Sunday 29 July 2018

FE Heroes: Arena Tiers 2.0

A year ago I analysed the tier distribution in FEH and now with the v2.8 update we are going to have a twist, with the new tier 21 [a.k.a. the WHALE TIER] introduced. It is no more than a simple application of queue theory and matrix diagonalization....well no, just solving linear equations to get the new distribution. Just a quick reminder, we can generate the ratio by the following equation

$p_{k+1} = r_{-1, k}^{-1}(\sum _{d =1,2,3} r_{d,k} p_{k-d} - (1-r_{0,k})p_k)$

from the bottom, where $r_{d,k}$ stands for the ratio of going up by $d$ tiers from tier $k$, and $p_k$ stands for the proportionality constant of players at tier $k$ in an equilibrium state.

Quite surprised that they looked extremely similar, right? If you look deep into the chart you will found that to be unsurprising at all. You may even wonder why are they different...

...that's because tier 20/21 are the only tiers with $r_k$ changed, after all.

(Apology for the stupid y-axis.) Numerically with the introduction of tier 21, players moved upward by 2%. There will be 2% more tier 18+ players [tier 18 is important due to the 4 orbs reward], and 2% more tier 20 players, which we, infantry lovers, do not care at all.

So here is the counter-intuitive part: why is there a difference on tiers below tier 20, given that the parameters were not altered? If we hide the flow between tier 20-21 and call the two tier 20+, this new tier actually has the same parameter as the old tier 20:

Old tier 20: 30% stay - 70% down 1

New tier 20: 20% up 1 - 10% stay - 70% down 1
New tier 21: 30% stay - 70% down 1
New tier 20+: 30% stay - 70% down 1

...or is it?

For sure that's not correct - by combining two tiers we brake the memoryless feature of the promotion system: if one stays in tier 21, he will 100% stay in tier 20+ in the next season, and that is not consistent with the global parameter for tier 20+. We can't really solve a set of linear equations correctly without the memoryless feature so the idea of tier 20+ does not really work. One simply go back to tier 20/21, do the math and 2% is really an existing difference.

...but why?

Even though the idea of tier 20+ does not help mathematically, it is true conceptually. Players are allowed to stay in tier 20+ longer [for those in tier 21] comparing with the old tier 20, hence expending the capacity at the top.

And, well, for me I do not bother getting top ranks anymore. I'm happy with my tier 19.5, or even 18.5 teams...

Wednesday 25 July 2018

25/7/18

~10年後の私へ~

今は幸せでしょうか?
それとも悲しみで
泣いているのでしょうか?


2008-2018,不多不少就是十年。最近在我身邊剛好出現了各種以十年為單位的懷舊復刻;
比如余兒<<九龍城寨>>的前傳<<龍頭>>在本年書展首發。
比如The World Ends With You的續作TWEWY -Final Remix-會在九月開賣。
比如RPG Maker VX的續作RMMV即將移到Switch上。嘛,我還是喜歡RMXP就是了。

仔細一想,現在的人生似乎也是以2008為轉捩點開始的十年所鑄造出來:
比如從公共圖書館轉戰書局和網絡文學。
比如從仙俠類MMORPG轉戰音游。
比如開始理解PHP,還有Discuz!架構。
比如那一狗票的八號風球。
政治經濟等等就更不用多說,怎樣看都是風雲突變的一年。

其實也沒甚麼啦。差不多的論調在2017還是2019一樣能用的上。不過等了好久的TWEWY終於出來讓我不感概一下不行如此而已。


啊,順便慶祝一下500道達成。算上DDR X3/13/14/ITG的話大概快破千,不過官方的500道就是500道啦。上面截圖的日期是7月2日,不過此刻[24/7/18]我都過550道了。之前順手打了一篇17譜面簡評不過還是吞回去好,都把18清了一大半還回去說17全破怎樣怎樣實在不怎樣,尤其在今天看到有人4miss過掉IX之後。

要加強的東西還有很多呢。