st1\:*{behavior:url(#ieooui) }

/* Style Definitions */
{mso-style-name:”Table Normal”;
mso-padding-alt:0in 5.4pt 0in 5.4pt;
font-family:”Times New Roman”;

In America’s
Pasttime as in its politics, democracy is a wonderful but fragile thing.  Ten years after Major League Baseball first
gave its fans the option to vote for All-Stars, Commissioner Ford Frick took it
away again after 1957, when Cincinnati Reds fans stuffed the ballot boxes to
send all but one Reds’ starter.  This was
not even a spontaneous upsurge of local pride: through the late spring, the Cincinnati Enquirer had printed ballots
to distribute them easily to fans, and local bars even required customers to
submit ballots before they would be served. 
Not until 1970 were the fans put back in charge of picking the starters,
but it’s been in their hands ever since–even surviving another sabotage attempt
when Massachusetts hacker Chris Nandor was able to create a program that voted
for Nomar Garciaparra nearly 40,000 times to edge out Derek Jeter.

But while the fans have held the All-Star franchise for
almost forty years now, the ways in which they experience and understand the
game have undergone revolutionary change over that time–with consequences that
show up in their All-Star choices.  In
1979, what the average fan knew about players on other teams was limited to
seeing who was on the relatively few nationally broadcast games, a handful of
games each year against the home town club, and box scores and leader boards in
the local paper.  By 1994, that fan might
well have had cable TV, which showed a few games a week and hours of sports
news offering highlights and recaps of every game.  Fast-forward to the present, and the same
fan–or his/her child or grandchild–almost certainly is on the internet.  I don’t need to tell you the effect that the
internet had on the game–you wouldn’t be reading this article right now without
the internet.  This informational tool
has provided fans with a vast array of baseball information and entertainment,
whenever they want it–from archived footage of every game, to current and
historical statistics cut in an endless number of ways.  Among other things, the internet has allowed
fans to vote for players with statistics right in front of them, follow teams
far from their current location, and exposed them to different strains of
thinking about how to evaluate baseball players and teams in general.

 I decided to parse the last forty years of fan voting into
three eras: the Early Era, the Cable Era, and the Internet Era.  Since ESPN began broadcasting in late 1979, I
classified the period between 1970 and 1979 as the Early Era, and designated
1980 as the dawn of the Cable Era.  Since
fans were first allowed to cast All-Star ballots online in 1996, I considered
that year the beginning of the Internet Era. 
(This is a rough designation, as any changes in fan voting patterns
should be observable from the start dates but not in full effect until some
years later: cable TV wasn’t nearly as widespread in 1980 as it would be by
1995, nor was online access in 1996–not to mention connection speed–anything
like what it is today.)  A look at the
starting lineups for every All-Star game revealed several large shifts in
voting tendencies that took place during these eras.  In this article, I will document some of the
more interesting trends.




st1\:*{behavior:url(#ieooui) }

/* Style Definitions */
{mso-style-name:”Table Normal”;
mso-padding-alt:0in 5.4pt 0in 5.4pt;
font-family:”Times New Roman”;


With less information in the earlier years of fan All-Star
voting, we might expect that fans would simply elect the players with whom they
were already familiar.  Especially before
the Internet Era, that is exactly what happened: carryover All-Star voting from
one year to the next was stronger the farther back we look.  In the Early Era, 51% of All-Star starters had
also started the previous year; the figure dipped a bit through the Cable Era,
with an average repeat starter rate of 47%. 
By the Internet Era, only 36% of All-Star starters earned starts again a
year later, which is a statistically significant drop.  The drop-off in persistence suggests that
fans were learning about new players having good years and voting them in
rather than simply poking the hole next to the name of the player who they
voted for the year before.  If a player
underperforms what he did the year before, it is easier for fans to see their
statistics on the internet and change their vote accordingly. 

 The internet also enabled fans to learn about good players
sooner.  Through the Early Era, only 6%
of All-Star starters selected had made their major league debut within the
previous three years.  During the Internet
Era, that figure had doubled to 12%, a statistically significant difference.

