NewPQS: PQS is Dead, Long Live PQS

Regular readers will know that my main long term research project has been the recording and analysis of the Giants pitching staff with regards to their PQS performance.  Well, BaseballHQ, the creator of the PQS methodology, announced in their 2017 book that they had revised the system in early 2016, changing their secret sauce, but I missed it, not reading the book until now.

Their new system, which they called NewPQS in their book, but will refer to as PQS going forward:

• Innings Pitched:  > 6 IP
• Hits Allowed:  H < IP
• Strikeouts:  K >= 5
• Command:  K/BB >= 3 (or if BB=0, K>=3)
• Home runs:  HR=0

Subtle differences, most of which were done to bring a more standard distribution to the PQS scores than had morphed with the rise in strikeouts over the years since PQS was first devised.  Plus, even with the HR explosion, 86% of starts could fall under the HR=1 rule, hence that change.  Also, the automatic PQS-0 for any start under 5 IP has been removed.

What this does is make PQS DOM starts much harder to achieve.  It also gives a name to the mid-tier PQS (2-3) which I've been calling MID but they are now using the term DEC for "decent".  

ogc thoughts

Wow, this will have huge implications for PQS.  According to the stats in the book, half of all starts were DOM starts from 2012-2014 under the old system, which means that the Giants staff was actually closer to average than I had originally thought and been presenting.  The new system will move most starts to the middle Decent starts region (2-3), resulting in less extremes, particularly at the high end of the scale, but ironically more extremes at the low end, pushing the percentage of DIS starts from 21% to 33%, while DOM starts fell from 50% to 27%, with DEC (I like this term since it continues the "D"; plus it fits in better with what it is in NewPQS, it was MID in OldPQS) starts rising from 30% to 41%.

• OldPQS:   DOM 50%/DEC 30%/DIS 21%
• NewPQS: DOM 27%/DEC 40%/DIS 32%  

And as a result, the ERA coming from each start is different:

• OldPQS:   DOM 2.32 ERA/DEC 4.54 ERA/DIS 11.20 ERA
• NewPQS: DOM 1.56 ERA/DEC 3.70 ERA/DIS 7.92 ERA

Changes:  Good and Bad

I'm not sure yet how much I like the changes.  The logic of how they came about with the changes makes a lot of sense.  As we know, baseball stats is not a stationary target that can be taken at face value.  I also like how DEC starts has changed from 4.54 ERA under the old system to 3.70 ERA under the new system.  I was surprised to see that more than half of all starts had zero home runs given up.  I like the subtle change for strikeouts, I wasn't happy with the -2 method before, and these changes makes it easier to create the logic in spreadsheets to calc PQS.  

And the changes align with some of the issues I've tried to note before.  That some auto-PQS-0 starts were actually on pace for an OK start but was shortened because they were being BABIPed, I've noted many times.  I've been wondering how the strikeout related components need to change given the continued surge in strikeouts, as hitters, further and further, goes for the homer more often and striking out more, as a consequence (under old, 67% of starts qualified, under new, only 48%).  

But that's the thing, right when they change the methodology in early 2016, the MLB has apparently juiced the ball in either 2015 or 2017 (538 had some articles, one and two, plus Fangraphs had a number of articles on this talking point, one, two, three, four, and argues it began in the middle of 2015; but pitchers were talking about the new stitching causing blisters in 2017), and that will greatly affect how the HR component contributes to PQS, it will be interesting to read what they have to say about this in their 2018 Baseball Forecaster book.   That's the problem, baseball is a moving target, the so-called Steroids Era (one can argue it was more a juiced ball than a steroids derived era) ran from roughly 1993-4 to roughly 2009-10, and PQS was built in that baseball environment, and NewPQS was built in the lowered Run Environment that had dominated until the ball was juiced again in recent seasons.

And that brings up a good point:  that this stat is geared towards Fantasy Baseball play, and not player analysis from a baseball perspective, per se.   I've liked this metric as a tool for pitcher analysis.  It is a vast improvement over the very old "Quality Start" metric of 6+ IP and 3- runs given up.  I also like it over Bill James Game Score, which needs a computer to compute daily (though Baseball-Reference.com does report on it the next day).  He intuitively built his, and it has worked out well in identifying how well a pitcher has pitched, but PQS has used the latest in baseball analytics regarding strikeouts, how important it is to strikeout more than you walk.  

So, overall, while I wonder how the HR change will affect things (I might investigate this), I think the changes are for the betterment of the metric, but will require a rejiggering in  our minds as to what is good and what is OK.  As such, I will be gathering the data on Giants pitchers and redoing as much of the data as I can, up to 2010 at least, maybe 2009 as well, we'll see how onerous a task this will be.  At some point next year, I'll tackle the playoffs again and see how things look with these changes.   I will start linking to this post instead of the glossary on BaseballHQ.com, as I can't access that link anymore without a big warning screen coming up.  

For an example of the massive change in recalibrating what is ace vs. what is good, let's take a look at Madison Bumgarner's history changes due to this:

• 2011:  old:  70% DOM/15% DIS;  
• 2012:  old:  66% DOM/13% DIS;  new:  34% DOM/44% DEC/22% DIS
• 2013:  old:  81% DOM/6% DIS;  new:  48% DOM/46% DEC/6% DIS
• 2014:  old:  67% DOM/12% DIS;  new:  48% DOM/37% DEC/15% DIS
• 2015:  old:  72% DOM/3% DIS;  new:  59% DOM/35% DEC/6% DIS
• 2016:  old:  76% DOM/0% DIS;  new:  56% DOM/35% DEC/9% DIS

Based on this, it seems like 30% is the new threshold for a good pitcher, 40% for great pitchers, and 60% for elite.  Bumgarner was severely hit by this change.   But as the book noted, "it brings the extremes into sharper focus.  It should be news-worthy  when a pitcher throws a PQS-5 or a PQS-0."

For another look at the change in recalibrating what is ace vs. what is good, let's take a look at Johnny Cueto's history changes due to this:

• 2011:  old:  54% DOM/13% DIS;  
• 2012:  old:  58% DOM/12% DIS;  new:  48% DOM/44% DEC/22% DIS
• 2013:  old:  64% DOM/27% DIS;  new:  18% DOM/46% DEC/6% DIS
• 2014:  old:  65% DOM/0% DIS;  new:  56% DOM/37% DEC/15% DIS
• 2015:  old:  59% DOM/6% DIS;  new:  34% DOM/35% DEC/6% DIS
• 2016:  old:  66% DOM/6% DIS;  new:  53% DOM/35% DEC/9% DIS

2013 was a short season for him, but that's a stark example of how changed that PQS season is for him.  2015 as well, but his other seasons (2012, 2014, 2016) were pretty close to the oldPQS.  But the 30/40/60 seems to hold well for Cueto, looking at his ERA and ERA+ for each season.

I will have to look over the top pitchers and see what their DOM looks like in NewPQS, and get a better idea of what the changes mean for interpreting.  I'm not happy because I thought that the OldPQS captured Bumgarner's dominance properly, but now his 2013 and 2014 seasons do not look as dominant.  But perhaps I'm picking at hairs, we'll see.  Plus, I'll have to learn the new methodology, though it seems easy to pick up once I get used to the new rules.