I think I successfully did a “Find and replace” for 2018 to 2019 and the various plotting apps/charts here will now compare 2019 stats to 2017. That was easy, comparing to both 2018 and 2017 wolud be harder. We’ll see if I can get around to that someday. Anyway, none of the charts are particularly interesting 5 games into the season - though it’s safe to say our pitching is still pretty good.
The stat graphs all still work. I don’t see updating any for the post season, but I can proabbly stop them from updating every morning now, can’t I? 103 wins, like 4 posts? Not the best ratio! I do tweet more often.
I wrote a few weeks ago about the Mariners and Astros “luck” relative to their runs scored / runs allowed. Since then, the Astros won 11 straight and opened up a 4 game lead or so on the Mariners. But, then since then the Astros dropped 3 of 4 to Tampa and the Mariners won 7 straight (including 3 by 1 run!), and now the division race is back to a half game and the madness continues:
Remember, those red dots are every other team in the liveball era. And it’s not just the Mariners are historically over-performing their modest run-differential, the Astros “should” have won about 5-6 more games than they have. It’s a perfect storm of opposite luck that’s keeping the AL West race alive and kicking on July 1.
Even in their respective last 10 games, while going 5-5 the Astros run differential is +9, the Mariners are 8-2 and their run differential is +8.
Now, we have the 2018 Mariners. The Mariners are 21-9 in one-run games. For this Astros’ fan, it’s “Oh no, not this again.” While The Ringer finds this charming, it has Astros’ fans shaking their heads. (Note: I wrote this post before I saw the Ringer piece, and then added that sentence and the link.) I mean just look at this:
That’s every team’s season since 1920, with our AL West friends highlighted. Now, one-run records lead to exceeeding your “expected” wins. Because those formulas use total runs scored and runs allowed, so winning close games, and losing non-close games, will make a team overpeform their expected wins. Take a look at the adjusted Baseball Prospectus standings.
The Mariners are exceeding their expected win percentage by .101! That’s 16 extra wins over a full season. (I’m using pythagenpat, or the BB reference first order expected win perentage). That would shatter the 162 game record that the 2016 Rangers hold (.081, or 13.2 “extra” wins). (The 1981 Reds playing in a strike-shortened 108 game season hold the record for “luck” of .086 above expected win percentage. Then they missed the playoffs anyway, because the division winners went to top teams in the first-half and second-half). So maybe they weren’t so lucky.)
That’s not all! If you looked at those Baseball Prospectus adjusted standings you would see the Astros are at the bottom. So, the Mariners are exceeding their expected wins by 6.8 games while the Astros are 5.1 games below their expected wins. This combines for a difference of 12 games in the standings!
The Astros bad “luck” of -0.075 if it lasted until the end of the season would be the third worst of all time going back to 1920 - behind two bad teams such that no one was that concerned about their run differential: the 1993 Mets (59 wins, expected 73), and the 1986 Pirates (64 wins, expected 77). Look at the plot from above with the Astros added: (Apologies the shade of red changed.)
Thankfully, even with this undperformance relative to runs scored/allowed the Astros are still on pace to win 102 games, based only on their current actual win percentage. I suspect/hope over the course of a long season, these numbers even out. Here in June though, yikes. Some of the data behind these plots below.
Standings and Expected Win Percentage as of games through June 12
Liveball Teams (1920 and later) With Absolute Value of “Luck” greater than .06
Our offense really is hit or miss. After AstrosCounty noted in today’s link list that the 2018 Astros have had a lot of games with 5 hits or less, I decided to make a histogram. It turns out density plots look a little better to me, and are normalized - but it’s like a smoothed, normalized, histogram. Peaks on the graph mean more games with that many hits. So, we have a lot of games where the Astros got around 5-6 hits. Behold the Astros’ bimodal hit distribution:
Look at that ridiculous peak in this density plot at 5 hits/game. Our average hits per game isn’t that different than last year, but the distribution of hits across our games in 2018 is double peaked. pic.twitter.com/5WN9LRYp2p
Maybe the offense is just strugling, regardless of at home or road? The Astros are averaging 1.76 more runs per game on the road than at home, and we’ve all been waiting for things to “average out” - i.e. for the home numbers to improve.
