From 0bac9ec4fab33ae4388294281bae0d0ec18dd73b Mon Sep 17 00:00:00 2001 From: Niklas Fiekas Date: Fri, 8 Nov 2024 08:45:12 +0100 Subject: [PATCH] mention quick pairing pools --- research/README.md | 8 ++++++-- 1 file changed, 6 insertions(+), 2 deletions(-) diff --git a/research/README.md b/research/README.md index 51e5617..43e6508 100644 --- a/research/README.md +++ b/research/README.md @@ -9,8 +9,12 @@ Why work with such large data sets? Replaying the entire history of Lichess encounters takes a long time, but I don't know how to avoid it. -* The observed period of time should be long, because rating periods are on the - scale of months. +* The observed period of time should be long, because optimial rating periods + may or may not be on the scale of months. +* The observed period of time should be long, because long-term changes of + rating distributions may be an important effect. +* The observed period should cover important changes, like the introduction + of quick pairing pools at the start of 2017. * Its not clear that sampling players does not introduce bias (for example, how often players around a specific rating meet).