Project Euler Forum

hk wrote: How about the following levels:
Level 1, Tetrahedron, 50-99 problems solved.
Level 2, Cube, 100 or more but less than 70% of the probems solved. (currently =100-140 solved)
Level 3, Octahedron, 70%-80% solved
Level 4, Dodecahedron, 80%-90%
Level 5, Icosahedron, 90%-99%
Level 6, Sphere, >=99% solved.

Personally, I like this suggest rather than "fixed" levels... it means that people who quit will eventually lose their levels, making way for the active users!

The relative levels idea is - excuse me, Hans - bonkers. I mean, what benefit does it add to a system that already reports your percentage to also report a coarsened version thereof. I mean, if it already says I'm a 73%er then SO WHAT if it also reports "you're in the 70-80% band". Unless the payoff is better than some bog-standard graphics, I think that coarsened percentages (admit it: that's what relative levels are) are a waste of time.

greekcandyman wrote:

hk wrote: How about the following levels:
Level 1, Tetrahedron, 50-99 problems solved.
Level 2, Cube, 100 or more but less than 70% of the probems solved. (currently =100-140 solved)
Level 3, Octahedron, 70%-80% solved
Level 4, Dodecahedron, 80%-90%
Level 5, Icosahedron, 90%-99%
Level 6, Sphere, >=99% solved.

Personally, I like this suggest rather than "fixed" levels... it means that people who quit will eventually lose their levels, making way for the active users!

I also prefer all fixed levels, although only the first level needs to be fixed for someone to never lose being ranked. If people like the challenge of endurance and staying up to date, we can have another relative level system.

Ah, we crossed in the post.

Well, so the point's not lost: I say no, no, no! to relative levels.

Just my 2p.

I think we hit the right point now:
The only conclusion I can draw from the bonkerniness of realative levels is:
All levels are bonkers, relative as well as absolute.
Simply display, paginated, all people that solved more than 50 problems.
Paginating does the rest. No need to introduce levels whatever they are named.
As fighting your way through the problems is an individual affair everybody can set his own goals.
No need for us to introduce fancy levels.

I would have to agree...fixed levels are better than relative levels, but I would rather not have any levels at all. The numbers speak for themselves.

An alternative to all of this would be to combine the stats and scores sections into a new feature that can show you all the scores within a certain criteria, including, but not limited to, programming language, country, and score range. It could also rank people based on certain problems (e.g. last 100 problems only, or problems 23 and 83-123). The default on the feature would be "all scores greater than or equal to 100, any country, any programming language, all problems", which is essentially the top-scorers list. It would be great addition to the site.

But since I use Project Euler to impress girls, I think they'd find an octahedron more attractive than page #6 or 102 out of 202.

On a serious note, I think having a fixed level/rank, as opposed to the ever-changing pagination or relative ranking, would be not "bonker" to some people. Especially the newcomers, who would probably want to achieve a certain rank before taking a moment's breath, and not want to come back to find out they are wiped off the list.

We are caching all pages to lessen the burden of the mySql server.
If everybody is allowed to think up his own queries the server will be down before you can say: "2 is the smallest prime".

Makes sense, but nonetheless, it'd be a great add-on and whenever Project Euler finds the server capacity, it's always a possibility.

Anyhow relative levels or absolute levels I don't see the use of levels for which it will take one or two years from now before we have enough problems so that anybody can reach that level. So I doubt the increased bonkerniness of relative levels over absolute levels.

hk wrote:We are caching all pages to lessen the burden of the mySql server.
If everybody is allowed to think up his own queries the server will be down before you can say: "2 is the smallest prime".

Is it intractable to cache another set of pages?

For your information on how the caching modules I wrote work... whenever a query is issued the server checks for the presence of a cached file. If the file does not exist or its time stamp is over one hour old then a new database query is issued and the results are stored in the cached file. Some cached files contain the raw HTML output (e.g. the score tables) whereas other files contain the pre-processed variables necessary to display the page (e.g. the problems table, sorted by ID or difficulty). The advantage of this technique is that caching only takes place when a particular query is issued and every page does not have to be re-cached every hour. For example, someone clicks on the England flag and because the old cached file had expired a new one is created. If anyone clicks on the England flag within the next hour then they will see the cached page. If no one clicks on the flag for a week then that page will not be cached again until requested.


I agree with hk in that it does appear strange (bonkers) to have levels that cannot be achieved yet. However, let us remind ourselves why we're thinking about introducing levels... I recognise that the real emphasis of Project Euler is solving the problems, but the current system only captures the top 1000 members. Those lucky enough to have entered the top 1000 table have found immense satisfaction in watching their username climb, but so far this encouraging experience has been restricted to them alone. For many members entering this table would remain near to impossible, as their battle is not with those talented members, but their own personal development. By introducing "bands" with fixed boundaries, members are given definite targets. They know that if they can solve, say, 50 problems then they will find themselves immortalised in a score table. Once they've arrived in that table they can then can enjoy watching themselves climb this table towards the next level, and so on. With relative levels, like the top 1000, the goal posts continue to move and members would not be competing against their own stockpile of solved problems but which percentile they find themselves in.

Of course, there are benefits with both, but I still believe that the following fixed boundaries closely capture what we're trying to achieve:

Level 1, Tetrahedron, 50-99
Level 2, Cube, 100-149
Level 3, Octahedron, 150-199
Level 4, Dodecahedron, 200-249
Level 5, Icosahedron, 250-299
Level 6, Sphere, 300+

If no one has any fundamental objections then when I get some time I will advance with this model.


As an additional option, the "performance" progress bar could be converted into a "genius" progress bar; it seems that some people are missing the genius rating. Or do we like the idea of using the bar to record the performance based on the last X problems?

