You once had a friend who enthusiastically recommended watching a TV show and said, “It takes a few episodes to get going, and the timeline gets weird at the end, and one or two of the main characters might be annoying. , but other than that it's GREAT.” And at first you might be put off, thinking that a really good show wouldn't need that many players. Sometimes you're right about that, but sometimes it turns out that the show is Parks and Recreation and even though the first season is about as interesting as living in a dungeon, the rest of the game is absolutely fantastic.
Sometimes small parts of a larger body of work do a poor job of representing the work as a whole. The oddity that occurs in small samples is probably not a new concept to FanGraphs readers, and it won't shock anyone to note that what constitutes a small sample depends on exactly what we want to measure. Recently, the good folks at MLB Advanced Media gifted us with a few metrics using Statcast's bat tracking technology. Every time we dig up a new metric, we must consider the appropriate serving size to satisfy our hunger for information, lest we find ourselves producing things we'll regret later.
In this article, we will try to determine the appropriate threshold sample for measuring bat bat speed; so that non-bat players don't feel left out, we'll do the same for the sword level from a pitcher's perspective. For most metrics, sample size is measured in pitches or plate appearances, but since both bat speed and swing rate are directly related to bat movement, their samples will be made up of swings. To determine reasonable sample sizes, I used the split-half method. The idea is to randomly select two samples of size X from a player's swing collection, calculate the player's average bat speed or swing average for both samples, lather/rinse/repeat a bunch of players, then take the full set of two player-sample pairs of all players and see if they correlate well how much. We complete the test by repeating the process for progressively larger sample sizes. And to be more careful, we will repeat the experiment several times and measure the associated values.
The theory behind this approach is that with large enough samples, the metric will contain more signal and less noise, thus representing the player more accurately. Therefore, two samples of sufficient size should be visually comparable to each other. Once we reach a sample size where the correlation is strong enough that the metric is considered what statisticians call “reliable,” that sample size becomes our lower limit for relying on the explanatory power of the metric. Six negative episodes shown since Parks and Recreation in its first season it did not end up providing a large enough sample to accurately reflect the quality of the series' episode. We needed to see more of the Pawnee people.
Starting with bat speed, the chart below shows the results for each test (in gray) and the average of all tests (green), with sample sizes on the horizontal axis and the corresponding correlation coefficient on the vertical axis. Statistical standards indicate that if the correlation has risen above 0.8, we are in good shape. With that in mind, the output suggests that average bat speed becomes a reliably descriptive metric at around 30 swings, which most players pile up over 20ish plate appearances.
To emphasize the importance of 30 swings, I decided to find the wackiest 20-swings in this metric's short life so far. By wacky, I just mean 20 swing lengths where a player's average speed is significantly different than his season average. Leading the leaderboard is Ildemaro Vargas, who earned his spot by attempting five of six consecutive pitches spread over two games on July 4 and July 5, leaving him with an average batting velocity over 20 swings that was 20 mph slower than his season high. average 69 mph. The first four bunt attempts were split evenly between the two PAs on July 4, when Vargas came up with a runner on first and no outs (the old bunting situation). On July 5, Vargas pinch-hit to start the bottom of the 11th with a second zombie runner (the classic modern version of bunting). His last effort registered a bat speed of 9 mph, which looks like this:
The Vargas example highlights an important aspect of calculating bat speed. Per Baseball Savant: “A player's 90% fastest swing, and any 60+ MPH swing that results in an exit velocity of 90+ MPH, is considered his 'competitive' swing. The average of this swing is equal to his season.” It's possible that complex logic is used in the background, but as far as I can tell, nothing is left of the check exchange, bunts, bad tips, etc. In addition, the mid-season field tests I calculated against Savant's bat. The fast leaderboard is well matched.
To me, this means that the calculation is very dependent on removing the bottom 10% of the turnover to remove these honest contributions. And in a sample of 50 swings, a bunting spree à la Vargas will be cut off (admittedly this bunting combination is unusual), but 10% of the 20 swings are only two swings, so three more attempts, and any other non-binding cycles. , sit down and count. To judge Vargas based on these 20 swings would be a judgment call The phone based only on the second season (which I liked, but many did not). Vargas went into bunting for a while, while The phone fully into the story line of the stevedores, patterns of behavior that end up being permanent.
While Vargas was hurt by a number of bunt attempts, others were frustrated by their check swing habits. Juan Soto is known for his knowledge of the strike zone and patience at the plate, but this means he likes to gather as much information as possible before committing to a swing, often pulling his bat back at the last second. During two games against the Mariners and their good pitching at the end of May, Soto pulled his bat back seven times, cutting swings at low bat speeds, and dragging his 20-swing average 15 mph below his full-season number. The “swing” below registered a bat speed of 10 mph, and because you checked, it also earned him a walk:
The TV coverage of Soto's rough 20-swing may be Ross' heavy hitter Friendsthat is, a good/hitting show that gives too much emphasis to an annoying character or character.
Fernando Tatis Jr. and he's a great pitcher, but during a series against the Mets in mid-June, a few abandoned swings mixed with foul balls dragged his small-sample bat speed 19 mph below his peak. – a sign of the season. The foul ball shown below resulted in a closed swing at 43 mph:
The anomaly of the Tatis 20-swing sample can be considered as a seasonal episode of gas leaks Communitywhich, after parting ways with the original creator, still looks like the same show only with poor execution, resulting in mishits and uncertain decision making.
Moving on to the measurement of the sword, finding an adequate sample size was a difficult ask, mainly because the correlation graph (which you can see below) resembles television static from back when TVs were big boxy things; if the cable breaks, you're left with nothing to watch but black and white chaos. Here we do not see a gradual improvement as the sample increases; the correlation is up to 0.2, shy of the target of 0.8:
This analysis suggests that getting hits at a consistent rate is not a reliable skill for pitchers, at least not given the currently available samples. Perhaps if we had a full season of samples to work with, the measurement would stabilize, but the lack of any distinct upward trend in the correlation makes that seem unlikely. Instead we can treat the level of the sword as It's SNL, which in its current form does not want to be viewed live at all. Instead, you can catch any clips from the Internet later, and while scrolling, check out any swords. Installing Ninja posted.
Out of curiosity, I looked at the swing rate from the batter's perspective, since the bat itself (and therefore, the action of swinging) is actually controlled by the batter, which suggests that skill may be more dependent on the player making the decision to swing. . The results were very promising, but even the 250 swing sample fell short of the 0.8 correlation cutoff, coming out with a correlation of 0.46.
Few players, even the best, are consistent performers with respect to any given metric. Diversity, disorganization, and externalities lead to noisy, unbalanced actions. Similarly, even the best shows have hits and gauges. A series may chop it up for holiday episodes, but insist on doing musical episodes or dream sequences, or decide to play later. When drawing conclusions about performance, it's important to make sure that the sample size is large enough to distinguish between a false positive and a new situation – like when Chris Davis forgot to hit and Michael Scott walked. The office.
Source link
