Check pitchers on the midway marker Zips ZSTAT

When we talk about predicting the future, the categories of “expected” statistics have utility. These data certainly inspired a sense of mixing among fans, but they performed an important task to connect what Statcast and similar non-traditional metrics say to performance in the field. Without what happens in baseball games, the tough rate of launch angles of X% or Y degrees does not actually mean anything.

I’ve been doing predictions almost half of my life now, so I have a vested interest in using such information effectively in predictions besides my normal curiosity. Just like Statcast estimates (like “X” before “X” in XBA, XSLG, etc.), Zips has its own version that uses “Z” very creatively.

It is important to remember that these are not predictions. Zips certainly not only looked at last year’s pitcher’s ZSO, but also said, “Cool, Bra, we’ll keep going.” But the data confusing the way events go through, and it’s more stable to a single player than the actual statistics. This allows the model to occlude projections in one direction or the other. Sometimes this is extremely important because for pitchers, home runs allow. Home runs are easy to turbulent in the wild neutral stats, and the pitcher’s home run estimator predicts future home runs better than the actual home run allowed. Similarly, in a specific statistic, the more pitchers “she is poor” or “beyond the grade”, the more Zips believe in actual performance rather than expected performance. For more information on accuracy and construction, see here.

As we did in the batsman yesterday, let’s take a quick look at how well the pitching performance and poor performance last season performed on the mounds for the rest of the season. Again, note that these are not predictions themselves, but performance indicators assist Make a prediction:

According to ZFIP, of the 19 largest FIP ​​super scores, I obviously can’t count to 20 when making the charts – 18 of the remaining 2024 timetables have at least 30 innings. The biggest, Trevor Williams, ended his season with flexor pressure a few weeks later. After June 13, all 18 have higher FIPs. Until June 13, the RMSE (root square error) between FIPs and the rest of the season had a FIP of 1.46, while the ZFIP had a FIP of 0.93 against the playoffs. In other words, ZFIP’s performance in projecting FIPs for the rest of the season is better than the actual FIP’s super results. Remember that there is no projection data or regression for the “Help” ZFIP built-in mean, which is derived only from statcast and similar data types. Let’s take a look at the tough results of last year’s FIP:

For 18 underdogs with at least 30 innings remaining this season, ZFIP won with a smaller profit with an RMSE of 1.16 vs. 1.30 FIP.

ZFIP is better than ZFIP with better grades, and seems to be a unique feature of 2024, rather than a consistent feature of the model. In the case of half-quarter data, ZFIP is usually 30-40% accurate than FIP when predicting future FIP.

Let’s start with the data before June 29 using ZFIP under ZFIP numbers and super score numbers. I use 40 rounds as the cutoff point: here:

Zfip didn’t completely save Bowden Francis’ bad performance, but at least when his shoulders were better, it made him a moderately useful situation. It was fun to see Walker Buehler appear here because I got a lot of comments in the chat last month and he looked much better than his actual results. It seems that some of you are doing something. Zach Eflin is better than his numbers, too late for the Orioles, but at least that could make him more gained by the trade deadline. It’s fun to see Hunter Greene here because he actually has a reasonable season already. This suggests he may be more persistent in the Cy Young game.

The estimated numbers have given some of the league’s best pitchers a bite, but many of them (Nathan Evaldi, Garrett Crochet, Hunter Brown, Mackenzie Gore) are still seen as outstanding contributors, not to the same extent. Joe Ryan and Michael King were unscathed. King has a tougher blow this season and has received more 1-0 charges. Ryan’s ZFIP is not too worried because his Sty score history outperforms ZSTAT, so much that Zips emphasizes less on expected statistics when making predictions.

Turn our attention to home runs:

Of these three FIP components, home runs are easily the most valuable pitcher’s ZSTAT. Unlike batsmen, pitcher home runs are often an absolutely terrible statistic from a forecast standpoint, and many long-term failed pitchers come from numbers that are high or very low for home runs. Indeed, home runs make the pitcher’s terrible transparency is what makes XFIP more predictive, although it assumes that the pitcher has no effect on whether the pitch becomes a home run, which is a ridiculous concept. It is much better to measure the suppression of a home run by things like exit speed data, so in reality any estimates using this data will do a superior job in predicting the allowed future home runs, rather than home runs Tally or XFIP.

Jameson Taillon is a great example here. His barrel rate is not good, his hard rate is average, but both numbers are so exaggerated that his home run permit rate increased by about 70% and no sudden lack of speed. He allows more pulling flying balls yes It’s a bad thing, but it only accounts for about four additional home runs.

walk:

Unlike allowed home runs, allowed walking (and strikeouts) are good stats for pitchers, so ZSTAT doesn’t dominate the actual numbers here. ZBB is still more predictive than the actual trail, mainly because it includes two board discipline statistics that are important indicators of future walking rates: out-of-regional swing rate and the percentage of first strikes.

Ben Brown is fun here because he has made a huge improvement in his professional walking rates and ZBB suggests he can get better. His improvement on the first course of the Bat was spectacular. He has increased from a 46% strike for minors in 2024 to this year’s Grand Slam. Alas, he is currently being bothered by 0.362 Babip, so the Bear tries to “reset” him in the minor. ZBB isn’t as shocked as you can this year as you do, especially early (actually his recent weeks have actually improved). After all, he is likely to end up being the most valuable trade candidate in July.

Now let’s take a look at the strikeout:

ZSO is only more predictive than actual strikeouts, but when predictions access two numbers, the prediction works best. ZSO’s strongest ability is to determine the player whose contact rate is a bit out of place with its strikeout rate.

One thing you may notice is that among side jobs, there are often more veterans. There is actually something! This is not my initial intention, but the relationship between board discipline and strikeout seems to be capturing some kind of ability, whether you call it “veteran Moxie” or “pitch,” or any other ability, the data cannot be measured well by the data. If you take service time as one of the inputs, I actually made a significant improvement in the ZSO model, but I excluded it here simply because I’m trying to take advantage of performance only and not these “extra” features. When Zips interprets these data in the projection, it believes that ZIPS scores higher for young pitchers, while for older pitchers, it scores less. It’s a work in progress; I’ve been exploring the interaction of tracks, sequencing data and strikeouts, and these data seem promising. For the moment, don’t be excited or panic about these data even if it’s still useful!