Data Science in Baseball

[masas-cave-blog]

Two of the best hitters in the history

The below graph shows the transition of the hits made by Ichiro Suzuki (2004) and George Sisler (1920). *Day 1 = 6th April

Sisler had accumulated more hits than Ichiro in middle of the season but Ichiro accelerated to yield more towards the season end and became No.1.

[masas-cave-blog]

Home Run Rate of each MLB team (2019 Regular Season)

There were 131 qualified batters this regular season, and the below picture shows the Interquartile range of the home run rate (numbers of home runs divided by at-bat) for each MLB teams.

LAD is of relatively high while SF, PIT and MIA are of relatively low. Regarding DET, Miguel Cabrera was the only one qualified batter, so there is not range.

There is no remarkable feature in that of WSH, this year’s World Champion: it looks very average. 

[masas-cave-blog]

T. Testing whether there is any difference in Hits, HR, RBI and SB among NYY, HOU and SD.

From the records of 2019 MLB Regular Season, checking whether there are any differences in terms of Hits, HR & RBI par AB and SB par G between NYY & HOU (two of the best teams in the league) and NYY & SD (one of the best & one of the worst teams in the league)

1) How to check

(1) Used t.test() in R

(2) Null Hypothesis is “there is no difference for each factor between the two teams”

(2) Significance Level is 5% (2.5% at each side)

2) Result

(1) NYY & HOU: Null Hypothesis was not rejected for all factors. There are no difference for those factors between NYY & HOU.

(2) NYY & SD: Null Hypothesis was rejected for HR & RBI and there are difference for the two factors between NYY, one of the most offensive teams in the league and SD, which won 30+ less games & scored 250+ less than NYY.

   #Two Sample t-test of HR / t = 2.1817, df = 28, p-value = 0.03768

   #Two Sample t-test of RBI / t = 3.5719, df = 28, p-value = 0.001307

[masas-cave-blog]

Data (HRs, SOs, Hits, Runs and Stolen Bases) from MLB History (since 1871)

The first is the relationship between HR par AB & SO par AB. the below plot shows that the more HR, the SO.

The second is the relationship between Hits par AB & Strikeouts par AB. The more H/AB, the less SO/AB. Those batters who can keep the good avg. tends to strike out less.

The third one is Run par Game & SB par Game. I wanted to know whether a number of Runs is related to ones’ ability to run fast. From the below plot. It can be read that the more SB (=if one can run fast) the more Runs.

[masas-cave-blog]

Batter’s Ranking in NPB (2019 Regular Season)

Here is the ranking of 60 batters from NPB 2019 Regular Season, who had 443+ PA (143 Games x 3.1 = a number of PA required to be qualified for Batting Average title), covering  (i) Batting Average, (ii) RBI, (iii) Stolen Bases, (iv) OBP and (v) OPS and (vi) the average of those 5 indexes.

<Scores>                                                                                                            >50 : Average                                                                                                

>40/60 (-10/+10) : Better/Worse than average (for 1 Standard Deviation /SD)    

>30/70 (-20/+20) : Much Better/Worse than average  (for 2 SDs / within the best/worst 2.5% of all batters if it comply with Standard Normal Distribution)

No.1 is S. Suzuki, Hiroshima Carp. He is the only batter who scored 70+ for multiple indexes (AVG, OBP and OPS) and the average of 5 indexes. 

4 Batters, T. Mori (Seibu Lions), T. Yamada (Yakult Swallows), M. Yoshida (Orix Baffaloes) and H. Sakamoto  (Yomiuri Giants), scored 60+ in the average of 5 indexes. 

28 out of 60 scored 50+ in the average of 5 indexes (consequently other 32 did not.) 

[masas-cave-blog]

Which Hitting Record Contributes Most to Wins

I was wandering which hitting records contributes most to team’s wins. Then I looked up 4 indicators of all 30 MLB teams from the 2019 regular season record,  which are OPS, HR, RBI and AVG, and researched which of them has the strong correlation to a number of wins. The result is as shown in the below plots:

OPS / Correlation 0.83 :  OPS is widely used to evaluate one’s hitting ability, and at the same time it has the strongest correlation to a numbers of wins. To win more games, a team needs to line-up higher OPS hitters.

HR / Correlation 0.74 : HR has a strong correlation to wins but weaker than OPS. A balanced hitter, with a good OBP & SLG rate, would contribute more to wins than a HR hitter with low OBP.

RBI : Correlation 0.65 : It was surprising that RBI has lower correlation rate than OPS & HR because, as we can easily imagine, the more RBI means the more scores, which would make a team win more. I need to study data deep in order to understand this.

AVG : Correlation 0.55 : AVG has only weak correlation to wins. A good hitter, say around .300, looks well in the stats but he may not contribute to a team’s win more than we imagine.

