Failure Streak Recovery
also Exponential Backoff in Difficulty · Reducing Difficulty After Failure
A control rule that cuts the next task's difficulty exponentially for each consecutive failure, rebuilding success before ramping difficulty back up.
Failure streak recovery is Elliot Temple’s explicit feedback rule for learners who hit several failures in a row. The remedy is not to try harder at the same difficulty but to drop the difficulty of the next task exponentially, once per consecutive failure, until you score a clean success. One failure: aim for roughly a third (Temple’s other formulation says half) of the difficulty. Two in a row: a third again, so a ninth. Three: a twenty-seventh. Recovery means deliberately doing something easy enough to win, then climbing back up.
CF treats this quantitatively because difficulty scales non-linearly: juggling two new things at once is more than twice as hard as one. So failures compound, and a streak signals you badly misjudged what you could do. The rule keeps your success rate up around 80-90%+, where learning actually works, and it directly serves error-correction cadence — you must reach a judgeable success-or-failure verdict frequently (Temple’s upper bound is weekly) so a streak is caught early, not after weeks of stalled effort.
The central oppositions are concrete. People take pride in attempting hard things and failing, treasuring impressive-looking struggle over progress; CF says prioritize success and count small wins. And people’s difficulty cuts after a failure are routinely far too small, producing 10+ failure streaks — Temple calls this one of the biggest mistakes learners make. Difficulty is also tracked per topic: a chess success doesn’t license a harder cooking project. With zero prior failures on a topic you may jump ahead to find your level, but the first failure forces the backoff discipline on.