You Are Here > Baseball-Reference.com > Blog >

SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all B-R content, so you can quickly and easily find the content you want.

Also, our existing B-R blog rss feed will be redirected to the new site's feed.

Baseball-Reference.com ยป Sports Reference

For more from Andy and the gang, check out their new site High Heat Stats.

Is Quality Starts a useful stat? (not really)

Posted by Andy on August 30, 2011

A Quality Start is an official stat assigned to a starting pitcher who goes at least 6 innings while giving up no more than 3 earned runs. Let's take a look at this stat in a bit more detail to see if it's all that useful.

Here is the fraction of all starts going back to 1919 that were Quality Starts, using the definition as stated above:

 

 

Like a lot of pitching stats, this one peaked in the 1960s and 1970s when scoring was quite low. It dropped dramatically in 1987 when home runs were hit at a record pace. In 1993, there was a sudden drop from 53.6% to 50.4%, and the rate continued to fall each of the next 3 years. Over the course of the steroid era, the QS% bounced between 46% and 51%. Then, in 2010 and 2011, as run scoring has dropped again, QS% has increased dramatically to pre-steroid era levels. This year's rate of 54.2% is the highest since 1988's 56.4%.

But are quality starts helping teams win? Here's the percentage of team victories when their starter throws a quality start:

 

The win percentage was around 70% for the first half of the 20th century. It started to drop in the 1960s as more and more quality starts were being thrown (see the first graph). Then in the steroids era, the win % started to climb again. That makes sense--when run scoring was up, if your pitcher was good enough to give up no more than 3 runs over 6 innings, you had a pretty decent shot at winning. The win rate peaked in 1996, reaching nearly 70% for just the second time since 1958 (with the other being the HR-aberrant 1987). But as scoring has fallen off in recent years, the team win % in quality starts has fallen off too. This year's rate of 66.0% very slightly lower than 1981 and 1976's rates, and the last year lower than 2011 was way back in 1972.
If we look at 2011 teams with the most pitchers throwing at least 10 quality starts, it shows that the stat is a bit flawed. Among the leading teams (with 5 such pitchers) are the Reds, Twins, and Pirates, all teams that are near or below .500, with the Twins in particular having a surprisingly poor record despite all the quality starts. Meanwhile the Red Sox have just two pitchers with 10 quality starts and yet are one of the best teams in baseball. Another near-playoff team, the Angels, have just 3 pitchers with 10 quality starts.
The quality start stat is interesting, but only at a surface sort of level.

 

 

This entry was posted on Tuesday, August 30th, 2011 at 10:10 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.

126 Responses to “Is Quality Starts a useful stat? (not really)”

  1. @100, Steve; That's because LaRussa is a genius. If he were of lesser intelligence, it would have been three relievers to get the last six outs (six outs by three relievers would be a "QBPA"-a Quality Bullpen Appearance).

  2. @99 I don't know the answer to your question, but I picked 5 dates from this year and looked at the box scores in both leagues. In 55% of the games, the game was undecided after the end of inning 6, meaning (to me) that at that point or some point after that there was a tie or a 1 run difference or a lead change. 45% of the time, a team had a lead of two or more after 6 innings and it never dropped lower than 2.

  3. @100 Jackson, RHP, started the 7th, but when they brought in Kottaras a LHB as a PH for the pitcher, they substituted Rhodes, a LHP, which brought in Josh Wilson, a RHB as a PH. Then he switched to Jason Mott a RH to face a righty(out), lefty(hit), righty(out). Then he brought in Rzepezynski a LH to face Fielder a lefty. Then he brought in Salas a righty to face 3 righthanders, all 3 of which have a higher % this year vs RHP, but RHB only hit 0.169 vs Salas, LHB hit 0.198. Kotsay, LHB, pinch hit for Hairston, RHB, hitting into a double play to end the game.

    All of this to protect a 2-1 lead.

  4. @99, Mike L -- I'm not seeing how your question arises from comparing "8 and 4" with "6 and 3." Why focus on the probability of scoring two runs over the last three innings? Can you clarify? Thanks.

  5. Slightly off-topic, related to 100, but god almighty do I hate LaRussa. Ugh.

  6. @104 John-forget it. I was looking at your stats for the W% of those at exactly 6/3 and exactly 8/4. Just one follow up question. Are those stats through six innings, and through eight innings?

  7. @15 I looked at the game logs for the 4 pitchers with 5 or more losses. Their teams were 51-28 team in their quality starts. In 24 of those losses the opposing pitcher also had a QS.

  8. Listen, the whole "6 IP and 3 ER is a 4.50 ERA and that isn't quality" argument doesn't really fit here. With 6.1 IP and 2.7 ER per game... 3.98 ERA. With 7 IP and 2.5 ER per game... 3.21 ERA. The average QS is somewhere between those.

  9. @104 John-I need more sleep. Simple question needed to be better phrased. The win loss stats for those 6/3 and 8/4-were those stats from the point in which the pitcher was removed, or through the end of the inning? So, if someone pitched the 8th, did you count the batting stats in the bottom of the 8th if it was a home game.

  10. @88 I don't know about the validity of comparing win percentage for 8 innings, because within that you have a subset of complete game losses for the visitors.
    @109 This will include any stats in the 9th inning for that pitcher as it goes by outs recorded.

    From 1992 to 2011
    Losses for the visiting team when starter pitches a complete game with exactly 24 outs (8 innings) and ER allowed
    19 (0)
    122 (1)
    213 (2)
    186 (3)
    126 (4)
    52 (5)

    Team record and starting pitcher ER when starting pitcher goes 8
    1222-157 (0)
    1651-504 (1)
    1057-622 (2)
    496-492 (3)
    169-285 (4)
    49-129 (5)

    The home team wins 41% (97-138) of the games when the starter goes exactly 8 and gives up 4 ER.
    The visiting team wins 33% (72-147 with 126 complete game losses).
    Of course, some games may be on both lists.

  11. This is in reference only to the 5 Minnesota pitchers with 10 QS. Performance in QS is comparable to the ML average.

    In 2011 Minnesota has 50% quality starts vs the ML 54% for their top 5 pitchers. The team winning percentage for those 62 starts was 68% (42-20) vs the ML 66%. 15 of their 20 losses (75%) were to a team with a pitcher with a QS. I estimated the ML average would be 78% (29/37) of the QS losses would be to a team with a QS starter.

    Their team record is 11-51 (18%) when they do not have a QS, far below the league average of 32%.

  12. The win-loss record for the 5 Minnesota starters (not team record) is a combined 26-7 if they go at least 7 innings. 13-42 if they exit before that.

  13. JA @104
    If you start with a baseline that the starter will give up either three or four runs (forgetting the bullpen for a second), then, at minimum, you know you will lose every game in which your team scores two or fewer runs (three, in the 4/8 guy). So, 6/3 or 8/4 is an irrelevant pick em-nothing that happens after they leave changes the outcome. By the same token, if your offense is cooking and scores a large number of runs before they leave, unless your bull pen implodes, again, once again, the difference in their performances is largely irrelevant and nothing that happens after they leave changes the outcome. The swing in outcomes comes largely in the games that are decided by a small number of runs. I asked about 2 runs in 3 innings because a) that's the minimum number of runs his team must score after he leaves for a 6/3 pitcher to change an outcome in a game he's down by one run, and b) that's the probable minimum number of runs they need to score to change the outcome if he leave with a one run lead (but the bullpen gives up at least one run in the next three innings). The 8/4 guy is a little more set in-not only has he given up one more run, but his team has only one inning to make it up, so it's much harder to change the outcome.

  14. This is in reference to the entire Red Sox pitching staff. Of the 134 games played, there were 66 QS (49% vs the ML average of 54%) and comparable to Minnesota.

    Their record in QS is 50-16 (76% vs ML average of 66%) with 12 of the losses to a team with a QS. They were 32-36 (47%) in non quality starts, 15 points above the ML average).

    Beckett and Lester 50 Starts, 36 QS (team record 29-7), 14 NQS (4-10)
    Rest of team 84 Starts, 30 QS (21-9), 54 NQS (28-26 !!!)

    Andrew Miller has 10 Starts, 3 QS (3-0), 7 NQS (6-1)

    The team has a winning record, in starts, for all 7 pitchers starting 7 games or more.
    The Red Sox and the Yankees are the only teams in the AL with Runs scored minus allowed is greater than 1. Minnesota is the only team in the AL where allowed exceeds scored by more than 1.

  15. Read a bit of this thread and now I'm going to say something similar to something I've probably said before, and if someone in the thread has already come up with a similar idea then I'm sorry about stepping on toes.

    QS is a fine stat but its limitations are specific to its inability to distinguish between a "gave your team a chance to win" performance and an excellent performance, as we all know.

    My solution for "fixing" the usefulness of the QS stat is not to alter it, but to complement it:

    ULQS - Ultra Low Quality Start
    LQS - Low Quality Start
    MQS - Medium Quality Start (most similar to QS)
    HQS - High Quality Start
    UHQS - Ultra High Quality Start

    I had something else written when I thought I figured out a nice little easy ratio and innings requirement for each of the five listed above. I will go from best to worst.

    UHQS = IP:R = 4.5 or greater (min. 8IP)
    HQS = IP:R = better than 3 but less than 4.5 (min. 7IP)
    MQS = IP:R = better than 2 but less than 3 (min. 6IP)
    LQS = IP:R = better than or equal to 1.4 but less than 2 (min. 5IP)
    ULQS = IP:R = less than 1.4 (min. 5IP, however, regardless of runs allowed, a start of less than five innings is always a ULQS)

    Now to throw out some hypothetical starts and see how it looks. As you can see, I don't use thirds of an inning, because in the end even though this is a lot it should be simple to figure out. Thus, for the purpose of simplicity, you always drop the thirds of an inning off, rounding down to the nearest complete inning. Thus each tier of QS is a little tougher to reach.

    Start of 9IP: 0-2R = UHQS, 3R = HQS, 4R = MQS, 5-6R = LQS, 7+R = ULQS
    8IP: 0-1R = UHQS, 2R = HQS, 3-4R = MQS, 5R = LQS, 6+R = ULQS
    7IP: 0-2R = HQS, 3R = MQS, 4-5R = LQS, 6+R = ULQS
    6IP: 0-3R = MQS, 4R = LQS, 5+ = ULQS
    5IP: 0-3R = LQS, 4+R = ULQS
    4IP and below = ULQS.

    I used 1.4 as the line between LQS and ULQS because 1.5 makes 7IP 5R an ULQS and I'd prefer it as a LQS. Other than specific IP/R scenario, 1.4 and 1.5 as the line is identical.

    I know that's rather lengthy and not easily digestible, and perhaps not quite spot on for everyone, but to me the real value in expanding and stratifying the QS stat (it could be distinguished into three or seven layers, or whatever, not five like I've done) would be that it would be a nice table to be able to spit out to make a good comparison of how a pitcher's performances compared to other excellent (or poor) pitchers, and the league average.

  16. For the American League
    7 of the 8 teams with a winning record have 72 or more QS vs 2 of 6 with losing records
    8 of 8 winning teams have 48 or more QS wins vs 1 of 6 losing teams
    8 of 8 winning teams have 15 or more nonQS wins vs 3 of 6

    For the National League.
    5 of the 6 winning teams have 73 or more QS vs 4 of 10 losing teams
    6 of the 6 winning teams have 45 or more QS wins vs 2 of 10 losing teams.
    2 of 6 winning teams have 20 or more nonQS wins vs 6 of 10 losing teams

    All the winning teams in both leagues have at least 66 QS, 45 QS wins, 15 nonQS wins, and 67 total wins. If any team had even one parameter lower they had a losing record. All the teams had winning records in their quality starts. 7 of 14 of the winning teams were above 70%. 5 of 16 of the losing teams were below 60%.
    Redefining a QS will likely toss out more losses than wins, so the differences between the teams will likely be more dramatic.

  17. RE: Wins and Losses

    I call BS for anyone that believes W & L are useless. Tell me a pitcher doesn't change his philosophy when he has a 5 run lead and when it's tied. While he does not want to give up a hit, walk or run anytime, he approaches each batter differently in those situations. With it tied, he might be less likely to throw a fastball to certain hitters. With a 5 run lead and no one on, he is more likely to challange the hitter. The pitcher that can last longer into games and gut out the Win is what I want on my team. Forget about ERA its about pitching to a situation. He is not in a vacuum.

  18. One more comment and I'm done.

    This year there were 2032 games, 4064 starting pitchers
    There were 2206 Quality Starts.
    QS starters were 1457-749 meaning 72% of the games were won when the winning team SP had a QS.
    If the starting pitcher went less than 6 innings, but gave up less than 3 ER, the team record was 266-266
    If the starting pitcher gave up 4 or more ER, his team was 309-1017.
    Overall in nonQS the record was 575-1283 meaning 63% of the games were lost when the losing team starter did not have a QS.

    You can tweak the definition of QS and it might make a difference at the team or pitcher level, but I think that the 72 vs 63 we have here is a good indication that a QS is a good stand-alone global indicatior of the potential for a win at the league level, but to go down to the team level, you need to use both QS and nonQS wins and losses as a factor in overall team success and to go down to an individual pitcher contribution to the team record, you also need to take into account QS losses to a team when the opposing pitcher has a QS, because the rest of the team needs to do their job.
    The ML average is 4.15 runs per game. If both starters give up 3 ER in 6-7 innings(ML ERA 4.04), and the the relief pitchers give up 1 ER in 2-3 innings(ML ERA 3.64), and the defense doesn't allow any unearned runs, we've got a 4-4 tie game. If the pitchers do better or the offense bombards the relievers or the fielding falls apart briefly, it's still a QS win and a QS loss.

  19. @117 Agreed Not just the pitcher but the manager will play differently when the game is out of hand in either direction.

  20. I'm back.
    I'd like to respond to #55, the lowest end of a QS, when a starter gets 3 ER in 6 IP.
    The record today is 75 team wins 103 team losses. The SP has a record of 49-65 with 64 no decisions, 26 wins and 38 losses went to the relievers, so I think we can get away with saying the SP got them in a position to win in 49+26+38 games 113 of the 178 (63%). So let's look at the 65 losses. Starting pitchers were 45-53 with 6IP, 3ER, 0 unearned. In 16 games he gave up 1 or more unearned runs so we can say four factors played a role in his being charged with the loss. He gave up at least 3 runs and possibly costly defensive errors occurred and possibly his team could not score at least 3 runs and possibly the relievers did not prevent runs from scoring. The Dodgers lead the league with 6 games with the SP going 6 innings with 3 earned runs and 0 unearned runs and charged with the loss. The final scores were 3-0, 7-0, 3-1, 3-0, 6-0, 3-2. Of course in all 6 games the opposing SP had a QS. These go into the he could not put them into a position to win category. It's likely a lot of those 65 losses came when the opposing pitcher had a QS. Toronto lead with 4 games where the pitcher got the win. The opposing starting pitcher did not have a QS. The final scores and scores when the pitcher left the game 7-6 (7-3), 6-4 (6-3), 10-3 (10-3), 7-3(7-3) : a team effort. These go into the he put them into a position to win category.

    What if we took these games out of the QS category and into the nonQS category? The 72% - 63% values @118 become 68% and 68%.

  21. Johnny Twisto Says:

    The pitcher that can last longer into games and gut out the Win is what I want on my team. Forget about ERA its about pitching to a situation.

    Provide evidence of pitchers who do this. Thanks.

    Must be nice to pitch with all those 5-run leads.

  22. @120 makes a very obvious but easily overlooked point: QS has a much higher correlation to victory when one pitcher posts a quality start and the other DOES NOT! If both pitch quality starts, of course the outcome will not be highly correlated. A pitcher cannot control (NL hitting excluded) the performance of their opponent. QS overall percentage would inherently not be overly high since in a large percentage of games, I'm going to guess 15%, both starters put up quality starts!

  23. @121...With all due respect, what's there to prove? Just because one can't wrap a pencil or a calculator around something doesn't mean it's not true. There are situations within the game itself that aren't easily explained away by "pure" stats. Wouldn't that be a fair assessment?

  24. Here's a sample of 14 games (CIN, DET, KC) where the pitcher pitched 6 innings, 3 ER, 0 unearned runs and had a no decision. They are included in the 64 no decision games which I argued should be included in the starter put his team in a position to win category. Someone may argue that if he left the game as a losing pitcher it shouldn't be counted, but the fact that he did not receive the loss means his team scored 3 runs during the game to offset his 3. A Y means the opposing pitcher had a quality start. He had the lead in 3 of the 5 games when he faced a nQS pitcher.

    Outcome, Score when pitcher left
    Wins
    7-5 (3-3) Y tie game
    6-3 (3-3) N tie game
    4-3 (0-3) Y offense came alive
    5-3 (0-3) N offense came alive
    6-5 (4-3) N Blown Win

    Losses
    4-5 (3-3) Y tie when he left
    3-5 (3-3) Y tie when he left
    3-6 (3-3) Y tie when he left
    3-5 (2-3) Y his team tied the game
    3-4 (3-3) Y tie when he left
    5-6 (2-3) Y his team tied the game or took the lead
    5-6 (5-3) N blown win
    3-6 (0-3) Y his team tied the game
    5-7 (4-3) N blown win

    In reference to post 122 Good points and what should also be obvious is that the probability of a QS team win vs a nonQS team is higher than graph 2 indicates in fact the graphs should show opposite trends. If QS were 100% graph 2 would be 50%.

    With respect to your 15% estimate, I don't know what the number is, but with a ML probability in 2011 that a pitcher will have a QS of 54%, the probability that both will have one is 29% (0.54*0.54), 50% for only one and 21% neither. 100 games, 200 Starts, 108 QS (29 times 2 + 50 times 1 +21 times 0). Since those double QS games (more likely as QS% increases) are counted as a win and a loss in chart 2 at the initial post, you should see a trend where if graph 1 drifts up or down for a few years, graph two should drift in the opposite direction, you see that in the last 5 years, 1950-1970, a four year period in the mid 40s, the 3 initial years, 1991-1996 and a few other spots. Graph 2 is noisy, so I'm only looking for a good trend over a few years.

  25. Johnny Twisto Says:

    what's there to prove?

    That a pitcher's W-L record can consistently outperform his run support vs runs allowed, because he's giving up runs when it doesn't matter.

    So: which pitchers do this?

  26. @125...about every pitcher known. I believe you are reading way more into @117's statement. He didn't state "consistantly" only that it does occur. Of course, he would have to be the one to argue this point, I'm just "guessing" as to what his statement meant.