 At least by one measure, fans’ decisions didn’t just change
over the entire forty years: they improved. 
The average VORP of All-Star starters for the full season in which they
were selected was 43.4 in the Early Era, just 41.6 in the Cable Era, but jumped
to 59.6 during the Internet Era.  The
caliber of All-Stars also rose in relative as well as absolute terms: during
the Early Era, 47% of All-Star starters led their positions in VORP at the end
of the year (for outfielders, this was defined as top three in outfield VORP),
and that dropped to 42% of All-Star starters during the Cable Era (for
designated hitters, I considered leading the position as having the highest
VORP among players at their position who were not All-Star starters).  Again, the Internet Era saw a great leap
forward: 55% of All-Star starters went on to lead their position in VORP that
season, another statistically significant jump.



 A truism of decision-making is that faulty information can
be worse than none at all.  Perhaps this
explains why the advent of cable TV not only failed to improve fans’ abilities
to send deserving players to the Midsummer Classic–it seemed to make them
worse.  Recall that both by absolute and
relative VORP, All-Star starter quality dipped from the Early Era and the Cable

 Did cable have a distorting effect?  Maybe. 
As it turned out, fans started to value different kinds of players
during the Cable Era–perhaps reflecting the difference between what’s
impressive in a box score and what’s memorable in a TV highlight.  As an example, remember that as late as
1996–the eve of widespread internet access–supposedly informed baseball fans in
New York
debated which rookie shortstop had the brighter future: Derek Jeter, or Rey
Ordonez, whose offensive game was nonexistent but whose glovework got him
almost nightly attention on “SportsCenter.” 

 One striking trend of All-Star voting in the Cable Era was a
lessened emphasis on the long ball.  Home
run rates changes a lot in general over the last forty years, so instead of
using raw totals, I adjusted the home run totals of every All-Star in every
year by that year’s overall HR rate relative to 2008.  With this measure, I counted up the number of
All-Stars that had hit the equivalent of 30 “2008 home runs” during their
All-Star campaign.  Throughout the Early
Era, 48% of All-Star starters hit the equivalent of 30 homers today; during the
current Internet Era, 46% of All-Star starters did.  However, only 37% of All-Star starters hit
the equivalent of 30 home runs during the Cable Era.

 If the sluggers weren’t seeing their three-day vacations
interrupted, who was?  Some of the more
prominent power hitters were replaced by speed demons, but not all that
many.  The Early Era featured 23.8% of players
who had 20 stolen bases or more, and during the Cable Era, that rose to 26.5%.
However, this number has remained high (25.5%) during the Internet Era.  These differences are not statistically
significant–though the home run difference in the Cable Era is.

 There did seem to be an increase in fan valuation of patient
hitters during the Internet Era.  The
difference in between OBP and AVG was .102 during the Early Era, .110 during
the Cable Era, and .151 during the Internet Era.  It seems that the valuation of walks did
start to take hold at the same time the internet did, although there was an
even larger difference in OBP and AVG during the Internet Era before Moneyball
came out than afterwards.

 I was curious if fans learned to account for park factors
better during the Internet Era, but apparently they already had been doing so.  My suspicion was that All-Star starters
before the Internet Era would have larger home park factors as fans looked at
raw batting numbers, but the average park factor of All-Star starters was right
around 100 (average) throughout the last forty years (100.7, to be exact).



 Of course, the internet doesn’t only give fans information
to make better-informed decisions; it also provides the tools to increase the
impact of their uninformed decisions, potentially rendering every online fan
base the digital equivalent of Cincinnati
in 1957.  Short of organized
vote-stuffing on a mass scale, however–and remember, we’re not talking about
the Final Vote here–the other big factor driving internet voting is the same
one that drives attendance (and, thus, paper voting): how the team is
doing.  During the Internet Era, there
was a huge, statistically significant spike in All-Star representation from
teams that sat in first place during the All-Star break.  In the Early Era, just 28.8% of All-Star
starters hailed from division leaders, and only 30.7% of All-Star starters did so
during the Cable Era.  Since the start of
internet voting in 1996, however, 40% of All-Star starters could look forward
to rejoining first place clubs to start the second half.

