應用機器學習方法分析中華職棒投手失壘表現

Abstract

本研究分析中華職業棒球大聯盟 32 年例行賽中 296 場之投球資料，共計 138 位投手之 55,985 顆投球，透過衡量球速、球種、投打站位、進壘位置以及好壞球數等變數對「期望失壘」進行估計。本文使用線性迴歸模型、決策樹迴歸模型、隨機森林迴歸模型作為基礎，對「期望失壘」進行估計，並比較不同模型間之優劣、變數間的重要程度等，而隨機森林迴歸模型對期望失壘進的預測效果最佳。

研究發現，在投手球數落入三壞球時，會對期望失壘產生顯著影響，其次為進壘位置與球速。因此，本研究歸納「控球」為影響投手表現的最重要因素。若以「期望失壘」衡量投球表現，會發現外籍球員普遍在此指標有較好表現，符合傳統上球隊找尋洋投之目的。此外，本研究也進一步將「期望失壘」與既有之投手指標進行對照與比較。發現由於「期望失壘」的計算包含每一顆投球的過程與結果，因此能與其他指標達成互補的效果。

若將「期望失壘」與本國籍投手薪資做比較，整體而言，本研究所估計的「期望失壘」越低的投手該年度月薪越高，但部分表現良好之新人投手之薪資明顯低於均薪，這表示中華職棒之隊伍仍依照年資與過去績效作為談薪依據，故本文建議球團可以積極培養新人，享受所謂「新秀紅利」。

This research analyzed 296 regular games of Chinese Professional Baseball League's 32th season to build a model of "expected-base-loss" by pitch speed, pitch types, pitching location, pitcher(batter) handedness and the current count. In this research, we compared the prediction performance and variable importance between the linear regression model, the decision tree regression model, and the random forest regression model. Eventually, we decided to use a random forest regression model to predict the "expected-base-loss".

We discovered there will be a significant impact on the "expected-base-loss" when a pitcher falls into the situation of three-balls count.On the other hand, pitching location and pitch speed are significant as well, so a pitcher should avoid pitching a "ball". Therefore, "pitching command" is the most important factor that influences the pitching performance. By the result of pitchers' "expected-base-loss", foreign players had better performance on this index, and it satisfied the purpose why a team wants to import foreign players. Also, we compare the "expected-base-loss" with other existing pitching indices, and the "expected-base-loss" index can complement other indices because it recorded all the pitching factors of each pitch.

In this research, we analyzed the relationship between "expected-base-loss" and the salary of local pitchers. We find they exist a negative relationship but we also find that young pitchers are underpaid according to their performance. We suggest those teams in CPBL should cultivate new bloods in order to get benefit form the "rookie bonus".