[masas-cave-blog]

RBI & OPS for Best & Worst 8 Teams in 2019 MLB Regular Season

The below picture shows RBI & OPS for each of best 8 teams, which advanced to Playoffs, and the worst 8 teams in 2019 MLB regular season. 

Basically, the best teams, those in blue, have more RBI & better OPS than the worst teams in pink. One exception is that both RBI & OPS of STL (.562) is worse than those of KC (.364), but they still has accumulated 32 more winning games than KC.

It’s interesting that MIN, which ended up in the 1st place in AL Central, had 300+ fewer scores than ATL, WSH and NYY but MIN recorded better winning % (.623) than ATL (.599) and WSH (.574).

To win more games, of course, the combination of good defense & offense is important; however, it’s clear that they have to be good at hitting to some degree in order to win the games.

[masas-cave-blog]

I learned how to use “function()” in R today, which made it possible to shorten the scripts to one-quarter. The below attachments are the previous script (approx 200 lines, way too looong) for the above blog AND the new script using  “function()” (approx 50 lines). I wish I would have known  “function()” before...

<Previous One with approx 200 lines>

<New one with function (), approx 50 lines>

[masas-cave-blog]

RBI & OPS for Best & Worst 8 Teams in 2019 MLB Regular Season

The below picture shows RBI & OPS for each of best 8 teams, which advanced to Playoffs, and the worst 8 teams in 2019 MLB regular season. 

Basically, the best teams, those in blue, have more RBI & better OPS than the worst teams in pink. One exception is that both RBI & OPS of STL (.562) is worse than those of KC (.364), but they still has accumulated 32 more winning games than KC.

It’s interesting that MIN, which ended up in the 1st place in AL Central, had 300+ fewer scores than ATL, WSH and NYY but MIN recorded better winning % (.623) than ATL (.599) and WSH (.574).

To win more games, of course, the combination of good defense & offense is important; however, it’s clear that they have to be good at hitting to some degree in order to win the games.

[masas-cave-blog]

World Series W-L Relation

According to the WS record since 1903 to 1993 for 90 years, the following 3 things can be read:

1) When a team wins the 1st game, it's 66% probability to win the series.

2) When a team loses the first two games, it's only 9% to win the series.

    If those 2 games were at home, it becomes 2%, while 8% at away.

    This means that it's crucial when a team loses the first 2 games, especially at home games. 3) No team between 1903 to 1972 had lost the first 3 games in a row and won the series.

L= loss at home l= loss on road W= win at home w= win on road 1903  LWLlwwwW  1927  wwWW  1950  wwWW  1973  WLwllWW 1905  wLwWW  1928  WWww  1951  LWlwwW  1974  wlWWW 1906  wLwLwW  1929  wwLWW  1952  lwLWLww  1975  lwWLWlw 1907  WWww  1930  WWllwW  1953  WWllwW  1976  WWww 1908  wWLww  1931  LWwlwLW  1954  WWww  1977  WLwwlW 1909  WLwlWlw  1932  WWww  1955  llWWWlw  1978  llWWWw 1910  WWwlw  1933  WWlww  1956  llWWWlw  1979  lwLLWww 1911  lWwWlW  1934  wlWLLww  1957  lwLWWlw  1980  WWllwW 1912  wLwWlLW  1935  LWwwlW  1958  llWLWww  1981  llWWWw 1913  wLwWw  1936  lwWWLw  1959  lwWWLw  1982  LWwllWW 1914  wwWW  1937  WWwlw  1960  WLlwwLW  1983  LWwww 1915  lwWWw  1938  wwWW  1961  WLwww  1984  wlWWW 1916  WWlwW  1939  WWww  1962  wlWLWlw  1985  LLwlwWW 1917  WWllWw  1940  LWlwlWW  1963  wwWW  1986  LLwwlWW 1918  wlwWLW  1941  WLwww  1964  WLlwwLW  1987  WWlllWW 1919  WWlwwLLw  1942  LWwww  1965  llWWWlw  1988  WWlww 1920  wllWWWW  1943  WLWww  1966  wwWW  1989  WWww 1921  LlWwLwWw  1944  LWlwwW  1967  wlWWLlw  1990  WWww 1922  WWwW  1945  LWLwwlw  1968  lwLLWww  1991  WWlllWW 1923  LwLwWw  1946  LWlwlww  1969  lwWWW  1992  lwWWLw 1924  LWlwlWW  1947  WWllwLW  1970  wwWLW  1993  WLwwlW 1925  LWllwWW  1948  lwWWLw  1971  llWWWlw 1926  lwWLLww  1949  WLwww  1972  wwLWLlw

The above data from StatLib of Carnegie Mellon University (http://lib.stat.cmu.edu/datasets/wseries)