I'd rather have the genius percentage because it makes me feel special!
And I like it better...

euler wrote:As an additional option, the "performance" progress bar could be converted into a "genius" progress bar

I think that would be nice. A genius bar calculated from last X problems solved to vary the scoring.

euler wrote:
Of course, there are benefits with both, but I still believe that the following fixed boundaries closely capture what we're trying to achieve:

Level 1, Tetrahedron, 50-99
Level 2, Cube, 100-149
Level 3, Octahedron, 150-199
Level 4, Dodecahedron, 200-249
Level 5, Icosahedron, 250-299
Level 6, Sphere, 300+

If no one has any fundamental objections then when I get some time I will advance with this model.

I fact I have objections. I still don't see the use of ranks that are unattainable for the next one or two years.
As stated earlier in this thread smaller bands are preferable. I also think that someone that solved 200+ problems does not need ranks anymore to stay hooked. Trying to solve all problems will be his/her target I'm sure.
From these points of view the following bands are preferable.

Level 1, Tetrahedron, 50-79
Level 2, Cube, 80-109
Level 3, Octahedron, 110-139
Level 4, Dodecahedron, 140-169
Level 5, Icosahedron, 170-199
Level 6, Sphere, 200+

A performance bar based on the last 25 problems seems a good idea to me.

neverforget wrote:I think that would be nice. A genius bar calculated from last X problems solved to vary the scoring.

When I introduced the "performance bar" my thinking was...

For most members when they join Project Euler they may come with little experience or already possess a lot of knowledge and computing skills, but even they would currently be unable to tackle some of the tougher problems that deal with some areas of specialism. Most members use the large problem base as a kind of training ground to learn how to tackle the more challenging problems. Due to the oscillating nature of difficulty of newly published problems, the last 25 problems should, in theory, cover the full range. So how well you could do on the last 25 is a proper measure of your current "genius" rating. This means that new members who already arrive highly skilled could perform well on the last 25 and in a relatively short space of time and it would be clear that they were indeed talented.

So the question is... do we use the "current performance" (measured from the last 25 problems) to replace the previous "genius rating" (based on the entire problem set), or do we have both have both, in which case do we want progress bars for both?

hk wrote:I still don't see the use of ranks that are unattainable for the next one or two years.

The reason I would still support having currently unattainable ranks is that if/when this system is introduced some members would be instantly transported to the top level. For those that might have enjoyed going through at least one advancement there would be nothing to look forward to. It may even leave some of them feeling that there is nothing left to achieve.

So perhaps the following boundaries are a good compromise in that the top level will be achievable by this time next year. I certainly don't think these gaps are too large, as we are not giving "levels" away and earning one is meant to be an achievement.
Level 1, Tetrahedron, 25-49
Level 2, Cube, 50-99
Level 3, Octahedron, 100-149
Level 4, Dodecahedron, 150-199
Level 5, Icosahedron, 200-249
Level 6, Sphere, 250+

With the additional condition that you are only immortalised once you've solved 100 problems; that is, members will disappear from the level 1+2 tables after 60 days of inactivity.

I believe that we should use a performance bar and a genius rating. The performance bar would indicate that the user has solved tricky problems, while the genius rating would simply provide the feeling of progress, little by little, despite not having gained a level.

I suggest, though, that the genius rating is displayed simply as a number, and not as a bar. The point of the genius rating (for me, anyway), is to be able to instantly see the change in percentage (in increments just a little bigger that # solved). However, a small graphic bar would not be able to show such change so easily. The change would only be about a pixel every few problems. That's not so easy to spot!

Alternatively, the use could select his/her preferences from a variety of options (#solved, percentage, performance, or more). Those could be stored in a cookie. Then, for the scores page, raw info could be sent, which would be assembled for view using javascript. However, that sounds like a nightmare to implement/keep up. In other words, it may be best to ignore this paragraph.

I think euler mentioned that the caching for score tables only memorizes any score tables viewed, refreshing (recalculating) if time stamp expires (1 hour i think he said). Each table has its own cache. It is not directly obvious if grouping the caches into "raw info" would help. Grouping all tables into one cache, and using server processing, still poses calculation overhead per page view. If the processing is moved out to browser php / js processing, it might screw the bandwidth. It may be simpler to display both genius % rating and the last-25 performance bar in the same table and have separate pages for sorting by either.

The idea of a ranking system is stimulating, but I too don't like the
military ranking very much.

I would suggest to borrow from sport, in particular judo and go, in using dan-grading.
Dan grades go on and on, so there is no need to invent names.
60 problems : black belt. 1th dan. Each 20 problems earns you a dan -grade.
80 : 2th dan
..
240 : 10th dan.
Above 9th dan one could be awarded a red belt.

Below sixty it is more colorfull, with kyu grades.
50 brown belt (1th kyu)
40 blue belt
30 green belt
20 orange belt
10 yellow belt
0 white belt (6th kyu)

To stimulate new-comers the white belt is subdivided
.. 8 white belt, brown slip
.. 6 white belt, blue slip
.. 4 white belt, green slip
.. 2 white belt orange slip
.. 1 white belt yellow slip

IMO medals should not be based on ranks, but on merits.
Merits could be associated with the difficulty of problems, as calculated by how many
persons have solved a problem. So a medal could be awarded for having solved halve of
the total difficulty of all problems combined.
An other, simple one is to state how many times a person was the first to solve a problem,
and how many times one of the first 10 and 100 to solve a problem. (Gold silver bronze medals,
with a number).
A more difficult one to judge is an award for an original algorithm, not foreseen by the
originator of a problem.

Groetjes Albert