# Gauging the First Half

Instead of making a post about mid-season awards, which we are sure to see a few of during the All-Star break, I figured I’d try something different. Let’s take a look at how individual plays affect a team’s postseason probabilities.

# Top Plays

Earlier in the season, I added a page that shows the top plays of the season in terms of win probability added. While placing a value on the importance of the individual game is interesting, we can take it one step further and look at how each play impacts a team’s playoff probability. A big hit in a game between two teams that are not in contention will have little to no effect. But a walk-off home run in a game between teams tied for the lead in a division will have a much greater impact. So let’s take a look at the biggest plays of the first half.

This list is sorted by championship win probability added (cWPA). Just as in-game win probability added shows the change in win probability in terms of percentage points, cWPA shows the change in World Series win probability. Your first thought upon seeing the cWPA values is probably how small they are. In fact, every play this season has a cWPA of less than 1 percentage point. This shows just how little of an impact, even the most important play of the first three months of the season, has on a team’s chances of winning the World Series. Another way to look at these numbers is to multiply them by 8, to see the change in probability of being one of the final 8 playoff teams.

## 1) Leonys Martin’s walk-off HR (0.71 cWPA)

With two outs and a runner on 2nd in the bottom of the ninth and his team down by a run, Martin fell behind in the count 1-2 on three Ryan Madson changeups. On the fourth pitch, Madson went changeup again and Martin deposited it into the right field bleachers. The walk-off increased the Mariners in-game win probability by 86 percentage points, but more importantly, it increased the Mariners probability of winning the World Series by 0.71 percentage points.
On a side note, this play is also 40th on our list, as it decreased the A’s World Series win probability by 0.38 percentage points. The change in percentage points is bigger for the Mariners since the game was of more importance, as they were ahead in the division by 1.5 games, while Oakland was 7 games back.

In the bottom of the 8th inning with two outs, Bryan Shaw was looking to send the game to the 9th with his team up by a run. He was facing Salvador Perez, who was 1 for 12 in his career vs Shaw. But on the 1st pitch, Perez gave the fans in left field a souvenir and his team the lead. This play decreased the Indians World Series win probability by 0.63 percentage points. You can actually see the moment when Bryan Shaw realizes that pitcher vs batter stats are too small of samples to trust.
This play is also 6th on our list, as it increased the Royals World Series win probability by 0.59 percentage points.

## 3) Ian Desmond’s go-ahead 2-R HR (0.62 cWPA)

The next two plays on this list are from the same crazy game in Oakland. The A’s were one out away from victory with Ryan Madson on the mound, when Ian Desmond gave Texas the lead with a 2-run HR off a changeup. This play increased the Rangers chances of winning the World Series by 0.62 percentage points. As Desmond rounded the bases, Rangers announcer Tom Grieve noted that Madson threw one too many changeups, which seems to be a recurring theme here.
This play is also 24th on this list, as it increased Oakland’s World Series win probability by 0.41 percentage points.

## 4) Khris Davis’s walk-off grand slam (0.61 cWPA)

The next half-inning, Texas closer Shawn Tolleson intentionally walked Josh Reddick to load the bases with one out. Next, Danny Valencia flew out to shallow right, which brought up Khris Davis, who had already hit two home runs in the game. Davis then ended the game on a walk-off grand slam, which left Adrian Beltre wondering “what the hell just happened”.

From the A’s perspective, this play is 27th on the list, as it decreased their World Series win probability by 0.40 percentage points.

## 5) Yasiel Puig’s walk-off single and error by Michael Taylor (0.60 cWPA)

Puig’s single would have put runners on 1st and 2nd with one out in the inning, but it was Taylor’s gaffe (and Puig’s hustle) that allowed both runners to score. This was the culmination of Michael Taylor’s horrendous game, where he also struck out in all five of his at bats. If you look close enough, you can see him calculating the cWPA in his head.