 Speaking of attendance, its importance as a driver of
All-Star voting has held steady over time. 
During the Early Era, the average All-Star starter played for a team
with attendance that was 28% above the league average, and during the Cable
Era, the average All-Star starter played for a team with 19% above average
attendance.  During the Internet Era, factoring
in online voting, the average All-Star starter had home attendance 22% above

 The real shift was around the middle of the Internet Era, as
more fans from big cities learned that they could move the vote themselves.  Only 35% of All-Star starters came from Boston, New York, Chicago, Philadelphia, Houston, or Los
Angeles (including all the various Angels’
“locations”) before 2002, but from then on, 53% of them came from those cities.  Stadium capacity no longer could serve as an
equalizer to limit big cities’ voting potential.

 Finally, fans have always rewarded players who have led them
to pennants and championships the previous year with All-Star selections, but
this has become less of a factor since the Early Era.  24% of All-Star starters played for teams
that had won the pennant the year beforehand during the Early Era, as compared
with 15% and 17% during the Cable and Internet Eras.  During the Early Era, 11% of All-Star
starters played for teams that had won World Championships the year beforehand,
compared to 7% during both the Cable and Internet Eras.



 New technology has always changed how we experience and
value professional baseball players.  This is especially apparent in All-Star
voting.  Fans have changed how they value
players for better or worse over time, and the All-Star game is an excellent
way to see that.

 While certain types of baseball purists will look down his
or her nose at the All-Star game as some unholy cross between an exhibition
game and a particularly crass marketing campaign, a deeper significance can be
gleaned from the signal of a more enlightened fan base cutting through the
noise of All-Star voting.  Fans whose
understanding of the game deepens over time vote with their money as well as
their All-Star ballots.  As the public
grasp of what contributes to a winning team becomes tighter, General Managers
must respond to that as well as the strategies of competitors–in other words,
they have to improve their players evaluation skills as well.  If the fans know that chasing the veteran who
has “won before” is not worth it, they will reward GMs who go after patient,
power hitters and other types of players who make baseball teams win–the sort
of guys the data tells us we should put on All-Star teams.



  1. Pat_Andriola says:

    Cool stuff, Matt. Glad to be writing alongside you.
    Just a question: Does the Final Vote follow these basic rules, or is it more along to lines of grass-roots voting and viral marketing?

  2. getdown says:

    the shift towards speed players in the cable era also reflects a shift in the game. the big turf cement doughnut stadiums influenced how teams played and guys like rock raines, juan sammuel, ozzie smith, vince coleman, etc became more desirable in vast outfields where balls sped over the green concrete and players around the basepaths. the number of guys with 20 or more stolen bases hasn’t changed but what about 40 or more or 50 plus?

  3. Matt Swartz says:

    for 40 SB or more, the numbers are: 7.5, 9.8, 6.4% for early, cable, and internet eras
    for 50 SB or more, the numbers are: 5.0, 6.4, 2.7%
    A lot of the issue that you are talking about are adjusted for by the fact that I normalized homeruns so it was really about who hit relatively more homeruns to their era.
    I’m also not quite sure that the concrete bowl stadium era didn’t cover the 70s as much as the 80s and the early 90s, right?
    I imagine the final vote will follow a similar rule, but that’s only a few years old I thought, isn’t it? So you couldn’t really compare. That’s also probably grass roots stuff like you mentioned too.

  4. JJNYC says:

    Interesting analysis — but I think you have your dates wrong for deciding what was the cable age and internet age. Your dates — while tied to certain events, are arbitrary.
    I bet this is probably why your starting hypothesis — that advances in technology helped improve voting — isn’t borne out by your numbers.
    In 1980, ESPN was in only 1 million of 20 million households that had cable — and there were probably close to 80 million households in the country at the time. So less than 1 in 80 households had ESPN. Also, the homes with cable in 1980 were largely affluent folks who were not punching all-star game ballots at a ball park. They were home watching their fancy WHT.
    I think 1987 is a better marker for the dawn of the cable era. In 1987, ESPN covered the Americas Cup in Australia — drawing coverage by the NY Times and applause from the world. Everyone wanted cable (and Nintendo).
    That said, the safest date would be 1993 when ESPN was in roughly 60 million households (of 96 million), and launched ESPN2.
    As for the internet — I think there are two flaws in the date you chose. First, 1996 was not be the “dawn of the internet era”. In fact, in 1997, only 36% of US households had COMPUTERS (much less the internet). Only 18% had internet access in 1997, and only 41.5% had internet access by 2000.
    It wasn’t until apprx. 2001 that 50% of US households had internet access, and that number only increased to 55% by 2003. In other words, the internet’s influence on the masses was minimal until relatively recently.
    (And again, there is the issue of the affluence of your voters. But we can put that to the side.)
    The other problem that your “Internet Era” stats suffer from is the fact that you have substantial overlap between the impact of Cable and the Internet.
    By 1997, there were 219 million TV sets in the US. And ESPN was in about 71 million homes, ESPN2 was available in 52 million homes, and ESPNews reached five million households. Chances are that the folks with internet access also had cable, which draws your conclusion (that it was the internet that made baseball fans and voters more intelligent) into question.
    I’d like to see what the stats/numbers say if you change the dates to:
    1987 and 1997 for Cable and the Internet, respectively.
    Also, I’d be curious to see what changing the dates to 1993 and 2001 for Cable and the Internet respectively show.
    (Or 1990 when Baseball Tonight went live).
    (Full disclosure, I had rabbit ears until 1996 when I got both cable and internet access at the same time).
    Also – as my buddy just noted to me when I was telling him about your piece – what impact did USA Today’s national coverage/circulation have on people’s votes.

  5. Dan Novick says:

    With regard to the difference between OBP and AVG…. hoe does that compare to the overall MLB difference? If the overall difference is growing, then wouldn’t we expect the all-stars’ difference to be growing as well?

  6. Matt Swartz says:

    Wow, JJNYC! You certainly put me to work! But you clearly put a lot of time and thought into your post so I’m more than happy to quickly go through the data.
    A couple things first, though, I admitted early in the article that the date choices were questionable and were only illustrating a general point.
    Next, that general point was that the internet made voters smarter– NOT cable. I even suggested cable might have made them dumber. And also, cable coinciding with internet is impossible to avoid. And I guess it’s still early in baseball too, so why can’t it still be the early era? We get into semantics 🙂
    Okay so running the same questions through the data…
    Using 1970-1986 as Early, 1987-1996 as Cable, 1997-2008 as Internet; Then, second numbers, using 1970-1992 as Early, 1993-2000 as Cable, 2001-2008 as Internet
    Same effect for percent of repeats by era: 47,52,35% first set of cutoffs;
    49,46,32% second set of cutoffs.
    Mostly same effect for three or less years since debut:
    8%,15%,12% first set of cutoffs;
    12%,5%,17% second set of cutoffs.
    Same VORP thing:
    40.8, 46.4, 59.6 by era first set;
    41.8, 55.2, 59.6 by era second set.
    Similar Stolen Base>20 thing:
    24,29,24% first set;
    26,27,21% second set.
    OBP-AVG diff– same effect:
    .102, .118, .152 first set;
    .103, .147, .149 second set.
    All-Stars in First place at break– same effect:
    31, 30, 40% first set;
    30, 35, 42% second set.
    Previous year pennant winners– clearly less of an effect:
    19, 17, 17% first set;
    19, 16, 18% second set.
    World Series– also less:
    8, 9, 7%
    8, 7, 10%
    Attendance above average– something fishy about 1987-1996 now that wasn’t observable in the data before:
    28, 14, 22%
    23, 21, 21%
    Big cities– not as much an effect:
    43, 30, 39%
    40, 24, 48%
    All in very similar results but a few small changes. I’d say the general point holds.

  7. Matt Swartz says:

    Dan, good question. I’ll check league OBP-AVG in 8 year intervals rather than adding it all up. We should have a clear picture of an effect if so…
    2008: .071
    2000: .076
    1992: .063
    1984: .064
    1976: .065
    So, some evidence of a change for sure, but the relative gaps are smaller than that in all-star voting. Consistent with teams valuing OBP more from their hitters and fans valuing OBP more by a lot.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: