balanced thought.