Re: [討論] Joe Maddon:“the bunt is an overrated

作者: gghh (GH)   2013-04-15 03:54:46
※ 引述《searoar (暗坑大豆)》之銘言:
: http://tinyurl.com/bqf3bb2
: is an overrated play.”
: Using the expected runs matrix at Baseball Prospectus (using 2012 data),
: http://www.baseballprospectus.com/sortable/index.php?cid=1122396
: runners on first and second with no outs yields 1.44 expected runs, while
: runners on second and third with one out yields 1.29 expected runs.
: 1.44>1.29 上太空的西瓜!!
: Theoretically, one would slightly reduce run expectancy by bunting. However,
: the certainty of scoring that one run goes up. Furthermore, bunting creates a
: more realistic opportunity for production than letting Loney, who posted a
: .630 OPS last season, swing away. Though I, like Maddon, think that bunting
: is overrated in many circumstances, that was not one of those situations.
: 但是給LONEY打就輸了GG
這段的重點不在Loney。
作者要表達的是,,雖然期望值較高,但變異數也大,在這種比賽最末段,
只要得一分就很關鍵的情況下,既然都能夠得一分,為什麼不選得分機率較高的bunt?
比賽前段或是打擊大戰用bunt當如不如用free swing
但也不能光用期望值就直接判斷bunt一定比較差 像比賽末段一分差決勝負
要重視的是variance rather than the expected value.
與其說西瓜上太空,不如說又看到一個亂用數據的人 (  ̄ c ̄)y▂ξ
作者: IanMason (伊恩)   2013-04-15 04:47:00
cannot agree with you more
作者: nolander (自己國家自己救)   2013-04-15 05:32:00
作者: OoyaoO (你今天崩潰了嗎 囧)   2013-04-15 07:26:00
講的好像bunt一定會成功一樣
作者: nolander (自己國家自己救)   2013-04-15 07:32:00
哪裡有講得像bunt一定會成功一樣= =?
作者: conjohn (conjohn)   2013-04-15 09:59:00
變異數應該就已經考量bunt失敗了吧只是考量後變異數還是較小
作者: searoar (暗坑大豆)   2013-04-15 10:13:00
重點是Loney啊 who posted a .630 OPS last season
作者: searoar (暗坑大豆)   2013-04-15 10:16:00
如果打的人是Pujols 你根本就不會去考慮觸擊 就算他是DP王
作者: RogerWaters (希望你在這裡)   2013-04-15 10:37:00
原文根本沒說什麼Variance的問題, 只是0出局1,2壘有人雖然比1出局2,3壘有人多得一點分, 但得一分的機率1出
作者: RogerWaters (希望你在這裡)   2013-04-15 10:38:00
2,3壘有人是比較高的, 打到九局1比1得一分就贏,還想多
作者: RogerWaters (希望你在這裡)   2013-04-15 10:39:00
得一點分, 何況還遇到上季ops只有.630的Loney
作者: CGary (下雨天也挺浪漫的)   2013-04-15 14:07:00
其實西瓜還有一個錯誤 統計數據可以廣泛應用在的結果 不可
作者: CGary (下雨天也挺浪漫的)   2013-04-15 14:08:00
直接套用單一案例(Loney)身上...很明顯這兩件事是不同的事

Links booklink

Contact Us: admin [ a t ] ucptt.com