September 21, 2014

More On Reinventing the Quality Start and How It Looks Historically

September 13, 2008 by · 8 Comments 

On Friday, on MVN Outsider (on a different site), I wrote an article called “Revisiting and reinventing the ‘Quality Start'”.  In the article, I took a different approach to the definition of a Quality Start and established new criteria.  Since then, after some other discussions, I have revised the criteria, making it a bit easier to follow and also took the data and applied it to historical data using Baseball-Reference’s Play Index to crunch the data from 1956-2008. 

The article at MVN shows the deficiencies of a Quality Start in its ability to capture a performance worthy of delivering one in the win column when that pitcher takes the hill.  The old criteria defined Quality Start as any start where a pitcher threw at least six innings and gave up three earned runs or less.  The problem with the number is that pitchers fail to post a winning record in games where they give up three runs in six or seven innings even if they are credited with 52% of the no decisions in the win column.  In addition, the article establishes new criteria which have since been tweaked for this latest experimental analysis.

The “New” Quality Start (sorry for the lame name) is credited to starting pitchers whose performance meets the following criteria:

  • At least five innings and two earned runs or less allowed
  • At least eight innings and exactly three earned runs allowed
  • At least nine innings and exactly four earned runs allowed

Sure, it does not provide the simplicity of the established criteria of the Quality Start as we know it today but it does eliminate the often criticized capturing of a six-inning, three-run outing.  After looking at this year’s pitchers and how the Quality Starts numbers would change in the MVN article, I thought the next logical step is to look at how the change affects historical numbers.

For this exercise, I took every pitcher that achieved 200 or more Quality Starts since 1956 and converted their performance to the “New” Quality Start.  Of those 100 pitchers, five pitchers — Juan Marichal, Mike Torrez, John Candelaria, Warren Spahn and Vida Blue — saw an increase and three pitchers — Luis Tiant, Mike Cuellar and Whitey Ford — saw a drop.  The other 92 pitchers all saw decreases in their Quality Start numbers including Tom Glavine who dropped from 436 to 386, Curt Schilling who fell from 288 to 239 and Andy Benes who fell from 235 to 190.

The next step was to rank and group the players into four groups of 15 and see how the groups changed from the established Quality Start statistic to the “New” Quality Start statistic.  How does the top 1-15 in established Quality Starts compare to the top 1-15 in “New” Quality Starts:

GROUP I – Top 15

Old vs. New Quality Starts - Top 15

ANALYSIS:  There are some changes here ranking-wise but the biggest one is who joins the group and who leaves the Top 15.  In the established Quality Start criteria, Frank Tanana ranks 13th but with the adjustments made, the “New” Quality Start ranking sees Tanana fall out and Jim Palmer slide in to the Top 15.  Under the “New” criteria, the Top 15 becomes a bit stronger with Palmer replacing Tanana.

GROUP II – Second 15

Old vs. New Quality Starts - 16 through 30 

ANALYSIS:  Three players swap groups here with Jamie Moyer, Jack Morris and John Smoltz dropping to Group III and Juan Marichal, Mickey Lolich and Luis Tiant jumping up.  It’s tough to compare the two sets of three but of the six Marichal is the best for my money.  Also, while Smoltz might be a better than Lolich and Tiant, that includes his time coming out of the ‘pen.

GROUP III – THIRD 15

Old vs. New Quality Starts - 31 through 45 

ANALYSIS:  Two more sets of three exchange groups with David Cone, Mike Torrez and Charlie Hough joining the Top 45 and Doyle Alexander, Chuck Finley and David Wells sliding down a group.  Torrez actually jumped from beyond the Top 60 into the Top 45 on the “New” Quality Start side.  As mentioned before, Torrez was one of the few in the Top 100 to see a jump in their number of Quality Starts.  If I had to pick between the two groups, I am more satisfied with Cone, Torrez and Hough over Alexander, Finley and Wells.

GROUP IV – Fourth 15

 Old vs. New Quality Starts - 46 through 60

ANALYSIS: Tough to argue that the group didn’t get weaker this time with the addition of Bob Forsch, Tom Candiotti, Ken Holtzman and Mike Cuellar and subtraction of Mark Langston, Kenny Rogers, Andy Pettite, Kevin Appier and Frank Viola.  Statistically, the numbers produced by the five that dropped are better than the four who moved up (the fifth was Torrez who jumped to Group III).

The challenge here in breaking down the numbers this way does not factor in Quality Start % which is a little more important than the raw number of Quality Starts a pitcher delivered.

The most telling argument for the new version of the Quality Start statistic is a comparison between Whitey Ford and Tim Wakefield.  Unfortunately, the Baseball-Reference data only analyzes numbers from 1956 on so Ford’s data is somewhat compromised.  However, for this comparison it is perfect.  Ford started 395 games from 1956 to 1967 and Wakefield has made 396 starts in his career.  When analyzing the two using the currently accepted criteria of Quality Starts, Ford had 211 compared to Wakefield who earned 210 Quality Starts.  I don’t know about you but that doesn’t look right.  Now, under the “New” criteria, Ford stays the same at 211 but Wakefield drops to 170.  Why the big difference?  One of the reasons is that Wakefield made 19 starts of exactly six innings and exactly three runs and in those games Wakefield went 5-6 and the team he started for went 6-13.  Is that quality?  I think not.

The “New” Quality Start statistic needs further analysis, of course.  How does it affect Quality Start % and is it a true representation of what a Quality Start looks like?  As I mentioned in my article on MVN, I was always a big fan of the statistic until a harder look revealed that the statistic captured starts that just did not help a team win more than it led a team to a loss.  Agree or disagree, any feedback is helpful to further tweak and play with this to make it more representative of what John Lowe set out to capture 23 years ago when he introduced the concept in the Philadelphia Inquirer.  What do you think?  Are we ready for change or are we happy with the same old flawed approach?

Comments

8 Responses to “More On Reinventing the Quality Start and How It Looks Historically”
  1. Mike Lynch says:

    I’ve been waiting for a change in QS since it was first introduced. I don’t think you capture “Quality Starts” with such a limited criteria. For example, a pitcher who allows three runs in six innings has a 4.50 ERA, which it typically worse than league average, but a pitcher who allows two runs in five innings has an ERA almost a full run lower, yet doesn’t qualify because he didn’t throw the minimum number of innings required. And a pitcher who allows four runs in eight innings doesn’t get a QS even though his ERA is identical to that of a pitcher who allows three runs in six. It makes no sense.

    A sliding scale makes much more sense, as does lowering the ERA required to earn a QS.

    Have you considered creating era adjusted criteria to account for the differences between, say, the 1960s and 1980s and beyond? That would be interesting to see as well.

  2. BJStone says:

    I think it’s real simple: If a starter lasts 5 innings or more with an ERA of under 3.00, it’s “quality”. An ERA of 3.00 to 5.00 is “average”, an ERA of 5.00 to 7.00 is “poor” and any ERA over that is simply pathetic.

    It doesn’t need to be so freakin’ hard.

  3. Bill Baer says:

    “any ERA over that is simply pathetic.”

    Also known as an Adam Eaton start.

    I couldn’t resist, sorry.

  4. James Farris says:

    Wow! Claude Osteen. Ive never seen career QS numbers before. Id like to see the numbers with run support alongside.

  5. Mike says:

    Why 8 innings and not 7?

  6. Brian Joseph says:

    At 7 innings and 3 earned runs allowed exactly, the winning percentage of pitchers is 43.2% since 1956. Pitchers turning in a 7 IP-3 ER performance are 1840-2418 with 1679 no decisions. Typically, relievers win 52% of their decisions which slightly increases the overall winning percentage of such a performance but it is still well under 50%. Hard for me to qualify such a performance as a quality start. The number doesn’t hit above 51% until 7-2/3 innings where it jumps to 59%.

  7. Derrick Reisdorf says:

    Okay. Here’s the point. If a pitcher goes 6 innings, that is good. Period. You can hopefully get away with a 2-inning setup and bring in your closer. The point is, a team did not have to go to their bullpen early in the game. For example: A guy’s walking batters all over the place and fielders are making great plays in the field, and come inning 5 the pitcher walks in 2 runs. He can’t even get out of the 5th. You think this is better than 3 runs after 6?
    I think we’d have to add more factors to all of this “research”.
    For instance, one factor many of us never think about that contributes to why a starting pitcher’s ERA is higher than relief pitchers is that more often than not (I wish I knew the percentage), a starter is yanked as soon as he gives up some runs- they don’t leave him in for the remainder of an inning. Often times (especially now days) managers will have specialists that come in to face same-handed batters, or have a setup man in the for one or 2 innings, or have the closer in there for one. They don’t have to stay in there until they give up a couple runs. Say a closer gives up a run once ever 3 appearances. That’s a 3.33 ERA. Not a bad ERA, but not good by any means for a closer. Now, take a starter- say he’s cruising along. He’s pitching a gem of a game and has only allowed 2 runs though 6 innings on 2 hits. Into the 7th, he allows a 2-run homer with no outs. The manager yanks him. Perhaps he can pitch another 2 scoreless innings. But that’s the thing- a starter isn’t in the game for a set amount of time (unless a team has a pitcher on a strict pitch count). Besides, most likely when a pitch count gets high, the pitcher’s done a lot of work. This still doesn’t take away from the possibility that they’ll yank the pitcher if he gives up a few runs fairly late in the game. He may still have pitched a good game though he didn’t get a quality start. The pitcher got his team into the 7th (and probably just went through the tougher part of a lineup, giving up the 2-run HR). So, now the pitcher just went from a 3.00 ERA to a 6.00 ERA.
    Now, picture a different scenario. A starter’s team has given him ample run support. His team is up 8-4 after 6. He’s not pitching great, but the manager wants to stay away from using his pen. He gets one out and puts two on. He has a lower ERA than the other pitcher but he didn’t pitch a better game.
    Keep in mind, people, that QS was determined to be a better indicator of a pitcher’s effectiveness than wins and losses. How many games has this pitcher went deep into a game and kept the run total down?
    A win is still a valid stat. It can be possible that a pitcher can pitch up or down to competition. I’m sure there are pitchers out there that stop trying to strike batters out when they have an 8 run lead. Some pitchers may want to keep the ball in the zone and let them put the ball in play to save themselves the work (if the pitcher has the skills to do so). I mean, you can make up any stat up to satisfy your needs. QS% can show how consistent a pitcher is. Sure, it’s possible with nothing but quality starts has a 4.50 ERA on the season, but not very likely. It’s more than likely he also has different quality starts throughout the season, and his ERA is probably in the mid-3s. I personally like the QS. It shows that a pitcher #1 started a game, #2 saved the team from having to use their bullpen early, and #3 didn’t give up a lot of runs. Their team has a pretty good chance of winning the game regardless of what their offense did for the previous 5-6 innings. The the bullpen keeps up the “minimum quality start pace of a 4.50 ERA” the most miraculous thing that offense is expected to do is score 5 runs in 3-4 innings. That is the absolute worse case scenario in for a quality start- that the pitcher gave up 3 in 6 and the offense hasn’t scored any runs.

    I don’t understand why people are comparing winning percentage to quality starts when quality starts was meant to REPLACE winning percentage. You know what stat is even more accurate than quality starts? Wins and losses! Every time a pitcher got a win, his winning percentage was 100%! Every time a pitcher was credited with a loss, his winning percentage was 0%! How weird is that???

    I think the best sabermetric stat I’ve seen for a pitcher is K/PA (and then add some degree to that bases per hit). Bases per hits + walks isn’t a stat that really has been adopted by anyone but to me it measure how well people hit the ball against the pitcher. Of course, it’s usually safer for a pitcher to give up a walk than a hit, but can require many more pitches.

    Anyway, sorry about the rant!

  8. Derrick Reisdorf says:

    But I do agree- it’s hard to determine whether it will be harder or easier to score runs in the future. Who’s to say it will be easier or harder to score 3 runs in 6 innings? However, the Win is not based on a specific number and is concrete- did that team hold onto the lead and win the game after the pitcher came out?
    Perhaps, average number of innings per start divided by ERA would give us a better indication of how valueable a pitcher is? He saves the manager from working his bullpen and doesn’t give up runs.

    You know what stat I’d like to see? What is the average winning percentage of a team when any player on that team gets a no decision. I would assume it’s a losing percentage. Because there are three possibilities of getting a no decision. You’re team is winning when you leave the game and someone else on your team loses it. This scenario presents itself with a team winning percentage of .000 pct. You’re tied when you leave the game and someone else gets the win or loss (which would result, theoretically in a team winning percentage of .500 pct). Or you’re losing the game, but you’re team ties the game or takes the lead, in which once again would result in a .500 pct. No matter what the likelihood of any of these scenarios happen, any percentage of each scenario in a no-decision will result in a team’s winning percentage below .500. I may not have covered all the logical bases in my head, but this seems right to me.

    Anyway, the most important point is that if you average the ERAs of quality starts, that ERA is pretty low. We should find out how many times the 6 inning & 3 run start occurs. Also, after 6 innings, a pitcher has went through a lineup at least twice, and with 3 runs allowed has most likely faced the heart of a lineup at least 3 times. A lot of times, it could be easiest for a manager to bring in an opposite-handed reliever (or some other specialist) to pitch to someone on the same side of the plate or to limit the starters pitch count or to make some other strategic move (e.g. defensive switch).

    So, let’s keep those ideas comin’ kids!

Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!