建構MLB前五局比賽的領先機率模型及驗證其準確性

Building MLB 1st 5 Innings Predictive Model and Quantifying its Success

張家豪(Chia-Hao Chang)

《統計與資訊評論》 19卷 (2019/12) Pp. 39-58

摘要

本文章的主要目的為：提供一個可以打敗線上投注賠率的模型方法，所考慮投注的項目是：美國職棒大聯盟（MLB）前五局的不讓分獨贏盤。線上博彩公司的投注賠率代表著兩個涵意，一、具體量化投注者獲利的金額，二、多家博彩公司彼此，和投注者多方妥協的市場平衡點。若投注者本身有一個相較投注賠率準確的預測模型，將可創造正的預期回報值。在本文當中，我們將預測MLB比賽的相對比分：以馬可夫鏈方法和跑者推進模型估計不同場上情況和球員能力而產生的球隊期望得分，進而獲得兩隊各自的獲勝機率。

關鍵字

投注賠率 ； 隱含機率 ； 馬可夫過程 ； 跑者推進模型 ； 預期回報

並列摘要

The main purpose of this article is to provide a model that can beat online betting odds. The games we considered are MLB first 5 innings betting. Betting odds represent that underlying chance of an outcome happening and the possible payout you can get if you happen to win the game. If the predicted probability for an outcome differs from the implied probability of the odds, positive expected returns are identified. In this article, we will predict the runs of baseball matches. The expected runs and implied probability to win are estimated from the Markov process method and the runner advancement model.

並列關鍵字

Betting Odds ； Implied Probability ； Runner Advancement Model ； Markov Process ； Expected Return