Here are the rest of our top 25 plays of the first half:

Rk Date Play Team VS cWPA Highlight
6 6/14 Salvador Perez HR KC CLE
+0.59
7 5/21 Matt Wieters HR BAL LAA
+0.58
8 7/08 Luis Valbuena HR HOU OAK
+0.56
9 6/22 Yasiel Puig little league HR WAS LAD
-0.56
10 6/12 Jayson Werth Single WAS PHI
+0.56
11 5/14 Albert Pujols HR SEA LAA
-0.52
12 6/05 Matt Wieters 1B BAL NYY
+0.47
13 7/07 Troy Tulowitzki 1B TOR DET
+0.47
14 4/12 Geovany Soto HR OAK LAA
+0.45
15 5/20 Melvin Upton HR LAD SD
-0.44
16 5/10 Ryan Rua HR TEX CHW
+0.43
17 4/12 Geovany Soto HR LAA OAK
-0.43
18 4/08 Starling Marte grand slam PIT CIN
+0.42
19 4/08 Starling Marte grand slam CIN PIT
-0.42
20 6/24 Adam Lind HR STL SEA
-0.42
21 5/21 Jayson Werth GIDP WAS MIA
-0.42
22 5/28 Drew Butera 2B CHW KC
-0.41
23 6/23 Adonis Garcia HR NYM ATL
-0.41
24 5/17 Ian Desmond HR OAK TEX
-0.41
25 6/11 Prince Fielder HR SEA TEX
-0.41

# Most Critical Moments

We can measure the importance that a particular play has on a game by using leverage index (LI), but this is limited to the situation in the game and it treats all games the same. Just as with WPA and cWPA above, we can take this one step further and measure the importance of the game by including the game’s championship leverage index (CLI). This number shows the importance of the game for each team, where the average game equals 1. If a win or a loss has a significant effect on the team’s playoff probability, the CLI will be greater than 1. By multiplying the LI and CLI, we can measure the importance that a play has on a team’s playoff probability. We’ll call this number pCLI (for championship leverage index by play). This number can be read as “how many times more important this situation was compared to the average play on opening day”.

Below are the top 10 most critical situations of the first half. As with the list above, some plays will appear twice, since they were important to BOTH team’s playoff chances.

Rk Date Team Inning Outs Runners Score pCLI Outcome Highlight
1
5/17 TEX Bot 9 2 Loaded 5 – 4 15.4 Khris Davis grand slam
2
6/12 WAS Bot 9 2 Loaded 4 – 3 14.2 Jayson Werth 1B
3
6/05 WAS Bot 9 2 Loaded 10 – 9 14.1 Ivan de Jesus fly out
4
6/24 TOR Top 9 2 Loaded 2 – 3 12.7 Michael Saunders pop out
5
6/11 SEA Bot 11 2 1st & 2nd 2 – 1 12.5 Kyle Seager fly out
6
5/17 TEX Bot 9 1 Loaded 5 – 4 12.2 Danny Valencia fly out
7
7/07 TOR Bot 8 2 Loaded 4 – 3 12.1 Troy Tulowitzki 1B
8
6/11 TEX Bot 11 2 1st & 2nd 2 – 1 12.0 Kyle Seager fly out
9
6/11 SEA Bot 10 2 Loaded 1 – 1 11.8 Ketel Marte fly out
10
5/06 BOS Top 9 2 Loaded 2 – 3 11.7 Hanley Ramirez strikeout
11
6/10 SFN Bot 9 2 1st & 2nd 3 – 2 11.5 Brandon Crawford strikeout
12
7/06 HOU Top 9 2 Loaded 8 – 9 11.5 Dae-Ho Lee strikeout
13
6/10 LAN Bot 9 2 1st & 2nd 3 – 2 11.5 Brandon Crawford strikeout
14
6/11 TEX Bot 10 2 Loaded 1 – 1 11.3 Ketel Marte fly out
15
6/05 WAS Bot 9 1 Loaded 10 – 9 11.1 Zack Cozart strikeout
16
6/05 BAL Bot 8 2 Loaded 1 – 0 10.9 Matt Wieters 1B
17
6/05 BOS Bot 9 2 1st & 2nd 5 – 4 10.9 Marco Hernandez strikeout
18
6/30 NYN Top 9 2 Loaded 3 – 4 10.8 Javier Baez pop out
19
6/18 TOR Top 9 1 Loaded 2 – 4 10.6 Josh Donaldson GIDP
20
6/21 BAL Bot 8 2 Loaded 7 – 6 10.6 Adam Jones ground out

If we revisit these lists at the end of the season, there is a good chance it will be dominated by second half plays. The reason for this is, just as the most important plays happen in the later innings of the game, the most important games occur near the end of the season. However, 2016 may be different since 5 of the 6 division leaders currently have at least a 5 game lead, which may lead to less enjoyable divisional races. For the sake of exciting plays and games, let’s hope some of these leads shorten.

