Posted by Nik Coughlin on 10/02/06 20:04

David Segall wrote:
> "Nik Coughlin" <nrkn.com@gmail.com> wrote:
>
>> David Segall wrote:
>>> "Nik Coughlin" <nrkn.com@gmail.com> wrote:
>>>
>>>> A work in progress, would like anyone who has a moment to please
>>>> take a look:
>>>>
>>>> Current site is http://rankmylist.com/
>>>>
>>>> Next version is http://rankmylist.com/testTabs/
>>>>
>>> The sort algorithm does not seem to terminate! If I enter 1,2 it
>>> just keeps asking me to compare them. You need an algorithm that
>>> uses the minimum number of comparisons and maintains the original
>>> order for equal values - the "I don't know" case. You also need an
>>> algorithm that never gives the user a chance to contradict himself.
>>> That is, if he prefers 1 to 2 and 2 to 3 then he must not be asked
>>> to compare 1 and 3 in case he decides on 3. I'd like to give you a
>>> suitable algorithm but I don't know one.
>>
>> In the next iteration it will only ask you to compare each possible
>> pair once, in a random order
> I doubt if any visitor will have the patience to do this. According to
> my rather rusty maths you will be asking them to do (n*(n-1))/2
> comparisons when, in the easiest case, they only needed to do n-1. For
> a list of five items that's ten instead of four which is tolerable.
> However, most people could order five items without the aid of your
> program. For your food list it's 4,465 instead of 94.

I see your point, I'm going to try that. In fact, I may make it the default
and hide the option of doing it the other way away somewhere for those who
are both highly indecisive and have a lot of time to kill :)

>> , I was just too lazy to do this initially which is
>> why it is always random. However, the user *will* be able to
>> contradict themselves. It seems to work better if they can, believe
>> it or not!
> I'm sure it does given your definition of better. My definition of
> better was minimising the amount of effort for your user.

My definition of better was that quite often people don't really know what
they think, so it's quite possible for them to think that they like 1 better
than 2, and 2 better than 3, yet because they don't consciously consider the
relationship between 1, 2 and 3 at the time they may choose 3 over 1 if
asked.

> I should add that I think your project is a good idea and my comments
> are only intended to optimise it for my use :)

Consider it on the to do list!

[Back to original message]