Back to the incidents log

Incident: Truncated Skewb scrambles in versions of TNoodle version 0.13.4 and earlier

Status: Resolved

4b1 4b3 4b3c tnoodle

BelgiumSummer2015 CubingUSANationals2018 HasseltOpen2014 HotelCube2018 ImaginaliaOpen2018 IndianNationals2015 LTCKids2018 MumbaiSummerOpen2014 SantaFeAutumnV22018 XianOpen2015

Incident description and resolution:

A bug in TNoodle 0.13.4 and earlier resulted in 86 Skewb scrambles for official competitions that were accidentally truncated. As a result, they were not generated in a way that satisfies the requirements in Regulation 4b3. Of these scrambles, 73 were for CubingUSA Nationals 2018. The WRC has decided to keep these results.

As described in the TNoodle 0.13.5 announcement:

Per the WCA Regulations, an official scramble sequence for Skewb must produce a random state (4b3) from those that require at least 7 moves (4b3c) to solve.

In order to make it hard to distinguish scrambles with short solutions on a scramble sheet, TNoodle was designed to ensure that all generated Skewb scramble sequences are at least 11 moves (even if there is an optional solution with fewer moves). However, a thread safety bug in TNoodle before version 0.13.5 made it possible for scramble sequences to be truncated, such that:

  • a truncated scramble sequence would be used if it required at least 7 moves to solve (still satisfying Regulation 4b3c), but
  • the truncation did not preserve the random scramble distribution (violating Regulation 4b3).

More details are available at the TNoodle project on GitHub.

An analysis of solution lengths for historical scrambles suggests that this did not affect the distribution of states for the vast majority of past competitions. However, it did result in a significant bias of states towards shorter optimal solutions for at least one major competition (CubingUSA Nationals 2018), which led to the discovery of this bug. We would like to thank Sébastien Auroux for bringing it to our attention, and Tim Reynolds for working with Jeremy Fleischman to identify and prepare a fix for TNoodle over the past weekend.

We have also verified that scrambles for other puzzles were not affected by a bug like this.

A full list of affected scrambles is available here. In particular, the distribution of optimal solution length for Skewb scrambles at CubingUSA Nationals 2018 was as follows:

Optimal Solution Length Expected # of scrambles Actual # of scrambles
7 moves ≈20 51
8 moves ≈78 84
9 moves ≈93 61
10 moves ≈5 0
11 moves ≈0 0

We believe that the large number of attempts for the competition caused conditions for the bug to happen with higher likelihood on the computer that was used to generate scramble sequences for CubingUSA Nationals 2018.

Although the impact on the scramble distribution is measurable, the WRC has decided to keep all Skewb results from the 86 truncated scrambles because:

  1. We do not believe that this bug was known to anyone, or purposely used by any organizers, Delegates, or competitors. This bug is the responsibility of the WRC and the WST.
  2. Delegates properly followed the Regulations by using the output of a current official version TNoodle without inspecting or analyzing the scrambles ahead of time, as required by Regulation 4b1.
  3. The scrambles all still required at least of 7 moves for an optimal solution.
  4. The affected results are not affected to the extent that the competitors received a clearly unfair advantage against future Skewb competitors.

For future incidents, please note that attempts should still be removed if their scramble sequences do not follow the Regulations because they did not come from a current official scramble program, as in the case of an incident where an old version of TNoodle was used by accident.

This incident was created on .