## 5 thoughts on “Gauging the First Half”

1. Simon says:

It would be interesting to sum the season total of WPA x cLI for each player, which would give you a nice ranking of championships added, which in turn would be good fodder for MVP discussions.

Would also be interesting to present standard seasonal batting/pitching stats weighted by pCLI. If two MVP candidates both had 35 home runs, but player A hit a lot of late-game bombs in key divisional games, while a lot of player B’s HRs were hit at the tail end of 9-1 losses with his team out of the race, then this method might adjust player A’s HRs up to 50 and player B down to 20.

It might be distorted by high-impact plays (eg the Khris Davis grand slam would count for 15.4 home runs), but it would be a good way of expressing standard stats in a way that many people believe the MVP is supposed to award.

2. Simon says:

Hi Dan, I really like your rankings of postseason series based on average championship leverage index. I thought maybe the same concept could be applied to the regular season. Wondering:

1. Which teams in baseball history have the highest average cLI in the regular season (ie averaging the cLI for each of the team’s 162 games)? I’d guess some team that won its division by a single game and was never ahead or behind by more than 2 games the entire season.

2. Which regular seasons as a whole have the highest average cLI for all teams (ie averaging the answer to question 1 for all of the teams in a given year)?

3. And on a slightly different note — does the average cLI for all teams steadily increase through the season, or is there a “hump” at midseason (ie averaging the cLI for all games played in history on June 15, June 16, June 17 etc — on what date (or game #) does the average cLI typically peak)? It seems that it would steadily increase over the course of the season, but I’d imagine high-impact late-season games are offset by many games with a 0.00 cLI between teams out of contention.

Thanks again for all of the very interesting information on your website.

1. I can’t answer these just yet, as I’m currently simulating previous seasons. I hope to be done in the next couple of months (it takes a while), so I can do this type of analysis.

As for #3, the highest average cLI would probably not be towards the end of the season, since so many teams are eliminated. However, the end of the season is where the biggest cLI games in history would be. I would imagine the highest average cLI day would be in the first half of the season, but it’s also very unlikely to have an extremely high cLI game during that time.

I hope to put something together once I finish all my simulations.

3. Simon says:

Hi Dan, I look forward to the results of your sims.

I made a crude attempt at determining which team had the most exciting season in 2016. Rather than use championship leverage index, I used the absolute value of change in playoff odds using the game logs on this site. If a team’s playoff odds increased from 30% to 35%, then it would receive .350 – .300 = .050 “points”. This isn’t exactly championship leverage index but it should come pretty close. Here are the 30 teams ranked in order of total “points”:

7.373 DET
7.260 BAL
7.090 SLN
6.948 TOR
6.477 NYN
5.916 BOS
5.747 SEA
5.311 SFN
5.260 MIA
4.972 HOU
4.959 PIT
4.456 LAN
4.203 KCA
4.059 NYA
3.781 CLE
3.297 TEX
3.195 WAS
3.145 CHA
2.931 COL
2.373 PHI
1.884 CHN
1.779 ARI
1.754 TBA
1.689 LAA
1.676 MIL
1.629 OAK
1.360 SDN
1.076 CIN
0.622 MIN
0.550 ATL

The next step would be to compare the average “points” this season (3.759) to other seasons, to see whether 2016 was more dramatic than other years.

I also tried adding up total “points” on the game# for each team. The results aren’t quite as interesting — a few big days late in the season (games 147, 151 and 155) and then a scattering of games early in the year before any teams really fell out of the races.

0.999 151
0.889 5
0.877 155
0.86 17
0.851 32
0.846 16
0.846 15
0.843 10
0.842 13
0.842 8
0.841 9
0.836 14
0.833 7
0.832 42
0.832 2
0.83 147

1. Excellent. I think we are on the same page here. This is pretty much what I want to examine once I have the data for all seasons.