News - Project Euler

The departure of the fastest solvers tables due to the data becoming increasingly polluted with cheats and members using genAI has meant more than just their removal from public view. It has also meant that one of the primary metrics we have used to rate problem difficulties has become increasingly unreliable. In an attempt to find a long term solution to this we have investigated other candidates, and we believe that we have found a balanced hybrid system that should provide measures that are less volatile and in theory can be used with confidence for many years.

Consequently, Project Euler will be taken offline briefly on Saturday to update the site with the new rating system. As mentioned, the previous metric relied heavily on the fastest solvers data, but with extensive correlation analysis it became clear that the fastest solvers data had a definite bias for certain types of problems. Fortunately, all of the new metrics we will be using: dynamic rating, demographic solves, and team ratings, generally agree with each other, which grants a greater degree of confidence in their combined reliability. However, we ask you to prepare for some surprising and, in some cases, seismic shifts in either direction. You will see some previously hard rated problems move down and at the same time other problems will shoot up towards the top.

We appreciate that this might be frustrating for those of you who had been working on that elusive problem at a particular difficulty level. So, ideally we do not want to make any significant changes after this, but there does remain scope to make small calibrations if necessary.

In addition to this, we have also decided to change the number of members at each rank to allow more to feel included. Moving from squares to cubes, we will have 100 Spheres, 800 Icosahedrons, 2700 Dodecahedrons, 6400 Octahedrons, 12500 Cubes, and 21600 Tetrahedrons, which means that 44100 members in total would have a ranked status.

After this update you will also notice below each problem that the information for each problem contains a lot more data. It will show a mean completion time of the most skilled solvers along with some fascinating demographic data, showing the percentage of members at each rank who have successfully solved the problem.

We would also like to thank Icy, griff, and Pacôme for joining a new moderator team. They are well known and respected members in the Project Euler Discord community. Their roles will include helping to keep the discussion threads up to the same high standards we have come to expect, including the removal of posts that disregard our genAI policy. They will also have the capability of viewing profiles of members flagged with suspicious solve patterns and taking action where necessary to help protect the integrity of the vast majority of our members, whilst ensuring that "cheats rarely prosper".

Finally, once we have completed this update and dealt with any bugs, we will be earnestly moving into the next major phase. We can't go into too much detail at this stage, but currently we are planning to introduce private and custom "groups" to Project Euler which will allow like-minded members to track each others progress. We are very excited by this and we hope you share our enthusiasm.

An update to the site-wide style has been made. The main reason for this was to improve accessibility for members if they wish to use the latest OpenDyslexic font. We took this opportunity to update the default font for the website from Segoe UI to one of the modern geometrical sans-serif fonts: DM Sans; it felt appropriate to make use of a font that has been designed with a focus on mathematical symmetries as part of its aesthetics. On your Account page is the option to switch between DM Sans, OpenDyslexic, or the old Segoe UI, if you don't like change. It should be noted that in order to accommodate OpenDyslexic, some minor compromises on the generic style settings have been made which will result in certain elements appearing slightly different to how you might have previously recalled.

There has also been changes to way that the default light mode and dark mode is handled. In particular, there is the option on your Account page to switch between Light (Default), Dark, or System; the latter will adopt the preference of either your browser or your operating system. It was a difficult decision as to whether or not to make System the default setting. Even though many members prefer dark mode, the signature look of Project Euler is light mode. So it was felt that any changes to this should be a deliberate choice of the member. Having System as the default would mean that new members who prefer dark mode settings might never see our signature look.

Hopefully these changes shouldn't have any negative unintended consequences impacting site usability, but as always, please report any issues you notice.

We are very grateful for the understanding shown after sharing the difficult decision to remove the fastest solvers tables in their current form, and as promised, we will be working in the near future to consider a replacement.

The reason for the removal of the fastest solvers tables means that the Eulerians table, which shows the "successes" of recent solvers, has become equally polluted with cheats. As a result, they will be removed in the next day or so. The good news is that at the same time we will be introducing a new data packed table in the Statistics section that shows the ranks of the problem set over the past 28 days, as well as the lowest/highest ranks seen (with dates). The data has only been captured across all problems since 17 April, which marks the time when we made the final tweaks to the rating algorithm, but it already provides useful insights. In addition, below each problem you will be able to see a graph showing the movement of ranks over the past 28 days. The minimal API will also be updated to allow members to capture current snapshots of problem difficulty ranks (using minimal=problems).

Once again, thank you for your understanding, and we hope that this does not detract from the main reason we have such dedicated and engaged members: the problems. We hope that you will enjoy the upcoming problems, especially as we celebrate the publication of the landmark problem 1000 shortly.

For many years, a number of our members have enjoyed racing to be among the first to solve newly released problems, and others have enjoyed marvelling at seeing the speed at which the problems have been conquered via the fastest solvers tables.

However, with the continued improvements in generative AI it is clear that more and more of those fastest solvers have been using genAI tools to solve the problems, undermining the accomplishments of our highly respected and talented members. We have now reached a point where genAI is able to solve almost every new problem we produce and it's only going to get better. Consequently, the fastest solvers tables contain a mixture of the most skilled problem solvers on the planet and others who have been shamelessly taking credit for genAI's ability to solve the problems.

As it is currently impossible to distinguish between those, it is with great sadness that we are removing the fastest solvers tables shortly, with problem 1000 being the last opportunity to earn a place. The team are working hard behind the scenes to find a reliable alternative. But ultimately, we may conclude that this is a fool's errand.

Despite the disappointment felt by the whole team, we can assure you that the regular release of high quality problems will continue to remain our number priority and will not be impacted by this decision.

In lieu of their absence, we would ask that members do not use the problem threads to boast that they were the nth person to solve the problem or make rushed posts as a cheap opportunity to do the same. The threads should remain a place of high quality, instructive dialogue. Any posts that abuse the use of the threads will have posts edited or removed and may result in that member being banned from posting in the threads.

It is hoped that you understand the need to take this regrettable course of action.