Jump to content

Talk:Weissman score

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Damn

Rationale & criticism

[edit]

Why was the formula built this way ?

Factoring in compression time obviously makes sense for any pragmatic purpose, but the rest...

  1. The log seems to be to keep the figures at bay, since compression times vary wildly.
    But shouldn't it happen after the time ratio ? Before is responsible for time unit changing the result.
    Shouldn't it happen after the size ratio too ? Before means a compressor reaching half the size in twice the time gets a lower score.
  2. Does a ratio of compression ratios make sense ?
    Isn't that the same as (reference compressed size / benchmarked compressed size) ?
  3. Why include a scaling in the formula ? Moreover an undefined one ?

All these make the scores less comparable between runs.
All things considered, this formula looks like a botched job.--Musaran (talk) 11:58, 1 January 2020 (UTC)[reply]

We'd need an interview for a citation, but I'd bet the answer to all of these is "it was made for TV".
It should be pretty obvious though that the "perfect" algorithm according to this score is just "return the input as-is, then wait for 1 {time unit}". You'll always end up with infinite score compared to any other algorithm. 96.87.87.25 (talk) 18:29, 8 October 2020 (UTC)[reply]