The June 1976 SI article was published shortly after Bert's first appearance with the Rangers, in which Bert and Mark "The Bird" Fidrych each went 11 innings, with the Tigers prevailing, 3-2. Including Bert's last two starts with the Twins, this made the third consecutive start where Bert had pitched well and been tied going into the late innings but lost. But the SI article wasn't a product of Bert's disappointing results in tight games over the preceding few weeks. The SI article was prompted by two conspicuous aspects of Bert's record that had persisted for years. First, Bert's W-L record never seemed to match the rest of his record - the superior ERAs, the shutouts, the complete games and the strikeouts. Second, Bert had a propensity to lose a lot of close games in the late innings.

Bert's fanatical supporters always have two deceptively simple explanations at the ready for Bert's mediocre W-L records in the '70s: he received poor run support, and he was unlucky. Each of these two rationalizations offered by Bert's backers fail completely to explain Bert's relatively poor W-L records, and each are particularly absurd for having been offered by people who purport to possess some degree of sophistication in statistical analysis. Each can be dismissed quickly and definitively.

Run Support. Bert's run support was slightly below average for much of the '70s. In this respect, Bert's backers are correct. But Bert's run support in the '70s was enough so that a pitcher with Bert's record of stinginess in allowing runs should have had a W-L percentage of approximately .600!

Bert had completed six seasons in the major leagues when the SI article appeared in June 1976, and had allowed an average of 3.16 runs/game while receiving approximately 4.0 runs/game from the Twins. Plug those two figures into the so-called Pythagorean Theorem with which any of Bert's statistically inclined supporters is familiar and you receive a projected winning percentage of approximately .615. Bert's actual winning percentage from 1970 to 1975 was .528. The Pythagorean Theorem suggests Bert should have been able to compile his .528 winning percentage while receiving only 3.35 runs/game from the Twins.

This disparity between Bert's actual record and his projected Pythagorean record persisted for the remainder of the '70s. At the end of the decade, Bert's winning percentage was .536. The Pythagorean Theorem says Bert should have had a winning percentage of .602.

To put it another way, if Bert's run support in the '70s had been 3.3 runs/game, Bert's backers would have a point about poor run support accounting for Bert's poor W-L record. But Bert's run support was approximately 4.0 runs/game during the decade, approximately 20% higher than the 3.3 runs/game that might have explained Bert's .536 winning percentage.

Bad Luck. It is particularly curious that Bert's backers would resort to this argument, one borne of superstition and anti-rationalism rather than the rigorous statistical analysis Bert's backers purport to favor. It is a wholly unworthy argument - even silly - for two simple reasons. First, the suggestion that Bert's "bad luck" could persist for a solid decade, across 350 starts and more than 2600 innings, following Bert from Minnesota to Texas to Pittsburgh, is pure nonsense as a statistical matter. Second, it is particularly nonsensical given the abundance of statistical evidence demonstrating that Bert had an astoundingly bad record in tight, low-scoring games and that this record was attributable to Bert's unusually poor performance in the late innings of tight games.

The rap on Blyleven, as expressed in the SI article, was that you could get to him in the late innings, and the statistics bear out that reputation. The following table shows Bert's performance in "late and close" situations (i.e., plate appearances against Bert in the 7th inning or later with the batting team tied, ahead by one, or the tying run at least on deck).

The pertinent numbers here are in the last two columns - the tOPS+ (Bert's OPS+ in late and close situations relative to his general OPS+ for the season) and the sOPS+ (Bert's OPS+ in late and close situations relative to the general league-wide OPS+, which by definition is 100 each year). Throughout the '70s Bert suffered declines in performance in late and close situations greater than any other elite pitcher of the era - his tOPS+ for the decade was 119. From 1980 to the end of his career the more mature Bert compiled a very good 86 tOPS+. Not coincidentally, Bert's winning percentage of .533 after 1979 almost exactly matches his projected Pythagorean winning percentage of .530, a stark contrast to the huge disparity between Bert's actual and Pythagorean winning percentages in the '70s.

Bert's improved performance in late and close situations from '80 to the end of his career coincided with a dramatic drop in the number of plate appearances against Bert in late and close situations. During the '70's approximately 14% of the plate appearances against Blyleven came in late and close situations. This figure dropped to less than 9% after 1979. This change is explained primarily by the fact that Bert was pulled from the game earlier in the '80's when he got in trouble in late and close situations, which contributed greatly to Bert's improved late and close performances later in his career. Consequently, Bert faced on average more than 8 batters per late and close game in the '70s, but just over 6 batters from '80 to the end of his career. This trend started in '79, when plate appearances against Bert in late and close situations dropped dramatically as a result of Chuck Tanner's decision to pull Bert at the first sign of trouble. The strategy worked spectacularly for the Pirates, who won 23 of Bert's 37 starts that year (a .622 winning percentage) despite the fact that in the great majority of Bert's 20 no-decisions he left the game either tied or behind.

The late and close statistics reveal that in the seven seasons preceding Tanner's '79 "quick hook" strategy, Bert had five seasons in which his tOPS+ was greater than 130 and his sOPS+ was 100 or greater. During these seven seasons Bert's tOPS+ was 128. No other elite pitcher of the era comes close to this record of performance decline in late and close situations over such an extended period.  Bert's sOPS+ was 105, meaning that in these five seasons Bert was an average or below average pitcher in late and close situations. Of the other premier pitchers of the era who, like Bert, had a dozen or more seasons in which they faced 100 or more batters in late and close situations, no other pitcher had more than three seasons in which their tOPS+ was greater than 130 and the sOPS+ was greater than 100. Steve Carlton and Gaylord Perry each had three such seasons and a fourth season that nearly qualified, but these were spread over careers more than 20 years long. The occurrence of five such seasons in a period of seven years during the peak of Blyleven's career explains to a significant degree the striking discrepancy between his actual W-L record and the kinds of W-L records projected by the Pythagorean Theorem.