However the start of the road series visiting the Angels led to only the Astros’ third loss when giving up 2 runs. They’re 23-4 when pitchers allow 2 runs or less. Yet, someone it sure doesn’t feel like it. Astros managed a meager 6 hits and 1 run.
I want to see this plot flatten out by more runs in the valley - not by the peaks getting smaller. I assumed the trick to averaging out this graph was the offense firing up when the pitchers give up 3 or 4 runs. Not that they’d start scoring less when giving up 2.
Here’s the thing about the Astros - despite a maddenly inconsistent offense, they’re still 26-17 with a pythagorean projected record of 32-11. Assuming the starters keep this going, the bats will wake up and the Astros’ best baseball in 2018 is ahead of them.
Only Astros who seem to have it together at home are Reddick and Fisher. I won’t call Jake Marisnick’s 61 WRC+ at home “having it together” though it is better than his -11 on the road.
So, with all those players not hitting at home, no wonder we have arguably the worse offense in baseball when we play at Minute Maid (WRC+ really punishes the Rockies, for example. Their home wOBA is better than ours but the park effects correction makes our WRC+ look better)
Now this is 20 home games out of 81. I’m … somewhat confident they get it together in Houston at some point. The offense is clearly capable of doing well. Maybe they should just start closing the roof all the time…
I posted this image on Twitter the other day, but it doesn’t tell the whole story. The averages make the offensive lull look worse than it is.
If you look at each game, you’ll see our missing offense when we allow 3-4 runs is just small number statistical noise. That’s going to fill in, and we’ll be far less frustrated because the Astros will win some games 6-4, instead of losing 4-1.
AJ Hinch said this the other night (in a videa clip I no longer find) after the Astros 16-2 win, that the offense was streaky or inconsistent or words to that effect. And it’s not anyone’s imagination. This rolling wOBA average graph is quite the rollercoaster, especially compared to last year at the same time. I made this a shorter 5-game rolling average so the ups and downs are more prominent than the 10-day. Try looking at all sorts of stats like this in the one of my interactive Astros web apps
Cole’s ridiculous 1-hit complete game shutout calls attention to our absurd rotation. I shared the plot below on Twitter, and have now added a starter/bullpen breakout option to the Scatter Plot interactive page.
The 2017 Astros were 22-11 through 33 games, the 2018 Astros are two games off that pace at 20-13. The Astros currenly have a pythagorean record of 23-10, according to Baseball Reference. This prompted me to add Pythagorean win percentage to the cummulative year to date web app. The 2017 Astros pythagorean W-L was 20-13 at this point (20.48 wins, to be more precise.)
So, while the offense is struggling, so far the resevoir of good pitching is keeping the 2018 Astros ahead of last year. I note that Pythogrean W-L rewards lower scoring wins - winning consitently 5-2 is much better than winning consistently 9-6, even if the run differential is the same. Of course, losing a lot of games won’t help any of these things, nor the only stat that matters - the Astros record.
The data hasn’t updated yet with the calamity that was Thursday’s loss, but yes - the offense is bad. (Update Actually this is going to update randomly whenever I update the data here, which is not the same as when I updat the data in the apps. Using R markdown documents that render on the fly may not have been a great choice) The offense didn’t have very many 10 game stretches this bad in 2017. Their wOBA is .300 for the 10 games ending May 2. Though, I think the wOBA should tick up once I get the game log for the May 3 loss, the Astros at least scored more than 0 runs….
You can tweak this plot interactively on the Rolling Averages app. Right now a variety of plotting apps all appear on the same document but I plan to separate them out soon.