Debating Blyleven's HOF Candidacy

[nanwynnfan]

Be my guest. Heck, I'd chip in the cost of a new shovel.

[bedir than average]

Triad, why should wins matter?

Its not as if Blyleven could control his offensive support

[WilsonC]

• Was there ever a three-year period in Blyleven’s career where he was one of the five best pitchers in baseball for that period? If so, when? If not, don’t you think that’s a reasonable standard to start with?

- I think he definitely has a case from about 71-78. I don't know if there's a three year period where he could conclusively be considered top 5, but he's certainly in the mix. I don't necessarily agree that there should be a hard rule in ranking, particularly over a short period. He needs to be one of the best of his time, to be sure, but it also needs to be considered who his peak overlaps with. Whether he's ranked fourth over his best years or sixth is less important than how good he actually was over those years.

• Do you feel 287 wins is enough to qualify a pitcher for the Hall of Fame, providing he has a winning record?

- Absolutely. It's not quite the magic number 300, but historically between 250 and 300, it's closer to the rule than the exception. His individual merits need to be evaluated too, obviously, but the win total is very much in line with deserving HoFers.

• What do you consider to be Blyleven’s two best seasons?

- 1973, and then probably 1974. One thing about Blyleven's peak is the remarkable season-to-season consistency. Kind of like Oswalt in that regard - he was consistently among the top, but with no real anomally seasons that standout as appreciably better or worse than his true level of ability. When looking at, say, his best five to nine years as a peak value, it's more favorable to him than looking at his absolute best seasons.

• Do you think Blyleven’s lackluster Cy Young voting (3rd, 3rd, 4th, and 7th) means that the voters somehow didn’t appreciate him, and if so, why didn’t they?

- Wins/Losses is the obvious answer. Really, that's the one question mark on his track record of his entire career. If he was 300-237, would we be having this discussion? Of course not. The only question is whether or not there was something about his performance that made him less likely to win than his peripherals showed, or whether he simply got unlucky in a handful of decisions throughout his career. It's about a dozen decisions between where he is now, and being a virtually unanymous no-brainer. It doesn't really seem too improbable that someone could lose a dozen more games than expected based on his performance over the course of his career. The question is, does this apply to Blyleven?

[Triad]

Originally Posted by bedir than average

Triad, why should wins matter?

Its not as if Blyleven could control his offensive support

There's much more that goes into wins than offensive support and runs allowed. Those are just what's on the surface.

A couple key reasons pitching wins should matter:
• They give an indication of how far into each game a pitcher was going (and how frequently). A pitcher who lasts 5 or 6 innings a game will have more no-decisions, and even many of his 1-run affairs will have time to turn over in the final 3-4 innings. A pitcher who goes 7+ innings will have more control over his W-L record.
• Pitching wins are context-sensitive. In other words, they treat all wins as no more than a value of 1 each, and all losses as no more than a value of one each. Thus, a terrible outing of 8 ER is washed away as just a regular loss, and a terrific outing of 0 ER is not overvalued as something more than a win. (I'll point out that ERA is still a good balancer with W-L, but the two components keep each other honest, and combined with total number of IP, they show a great representation of what the pitcher did.) ERA doesn't measure consistency, but rather gives a flat average. If a pitcher has several shutouts and 1-run games, but then also has many games where he blows it, ERA evens those out and still makes him look good. But a shutout is worth only 1 win, just like a 2-run win is worth only one win. A shutout increases a team's chances of winning by a few percentage points, but not to the extent that the ERA difference would suggest.

Statistical analysts often look at pitching wins and think they're supposed to represent team wins, which they're not. They simply have the same name, but they represent different things. A pitching win doesn't really mean that the pitcher "won" the game, because it was the team that won the game. For the pitcher, it's merely an indicator that he made a strong contribution to the win. Probably about 75% of the time, a win or loss by a pitcher is a good indicator of how well he pitched. Mixed in with ERA, it becomes a solid one-two punch.

Estimates from various models have pitching weighted at about 35% of the overall value of all the players on the field (with fielding being about 13% and offense being about 52%). 35% for one player is huge. That is a lot of influence on the outcome of the game. In one season, or 30-35 starts, there can be some fluctuation, where you get situations like Matt Cain's W-L record being completely out of whack. However, over the course of 100 starts (3 seasons or more), these things do tend to level out more. And over a career, they are very level. I can't think of any cases over an entire career where a true HOF-caliber starting pitcher had close to a .500 record. There is a correlation.

For the best pitchers, run support does not play as much a role as it does for the average pitcher. In the current game, if you're giving up 1-2 runs in many of your outings, most any team is going to cover that most of the time. It's the pitchers who regularly give up 3-4 runs a game where run support comes more into play. Matt Cain did get the shaft this year, but if he'd pitched like Jake Peavy, he would've been something like 9-13 instead of 7-16. The Giants scored 4 or more runs in his games only 11 times, and they were 7-4 in those games. When he gave up 3 or more runs, he was 2-11. When he gave up 2 runs, he was 0-1. When he gave up 1 run, he was 4-4. When he gave up 0 runs, he was 1-0. He had some control over it, but in this case not as much as most pitchers would.

