Over at Grantland, Jonah Keri has done an admirable job of running a weekly power ranking, of sorts. He’ll typically create some different tiers and breakdown a team from each level to give the casual fan a better understanding of what is affecting each franchise. This is great for those that typically only follow one team, because it gives that person something to talk about when they end up chatting with fans of other teams. Well, Jonah focused on the Rays in the most recent article. He got around to mentioning something that he has gone over before that many fans can easily agree with because it’s just so intuitive. That idea is the notion of “clusterluck”.
Jonah isn’t the first, of course, but he’s a vocal and respected leader with great connections so he’s probably the initiator for many folks’ first foray into this concept. The link above does a pretty good job of giving the gist and here’s a few more from Dave Cameron whom wrote a couple of really good posts earlier this year looking into this concept. They’re all pretty good and build upon one another so if you have some time you should check those out.
The basic concept is that, in the aggregate, using runs scored and allowed works pretty well, but due to the timing of events smaller samples can be deceiving. An offense estimator like wOBA does a better job of portraying what should have happened due to the timing of events. Sometimes a double drives in three runs, sometimes that guy ends up stranded on 2nd. There is a wealth of lost information in between so it makes sense to use the average number of runs scored to give a better idea of what should have happened. To that end I’m using wOBA values from The Book for each event:
BB: .72, HBP: .75, 1B: .90, ROE: .92, 2B: 1.24, 3B: 1.56, HR: 1.95
I do not differentiate for double plays or fielder’s choice. An out is an out and carries a value of 0.00. I’m also creating a baseline using the matchup tool created by myself and Ian Malinowski to give an idea of where the Rays were facing better competition. You’ll notice gaps where, on paper, you’d expect the Rays to be much better or worse than they’re opponent, so it’s always interesting to compare reality to the expectations. Let’s start by looking at the Rays expected and actual wOBA. All following data through July 29th:
These are 10-game trends to cut down on some of the noise. The blue lines show the output of the matchup tool with the dotted line being for our pitchers. The red lines show actual wOBA with the dotted line, again, being used for our pitchers. Readily evident is the looooong stretch of allowing a higher wOBA than we posted, but you can see where that turns over around the mid-60’s and carries through basically the rest of the season.
Now this is wOBA which is really cool, but it doesn’t really cut to the heart of scoring runs, per se. Well, the beauty of wOBA is that we can easily convert these figures to runs by calculating wRAA ((wOBA-lgwOBA)/1.15*PA) and then we can just add that to league average runs per game (4.14) to get an idea of how many runs should have been scored for either team. To cut down on the clutter I’m going to separate offense from run prevention and start with the hitters:
Again, we see our expectation in blue and in this case aRuns is the amount of runs we would expect extrapolating our actual wOBA. The black line shows the number of runs actually scored in the game. Where red differs from blue you can consider that over/under performance from expectation, but where black differs from red is where we see evidence of cluster luck. This mostly tracks along, which is what you should expect, except for the grand departure from games 60 – 80 where we see aRuns start to trek upward due to better wOBA figures, but actual runs scored in the game sits well below what we think the team should have scored. Let’s flip over to the pitching side:
On the runs allowed side of the equation we see some poor luck early which went the other way for awhile, tracks along, and then a pretty bad stretch, again, between 60 and 80 where performance improved drastically, but runs allowed took some time to catch up to that level. We certainly see more variation here as compared to hitters which matched up quite well for the most part. Putting these two parts of the equation together will give us an idea of how these things balanced out so let’s look at run differential across these three different metrics:
The blue line tells us what the matchup tool thinks and you’ll notice that it generally likes us across the board other than that tough stretch early when the team’s resources were depleted. You can see an awful lot of time spent in the negative for both aRuns and actual Runs, but again we see that performance improved before seeing an actual turn around in run differential. This lag period could be an indicator or precursor for a team’s changing fortune, or it could be coincidence, would be interesting to see more of this stuff. Now that’s cool and all, but it’s tough to get an idea of impact here so let’s take a look at the cumulative of this stuff:
The expectations like the team to be around +20 by run differential this late in the season and took more of a slow roll to get there. We know that reality features more bumps in the road and that’s apparent in the aRun differential line. Early on the Rays were playing much better than the actual runs scored and allowed, the black line, tough we do see the massive dip following the injuries and poor performance. We also see the team pulling out of their tailspin around the 70th game and they haven’t looked back since. The neat byproduct of this research is to see actual run differential mirror our wOBA run differential, but the gap starts early and like compound interest that early divergence is carried throughout the entire season.
The Rays performance throughout this season was better than what was reflected on the scoreboard, and while the latter is all that matters in reality, we should always be keeping an eye on the performance because in the long run actual runs will end up approximating what the wOBA figures tell you. The season may be lost, but the Rays played much better than they will get credit for outside of one prolonged, and injury-fueled bout of poor play. For those that want to see what this stuff looks like modeled together on one chart: