QUOTE(cs987987 @ Dec 1 2021, 01:30) 

有週期嗎
還是看數量決定啊
(started / 1000 + voted / 10000).
(1000/1000 + 10000/10000)
兩個都滿的
但是準確率只有95%
一個是
(500/1000 + 5000/10000)
但他準確率是100%
分數怎計算啊
QUOTE(PrincessKaguya @ Dec 1 2021, 01:47) 

我也想知道
菠蘿跟我說了公式之後
我花了老半天時間也弄不明白這東西如何計算 (IMG:[
invalid] 
style_emoticons/default/heh.gif) 
不是很简单吗?
- As part of an internal reword of tagging statistics, the mod power formula has been changed. It now looks at started and voted accuracy separately, and for each, gives +0.55 per percentage point over 90% accuracy and subtracts 0.5 per percentage point below 85% accuracy, with a total cap of (started / 1000 + voted / 10000).
也就是说,准确率95%,也就是0.55*5,但是超过了 (1000/1000 + 10000/10000)=2,所以是 tag mod +2。
而准确率100%就是 0.55*10,同样超过了(500/1000 + 5000/10000)=1,所以 tag mod +1。
算法大概是
x1=0;
if(started_percent>=90) x1=0.55*(started_percent -90);
if(started_percent<=85) x1=-0.5*(85 - started_percent);
x2=started / 1000;
x3=0;
if(voted_percent>=90) x3=0.55*(voted_percent -90);
if(voted_percent<=85) x3=-0.5*(85 - voted_percent);
x4=voted/10000;
tag mod = x1+x3 < x2+x4 ? x1+x3 : x2+x4;
自己试试,应该没错。
(本该看paper的我却在研究这玩意,真是蛋疼 (IMG:[
invalid] 
style_emoticons/default/wacko.gif)