[Triad]

Originally Posted by WilsonC

I think he definitely has a case from about 71-78. I don't know if there's a three year period where he could conclusively be considered top 5, but he's certainly in the mix. I don't necessarily agree that there should be a hard rule in ranking, particularly over a short period. He needs to be one of the best of his time, to be sure, but it also needs to be considered who his peak overlaps with. Whether he's ranked fourth over his best years or sixth is less important than how good he actually was over those years.

That sounds reasonable. I was just suggesting a 3-year period as a minimum. A 2-year period wouldn't tell us as much. Ron Guidry, Dwight Gooden, and the like would creep through.

Let's check your period of 71-78. And we can extend this to the top six pitchers of that time. It has to start with Seaver, Carlton and Palmer. Blyleven was nowhere in their class in those years, so that cancels out three spots right there. Blyleven's main competition for the last three spots would seem to come from Catfish Hunter, Gaylord Perry, Ferguson Jenkins, Vida Blue, Luis Tiant, Nolan Ryan and Don Sutton.

During that time, Blyleven had 7 seasons of 120 ERA+ or more, three of those at 140 or more. His record in his "prime" was 126-114 (.525).

Here's a good place to start. In that period, here are the Cy Young finishes of each.

Perry - 1st, 1st, 4th, 7th
Jenkins - 1st, 2nd, 3rd, 6th
Blue - 1st, 3rd, 6th, 6th, 7th
Sutton - 3rd, 4th, 5th, 5th, 5th
Hunter - 1st, 2nd, 3rd, 4th
Tiant - 4th, 5th, 6th
Ryan - 2nd, 3rd, 3rd, 8th
Blyleven - 7th (gulp!)

Do you really think the Cy Young voters got it that wrong? Is he the only pitcher in history that got shafted like this during his prime? If it's just a matter of overvaluing wins/losses, then we would find a real pattern throughout history of not giving certain pitchers enough credit. Where do we find this?

Are you trying to tell me that a pitcher that only got Cy Young votes in one of eight years in his prime, and finished 7th that year, would be one of the six best pitchers of that period? It seems inconceiveable. About the best one could stretch this would be to say that Blyleven was the 8th or 9th best pitcher during that time. Not all that impressive.

Originally Posted by WilsonC

• Do you feel 287 wins is enough to qualify a pitcher for the Hall of Fame, providing he has a winning record?
Absolutely. It's not quite the magic number 300, but historically between 250 and 300, it's closer to the rule than the exception. His individual merits need to be evaluated too, obviously, but the win total is very much in line with deserving HoFers.

This would mean you would also elect Tommy John and Jim Kaat, correct? Where's the cutoff? 270?

Originally Posted by WilsonC

• What do you consider to be Blyleven’s two best seasons?
1973, and then probably 1974. One thing about Blyleven's peak is the remarkable season-to-season consistency. Kind of like Oswalt in that regard - he was consistently among the top, but with no real anomally seasons that standout as appreciably better or worse than his true level of ability. When looking at, say, his best five to nine years as a peak value, it's more favorable to him than looking at his absolute best seasons.

This seems a very unusual pattern for a Hall of Fame pitcher. Can you think of any other worthy Hall of Famers who fit this pattern at the same level? I can't think of any. Phil Niekro might be close, but then I don't think he's a worthy Hall of Famer either. What is reasonably concluded from this is that Blyleven was a very good pitcher for a long time, but was never a great pitcher. Is the Hall of Fame a place for very good pitchers?

In 1973 and 1974, what place would you have voted Blyleven in the Cy Young? I can't see him deserving higher than 5th either time. That doesn't say much for your career when that's your best.

1973 - what you and others would call his best season...

Blyleven was 20-17, with 3 no-decisions. The Twins actually bailed him out in two of those three ND, where he gave up six runs in both of them.

Run support. The Twins scored 165 runs in his 40 games (average of 4.13 per game). He allowed 109 runs, and the bullpen allowed another 12, for a total of 121 runs allowed.

165 runs scored, 121 runs allowed. He was given an average of a 1-run cushion in each game. How does this translate to poor run support?

They scored at least 4 runs 18 times, and 3 runs 6 times.

Blyleven's ERA in 1973:
0.96 in wins
4.48 in losses

I need to revise what I said earlier about the gap between Blyleven's inconsistency. When he was good, he did very well, and when he was bad, he was just bad enough to lose. This made his ERA look good, but he still deserved to lose the games he did, for the most part. In contrast, Catfish Hunter (3.34) and Vida Blue's (3.28) ERAs that year looked worse, but that's mainly because when they lost, they tanked. In reality, one terrible loss should not reflect badly on three or four other wins.

Blue:
1.79 in wins
6.84 in losses

Hunter:
2.53 in wins
9.13 in losses

Even Jim Palmer (2.40) did worse in wins than Blyleven, but he had more of them, and he did worse in his losses, but a loss is only one loss.
1.31 in wins
5.11 in losses

What you're doing is overvaluing Blyleven's ERA due to his many games where he allowed 0 ER and having a "decent" ERA in his losses. A 4.48 ERA in his losses is still not good enough to win, though it keeps his team close. There's no bonus credit for giving up 6 runs instead of 9 runs. Either way, you lost. ERA incorrectly blends those bad losses in with the other good games, and makes Blue's and Hunter's ERAs look worse than they really were.

Originally Posted by WilsonC

• Do you think Blyleven’s lackluster Cy Young voting (3rd, 3rd, 4th, and 7th) means that the voters somehow didn’t appreciate him, and if so, why didn’t they?

Originally Posted by WilsonC

Wins/Losses is the obvious answer. Really, that's the one question mark on his track record of his entire career. If he was 300-237, would we be having this discussion? Of course not. The only question is whether or not there was something about his performance that made him less likely to win than his peripherals showed, or whether he simply got unlucky in a handful of decisions throughout his career.

Yet in 22 seasons, he NEVER had that stellar season. Not once. He was never spectacular for a full season. This should make us very suspicious.

The damning evidence against Blyleven is that during this time, his teams scored 2% above the league average, and the other pitchers on his teams' staffs (min. 20 GS in a season) had a cumulative winning percentage that was BETTER than Blyleven's! How could he be the only unlucky pitcher on his staffs over the course of his first ten seasons? It's unthinkable.

Originally Posted by WilsonC

It's about a dozen decisions between where he is now, and being a virtually unanymous no-brainer. It doesn't really seem too improbable that someone could lose a dozen more games than expected based on his performance over the course of his career.

I could understand that, but the law of averages says that a great pitcher would not have 10 consecutive seasons during his prime of being unlucky. That's statistically impossible within the realm of all the pitchers who have ever pitched in the last 130 years.

You seem to be saying that Blyleven's redeeming value is that he is close to being great.

[WilsonC]

One of the truths of statistics is that often times, an occurence that seems highly improbable for an individual becomes rather likely given a large number of samples. While it is reasonable to propose a theory that Blyleven's ERA and peripherals are misleading due to possible inconsistency (or for that matter, excessive consistency) that reduced his overall winning percentages, the fact that it seems improbable for him to be that unlucky doesn't really hold weight as a means of validating the theory.

I've started going through Blyleven's game logs to see whether there is, in fact, some inconsistencies in performance that are responsible for the discrepency in his W/L record versus his actual performance. I've started graphing his actual performances into some matrices, based on innings pitched in each appearance and runs allowed. For simplicity sake, all innings are rounded down and all runs (earned and unearned) are creditted to the pitcher. I've also taken a few shortcuts, like grouping all 5 or fewer innings starts together and grouping all starts allowing 6 or more runs together, mainly to save a little time for a preliminary study. My basic premise is the assumption that, say, a 6 inning 2 run start by Blyleven is valued the same as a 6 inning 2 run start by any other pitcher under the same context.

In looking at Blyleven from 1970-1978, even just eyeballing the results there are a few places where he seems strangely unlucky. For example, when he went 8 innings and allowed 1 run, he was 3-6 over that span, he was 3-8 in 7 inning 2 run games, and 1-8 in 7 inning 3 run games. Additionally, over nine years, he only won 3 games where he allowed more than 4 runs, which seemed awfully low to me.

Next, I started doing the same thing for his teammates (the over 20 GS samples) to get a comparisson of how the rest of his team did with identical performances. So far, I've only gotten as far as 1970-1972 for Perry and Katt done, but did did a test anyway out of curiosity.

I determined the winning percentage of his teammates with each Runs/IP permutation in the matrix, and multiplied that by the number of decisions Blyleven had with each Runs/IP permutation to get his 'expected' wins, if he had the same 'luck' as his teammates. While my data's incomplete, Blyleven gains 24 wins over these 9 years in this preliminary test of the methodology, boosting his winning % almost a hundred points to .623.

As a secondary test, I tried tried estimating the expected winning percentage of each permutation by calculating a pythagorean winning %, using the league average R/G under for the runs scored variable, and the ERA of each Runs/IP permutation as the runs allowed variable. It's quick and dirty and doesn't account for bullpen support, but I was curious what this would show. As it turned out, this method proided almost identical overall results to the teammate study. He gained 23 wins, and had an expected winning % of .618.

What do these studies show so far, aside from the fact that I had too much time on my hands? They don't show anything conclusive, but they suggest that Blyleven was likely unlucky, not only when looking at his overall averaged stats, but also when looking at his game-by-game performance record when compared to his decisions.

Whether it's based purely on luck, or whether there's a valid reason for these discrepecies - for example, if his managers took particular care to match him with other aces - it's difficult to tell from this. However, if his misfortunes with regards to W/L throughout his first nine years are even half as extreme as they appear in this preliminary study, that makes up the entirety of the dozen decisions I suggested separating him from the no-brainers. I'm not suggesting he was almost great, I'm suggesting that he was a great pitcher with just enough random quirks in his results to require a closer examination to judge his true level.

[Triad]

Good work, Wilson. What are the leading variables? I'm thinking some of them could be:

Blyleven's fielding
Blyleven's batting (not an issue on his AL teams from '73 on, however)
The context of each game situation which could have kept scoring down (weather, day/night)
Fly ball/Ground ball pitcher
Pace of pitcher effecting fielders and/or batters
Distribution of runs allowed by innings
Different catcher than some of the other pitchers?
The other players on the team didn't like Blyleven?

As for luck, luck evens out over time, not the other way around. This is a mathematical constant. Over a few thousand innings, a random occurrence would have negligible effect. I would find it hard to accept that Blyleven had significantly worse luck than any pitcher in history with at least 300 starts. The pattern of luck (which is a contradiction in terms) strongly suggests an underlying cause, or group of causes.

Originally Posted by WilsonC

In looking at Blyleven from 1970-1978, even just eyeballing the results there are a few places where he seems strangely unlucky. For example, when he went 8 innings and allowed 1 run, he was 3-6 over that span, he was 3-8 in 7 inning 2 run games, and 1-8 in 7 inning 3 run games. Additionally, over nine years, he only won 3 games where he allowed more than 4 runs, which seemed awfully low to me.

Those do indeed seem low. We know that the Twins and his other early teams were not typically poor offensive teams, so their ability to score runs was there. What is this x factor?

Originally Posted by WilsonC

I determined the winning percentage of his teammates with each Runs/IP permutation in the matrix, and multiplied that by the number of decisions Blyleven had with each Runs/IP permutation to get his 'expected' wins, if he had the same 'luck' as his teammates. While my data's incomplete, Blyleven gains 24 wins over these 9 years in this preliminary test of the methodology, boosting his winning % almost a hundred points to .623.

This will only ultimately tell us something if we can account for what that luck factor is.

Originally Posted by WilsonC

What do these studies show so far, aside from the fact that I had too much time on my hands?

It's a great way to spend the off-season!

Originally Posted by WilsonC

They don't show anything conclusive, but they suggest that Blyleven was likely unlucky, not only when looking at his overall averaged stats, but also when looking at his game-by-game performance record when compared to his decisions.

The game-by-game performance is going to lend us more clues than the overall totals, I believe.

Originally Posted by WilsonC

Whether it's based purely on luck, or whether there's a valid reason for these discrepecies - for example, if his managers took particular care to match him with other aces - it's difficult to tell from this.

It would seem that the chances of it being attributable largely to luck are almost nil. We can examine the requirements for such conditions later.

Originally Posted by WilsonC

However, if his misfortunes with regards to W/L throughout his first nine years are even half as extreme as they appear in this preliminary study, that makes up the entirety of the dozen decisions I suggested separating him from the no-brainers. I'm not suggesting he was almost great, I'm suggesting that he was a great pitcher with just enough random quirks in his results to require a closer examination to judge his true level.

Yes, that would make him a legitimate Hall of Famer. If he deserved to be more like 307-230 and it indicates he had a better peak, then I'd support him.

[WilsonC]

Originally Posted by Triad

Good work, Wilson. What are the leading variables? I'm thinking some of them could be:

Blyleven's fielding
Blyleven's batting (not an issue on his AL teams from '73 on, however)
The context of each game situation which could have kept scoring down (weather, day/night)
Fly ball/Ground ball pitcher
Pace of pitcher effecting fielders and/or batters
Distribution of runs allowed by innings
Different catcher than some of the other pitchers?
The other players on the team didn't like Blyleven?

I think some factors could definitely explain the discrepency in his ERA versus his RA, if indeed he had a pattern of a high discrepency. His fielding may have a slight impact, and his flyball/groundball rates, as well as any quirks of the type of contact he may have induced could have an impact, especially if he induced a lot of contact to a defensive weakness on the team. He was known for his devastating curveball, which may have been difficult for some catchers to handle, leading to more pass balls with him on the mound. Note though that these types of factors would contribute directly to runs allowed, so any such defensive quirks would already be accounted for with my methodology.

One potential variable I could see having an impact in his performance versus win/loss record would be managerial tendencies. It's possible his managers used a different lineup with him on the mound (different catchers would be a possibility - easy enough to check). His managers could have taken particular care matching him up with other aces, giving him a disproportionate number of pitcher's duels with other top-notch pitchers (a little trickier to check, as we'd need to compare his opponents to the norm, and to the opponents of other aces). Also, given the number of complete games he threw, it's possible that he was simply rode too hard at times, and his managers did a poor job of relieving him when he was fatigued (This would likely show up as a disproportionate number of late runs allowed in his losses).

Originally Posted by Triad

As for luck, luck evens out over time, not the other way around. This is a mathematical constant. Over a few thousand innings, a random occurrence would have negligible effect. I would find it hard to accept that Blyleven had significantly worse luck than any pitcher in history with at least 300 starts. The pattern of luck (which is a contradiction in terms) strongly suggests an underlying cause, or group of causes.

Careful now, as this is only partially true. Firstly, we're not looking at random occurences over a few thousand innings, as those would be represented in his ERA, WHIP, K, etc. - all of which stack up nicely with most of his contemporaries who made the HoF. Rather, what we'd be looking at is random occurences over 527 decisions. However, even there it's not really 527 decisions we're looking at - we can remove the 60 shutouts from that, as there's essentially no luck involved as to whether a shutout will be worth a win. 1-run complete games would be less subject to luck as well, and he had plenty of those too (though I don't have his career totals handy). Also, his worst outings would be less subject to luck, as they would be almost assured losses. So we'd really be looking at a sample of no more than about 400 or so that luck plays a part of, a solid sampling, but not as assured as thousands of innings to wash out the noise.

The other thing is that it's not quite true that luck evens out over time - rather it converges and is washed out. That is to say if you take a random sampling of a player's career where he happened to have a stretch of winning 5-10 less games than expected, it won't balance out with him making up those 5-10 games elsewhere - but rather given enough decisions, those 5-10 games would become irrelevant. When we're talking about a sample of 500 or so, a particularly unfortunate stretch of losing 13 more games than expected over a few seasons doesn't have time to wash out - it's the difference between 287-250 (.534) and 300-237 (.559). A big difference.

Consider this: If you were to take the best 4 year stretch performance-wise of any given pitcher's career, there's a 0.81% chance that he will fall under the 30th percentile mark for wins expected in every one of those 4 years, purely due to luck. That would be worth at least a few wins a year in those 4years, and would be extremely unlikely for any given pitcher to suffer such bad luck. However, is you take a random sampling of, say, the 40 best pitchers in history, the probability that one of those 40 guys has a 4 year string as unlucky as that jumps to 27.8% - roughly the probability that Carlos Beltran gets a hit in his next at bat - so it's certainly not unthinkable that a HoF caliber pitcher would have that kind of string of misfortune in his 4 best years - and where there's a lot more second-tier HoFers than inner circle guys, it would likely be a second tier guy if that happened. And for most second tier HoFers, if their W/L record in the best four year stretch of their careers is below the 30th percentile, it would indeed call their HoF credentials into question by many. So while any given HoF pitcher, taken at random, would have a relatively small chance of luck playing enough of a factor to impact his HoF support, the chance of that being the case for at least one HoF caliber pitcher is quite tangible, and if that is the case, the chance that people would be debating that one pitcher's candidacy would be quite high. If we extend our criteria to, say, at least 4 of his best 5 years, the probability jumps greatly.

From just eyeballing his stats, it's apparent that:
- His best performances, in terms of rates compared the the league, tend to be as a young pitcher, but he also had a handful of near-peak seasons after 30.
- While his performances were better in the first half of his career, his W/L records are much more reflective of what would be expected from his performances later in his career.
- His decline phase W/L record doesn't seem appreciably different than his peak ERA.

If he had some unusual quirk in the patterns of his performances that cost him wins early in his career, it doesn't seem to show up in the second half of his career. That suggests that only a relatively small subset of years in his career is subject to unusual patterns of luck, making it far more plausible than it seems at first glance that the differences could, in fact, be totally random.

Now, that's not to say the differences are totally random, but it certainly suggests that the degree of randomness involved is not unthinkable.

[Triad]

Originally Posted by WilsonC

I think some factors could definitely explain the discrepency in his ERA versus his RA, if indeed he had a pattern of a high discrepency. His fielding may have a slight impact, and his flyball/groundball rates, as well as any quirks of the type of contact he may have induced could have an impact, especially if he induced a lot of contact to a defensive weakness on the team. He was known for his devastating curveball, which may have been difficult for some catchers to handle, leading to more pass balls with him on the mound. Note though that these types of factors would contribute directly to runs allowed, so any such defensive quirks would already be accounted for with my methodology.

These are all plausible factors, I think. But, yes, they are mostly accounted for (except maybe for their distribution).

Originally Posted by WilsonC

One potential variable I could see having an impact in his performance versus win/loss record would be managerial tendencies. It's possible his managers used a different lineup with him on the mound (different catchers would be a possibility - easy enough to check). His managers could have taken particular care matching him up with other aces, giving him a disproportionate number of pitcher's duels with other top-notch pitchers (a little trickier to check, as we'd need to compare his opponents to the norm, and to the opponents of other aces). Also, given the number of complete games he threw, it's possible that he was simply rode too hard at times, and his managers did a poor job of relieving him when he was fatigued (This would likely show up as a disproportionate number of late runs allowed in his losses).

Good points. Definitely worth looking into.

Originally Posted by WilsonC

Rather, what we'd be looking at is random occurences over 527 decisions. However, even there it's not really 527 decisions we're looking at - we can remove the 60 shutouts from that, as there's essentially no luck involved as to whether a shutout will be worth a win.

Blyleven did play during an era when shutouts were more common than most, so in that sense he was riding the wave. He was among the leaders in shutouts quite a bit, but that tells us more about his patterns than it does his overall value. But you're right that we can essentially throw those out of the equation.

Originally Posted by WilsonC

1-run complete games would be less subject to luck as well, and he had plenty of those too (though I don't have his career totals handy). Also, his worst outings would be less subject to luck, as they would be almost assured losses. So we'd really be looking at a sample of no more than about 400 or so that luck plays a part of, a solid sampling, but not as assured as thousands of innings to wash out the noise.

That sounds about right.

Originally Posted by WilsonC

The other thing is that it's not quite true that luck evens out over time - rather it converges and is washed out.

Same basic effect though. When speaking of luck, it's hard to speak in technical terms, because luck isn't really a condition, but the description of results.

That is to say if you take a random sampling of a player's career where he happened to have a stretch of winning 5-10 less games than expected, it won't balance out with him making up those 5-10 games elsewhere - but rather given enough decisions, those 5-10 games would become irrelevant. When we're talking about a sample of 500 or so, a particularly unfortunate stretch of losing 13 more games than expected over a few seasons doesn't have time to wash out - it's the difference between 287-250 (.534) and 300-237 (.559). A big difference.

That's possible. But how possible? Every pitcher has had some good luck and some bad luck. A lot of it is to be expected. The best pitchers should be able to overcome a lot of it. Run support isn't as important a factor for the best pitchers, for example. It's most important for those pitchers who are right around the league average. We'll have to look closer at his run support vs. the average staff member's run support on his teams.

Also, 13 wins instead of losses in his career still wouldn't necessarily translate into a better peak. It might affect his best seasons by only one game or no games, and be borne out in a 12 or 13-win season, for example. However, that 13 would be enough to sway the voters.

Originally Posted by WilsonC

Consider this: If you were to take the best 4 year stretch performance-wise of any given pitcher's career, there's a 0.81% chance that he will fall under the 30th percentile mark for wins expected in every one of those 4 years, purely due to luck.

I'm a little dubious about wins expected methods. Where can I go to see this type of data? Part of the discrepancy, or error, may be in the process.

Originally Posted by WilsonC

That would be worth at least a few wins a year in those 4years, and would be extremely unlikely for any given pitcher to suffer such bad luck. However, is you take a random sampling of, say, the 40 best pitchers in history, the probability that one of those 40 guys has a 4 year string as unlucky as that jumps to 27.8%

If this is true, then we should see some other examples among the top 100 pitchers of all-time. What are your calculations on that?

Originally Posted by WilsonC

...so it's certainly not unthinkable that a HoF caliber pitcher would have that kind of string of misfortune in his 4 best years - and where there's a lot more second-tier HoFers than inner circle guys, it would likely be a second tier guy if that happened.

You're right, it wouldn't be unthinkable just based on that, but then if you consider that the trend carries over the bulk of Blyleven's career as well as his five best seasons, it says something about the validity of that trend. If Blyleven's best five seasons were anomalous to the rest of his career, then what you're saying would be more likely.

Originally Posted by WilsonC

- While his performances were better in the first half of his career, his W/L records are much more reflective of what would be expected from his performances later in his career.

Even though he had some good late seasons, he was all over the map from year to year. That tells us the data is less reliable.

Originally Posted by WilsonC

If he had some unusual quirk in the patterns of his performances that cost him wins early in his career, it doesn't seem to show up in the second half of his career.

Nine years is a pretty big sample, though. Pitchers get smarter as they get older, so he could have been more consistent later in his career, and not made the same types of mistakes he made in his 20s.

Originally Posted by WilsonC

Now, that's not to say the differences are totally random, but it certainly suggests that the degree of randomness involved is not unthinkable.

I think much of the reason for his ERA vs. W-L discrepancy could have to do with his apparent distribution of good games and bad games. When he lost, he was still fairly efficient, though not good enough to expect to win. But he didn't often fall completely apart. At the same time, a lot of his wins were of the 0, 1, 2 run variety, bringing his ERA down. But is a 0-run win really that much better than a 1-run win? ERA measures earned runs allowed straight across the board. But I think the difference between 4 runs and 3 runs, or 3 runs and 2 runs is more significant than the difference between 2 runs and 1 run, and quite a bit more significant than the difference between 1 run and 0 runs. I'm too lazy to do this right now, but we could look at how often each run total would've been good enough to win a game in his outings.

* * * * * * *

Incidentally, for what it's worth, Blyleven's main catchers were George Mitterwald, Phil Roof, Ed Ott, Tim Laudner, Glenn Borgmann, Jim Sundberg and Chris Bando. However, none of them were more than a full two-season's worth. Mitterwald had the high, with 79. Sundberg was stellar behind the plate. Not aware of the rest. Were any of them liabilities?

During Blyleven's career, a pitcher reached 20 wins in a season 146 times, yet he only had one of those seasons. 19 win seasons? 45 total, Blyleven one. 18 win seasons? 76 total, Blyleven zero. A pitcher that got above 17 wins only twice in his career makes you wonder, particularly in an era comprising the '70s when complete games were common and pitchers got more decisions.

Multiple 20-win seasons during Blyleven's career:

4 - Wilbur Wood
2 - Frank Viola
3 - Luis Tiant
4 - Dave Stewart
3 - Tom Seaver
2 - Bret Saberhagen
2 - Nolan Ryan
4 - Gaylord Perry
8 - Jim Palmer
2 - Phil Niekro
2 - Joe Niekro
3 - Jack Morris
2 - Andy Messersmith
2 - Dave McNally
2 - Mickey Lolich
3 - Dennis Leonard
2 - Jerry Koosman
2 - Jim Kaat
2 - Randy Jones
3 - Tommy John
4 - Ferguson Jenkins
5 - Catfish Hunter
3 - Ron Guidry
2 - Tom Glavine
3 - Mike Cuellar
2 - Joe Coleman
3 - Roger Clemens
6 - Steve Carlton
3 - Vida Blue
2 - Joaquin Andujar

Wow!

Some of those we'd say right away were better than Blyleven. But of the rest, can we really say they were all simply luckier than Blyleven to have multiple 20-win seasons? And remember, Blyleven had more than 17 wins only twice! How much bad luck can be attributed to someone?

[WilsonC]

Originally Posted by Triad

That's possible. But how possible? Every pitcher has had some good luck and some bad luck. A lot of it is to be expected. The best pitchers should be able to overcome a lot of it. Run support isn't as important a factor for the best pitchers, for example. It's most important for those pitchers who are right around the league average. We'll have to look closer at his run support vs. the average staff member's run support on his teams.

Again focusing on the 1970-1978 range (because that's as far as I've gotten in my worksheets ) the league average was 4.08. His teams (including Blyleven's starts) scored an average of 4.29 runs during that sttretch, or about 5% over the average. Isolating Blyleven's starts, his run support was 4.02 over that time, or 1% under the league average. It's certainly not criminally bad run support taken as a whole, but it's clearly well below that of his teammates. That's obviously not a factor in the pitcher's control, but selective matchups featuring more aces than usual is a definite possibility here.

Originally Posted by Triad

Also, 13 wins instead of losses in his career still wouldn't necessarily translate into a better peak. It might affect his best seasons by only one game or no games, and be borne out in a 12 or 13-win season, for example. However, that 13 would be enough to sway the voters.

That's why we'll look t some individual seasons, as well as the focus on his extended peak!

Originally Posted by Triad

I'm a little dubious about wins expected methods. Where can I go to see this type of data? Part of the discrepancy, or error, may be in the process.

The calculation here isn't using a particular method of win expectency, but rather it's assuming that given the sampling of performances by the pitcher, there is a range of wins and losses that he may end up with, due to variation in perfrmances of his team. I'm simply taking the bottom 30th percentile of whatever that range may be and stretching it over 4 seasons to give, by definition, a 0.30^4 chance of all four seasons falling below that 30th percentile, or 0.81%. As for the typical difference between the 50th percentile and the 30th percentile, I'm not sure how many wins that would be without further investigation - but it would certainly be enough to give the illusion of a pitcher's peak being lower than it is - especially compared with the 76% of pitchers on the opposite end of the spectrum with at least one 70th percentile season in their best four.

Originally Posted by Triad

If this is true, then we should see some other examples among the top 100 pitchers of all-time. What are your calculations on that?

We very well could see it, though I'd need to dig deeper to find examples. One important thing to note is there's a comparatively small subset of pitchers where it would really be the most heavilly debated. An inner circle HoFer like Walter Johnson or Roger Clemens could have that kind of bad luck and it would go unnoticed, because he'd be so far above the threshold that there would be no debate even if you severly dampen the apparent value pf his peaks. Someone even slightly below the HoF level would also go more unnoticed, because if someone has questionable credentials to begin with, a trait that makes him look less qualified will likely keep him out of most such discussions to begin with. Somebody has to have been the least fortunate of great pitchers - it may well have been Blyleven.

Originally Posted by Triad

You're right, it wouldn't be unthinkable just based on that, but then if you consider that the trend carries over the bulk of Blyleven's career as well as his five best seasons, it says something about the validity of that trend. If Blyleven's best five seasons were anomalous to the rest of his career, then what you're saying would be more likely.

Ahhh, but it's not uncommon for even a great pitcher to have a handful of .500ish seasons throughout his prime - it's just uncommon for them to be clustered as much as Blyleven's. Take the following record:

10-9
16-15
20-14
23-14
20-14
15-10
13-16
14-12
14-10
12-5
8-13
11-7
7-10
22-4
8-11
8-5
17-14
15-12
10-17
17-5
8-7
8-12

That's simply taking 4 of his good seasons (not necessarily his 4 best, it's just an illustration) and converting 3 losses to wins. It no longer looks like there's any real pattern - It looks like the typical record of a pitcher who had a great prime early in his career, regressed somewhat mid career, but occasionally put together a season worthy of his younger days later in his career. The handful of average W/L records are common, and not really noticed if he has the clearly excellent seasons to go with it. Take the four 20 win seasons, regress each by 3 wins (an arbitrary 'luck' estimate), and suddenly we see a pattern appear where there is not really any pattern.

Originally Posted by Triad

Even though he had some good late seasons, he was all over the map from year to year. That tells us the data is less reliable.

His later career is asctually much closer to what we would expect, generally. For the most part, he remained a good, but not a great pitcher after his peak. He had some very nice seasons where the record reflected it, some mediocre seasons where the record reflected it, and some years where he seemed a little unlucky. That's typical.

Originally Posted by Triad

Nine years is a pretty big sample, though. Pitchers get smarter as they get older, so he could have been more consistent later in his career, and not made the same types of mistakes he made in his 20s.

What's interesting about the theory of his records reflecting a lack of consistency is that it would be odd for someone who's unusually inconsistent from start-to-start to also be so unusually consisten from year-to-year. During the 1970-1978 stretch, his ERA never fell out of the 2.52-3.18 range. That's a very narrow range over a 9 year span. For it to be generated by an inconsistent pitcher would be a little unusual, albeit not impossible.

Originally Posted by Triad

I think much of the reason for his ERA vs. W-L discrepancy could have to do with his apparent distribution of good games and bad games. When he lost, he was still fairly efficient, though not good enough to expect to win. But he didn't often fall completely apart. At the same time, a lot of his wins were of the 0, 1, 2 run variety, bringing his ERA down. But is a 0-run win really that much better than a 1-run win? ERA measures earned runs allowed straight across the board. But I think the difference between 4 runs and 3 runs, or 3 runs and 2 runs is more significant than the difference between 2 runs and 1 run, and quite a bit more significant than the difference between 1 run and 0 runs. I'm too lazy to do this right now, but we could look at how often each run total would've been good enough to win a game in his outings.

From 1970-1978: (in parentheses: his record in games where he pitched at least 7 innings)
0 Runs Allowed: 41-0 (40-0)
1 RA: 39-10 (35-8)
2 RA: 26-17 (23-16)
3 RA: 15-27 (10-21)
4 RA: 12-29 (9-17)
5 RA: 2-20 (2-5)
6 RA: 1-20 (0-6)

Considering the average run support of this stretch was just over 4, he was definitely robbed in 3-4 run games. With ordinary luck, he'd be over .500 in 3 run games (particularly the 7+ inning ones) and just under .500 in the 4 run games (just over in his 4 run CG, but under in the others). Obviously you can't lose a shutout - but I'm a little curious about how his numbers here compare with other pitchers.

* * * * * * *

Originally Posted by Triad

Incidentally, for what it's worth, Blyleven's main catchers were George Mitterwald, Phil Roof, Ed Ott, Tim Laudner, Glenn Borgmann, Jim Sundberg and Chris Bando. However, none of them were more than a full two-season's worth. Mitterwald had the high, with 79. Sundberg was stellar behind the plate. Not aware of the rest. Were any of them liabilities?

Sundberg's the only one I'm familiar with.

Originally Posted by Triad

During Blyleven's career, a pitcher reached 20 wins in a season 146 times, yet he only had one of those seasons. 19 win seasons? 45 total, Blyleven one. 18 win seasons? 76 total, Blyleven zero. A pitcher that got above 17 wins only twice in his career makes you wonder, particularly in an era comprising the '70s when complete games were common and pitchers got more decisions.

Multiple 20-win seasons during Blyleven's career:

4 - Wilbur Wood
2 - Frank Viola
3 - Luis Tiant
4 - Dave Stewart
3 - Tom Seaver
2 - Bret Saberhagen
2 - Nolan Ryan
4 - Gaylord Perry
8 - Jim Palmer
2 - Phil Niekro
2 - Joe Niekro
3 - Jack Morris
2 - Andy Messersmith
2 - Dave McNally
2 - Mickey Lolich
3 - Dennis Leonard
2 - Jerry Koosman
2 - Jim Kaat
2 - Randy Jones
3 - Tommy John
4 - Ferguson Jenkins
5 - Catfish Hunter
3 - Ron Guidry
2 - Tom Glavine
3 - Mike Cuellar
2 - Joe Coleman
3 - Roger Clemens
6 - Steve Carlton
3 - Vida Blue
2 - Joaquin Andujar

Wow!

Some of those we'd say right away were better than Blyleven. But of the rest, can we really say they were all simply luckier than Blyleven to have multiple 20-win seasons? And remember, Blyleven had more than 17 wins only t