From 9ce80a21b15373f4e4a4ed6e8f5d90cbc57bf499 Mon Sep 17 00:00:00 2001 From: Trawlnetter <250747469+Trawlnetter@users.noreply.github.com> Date: Sun, 21 Dec 2025 02:36:59 +0000 Subject: [PATCH] Remove "similar members" explanation entirely Based on https://github.com/iftechfoundation/ifdb/issues/449 The "similar members" recommendation system has been dropped. The text explaning how it worked was still in the code, but this removes it entirely. --- www/help-crossrec | 35 ----------------------------------- 1 file changed, 35 deletions(-) diff --git a/www/help-crossrec b/www/help-crossrec index bf50cfa5..cd7e26b0 100644 --- a/www/help-crossrec +++ b/www/help-crossrec @@ -15,12 +15,6 @@ members and review sites like target="_blank">Baf's Guide. We're not being paid to promote one game over another. - -

The system picks front-page recommendations by randomly selecting a few games with the highest ratings, excluding games that you've told us you've already played. @@ -36,35 +30,6 @@ confidence in the game, by adding five "fake" ratings to the average (one 1-star, one 2-star, one 3-star, one 4-star, and one 5-star rating) and subtracting the standard deviation from the result. - - -

The system picks front-page recommendations by looking for other -members with similar patterns of likes and dislikes to your own, as -expressed in the ratings they gave to the games they reviewed. IFDB -tries to match you up with a few other members, then recommends games -that those other members rated highly. - -

In principle, the more ratings you and other members provide, -the more accurate the matching will become. So the recommendations -should get better and better as you rate more games. - -

This approach is sometimes called "collaborative filtering." Some -people think it's great, others are skeptical. The obvious objection -is that it doesn't capture the reasons that you like the games -you like, so it might match you up with someone who happens to like -some of the same games, but for completely different reasons. There's -obviously no guarantee that the approach will actually produce good -advice, but we hope that it gives you at least a few leads on games -that you might otherwise have overlooked. - -

If the algorithmic recommendations on the home page don't work for -you, remember that IFDB still offers several ways to get -personal recommendations from other users, such as member -reviews and Recommended lists. -