Reads Bad: Looking at Yahoo's Robo-Roto Fantasy Recaps

The bad news: Yahoo has rolled out some robots to write fantasy football recaps. The good news: the robots write like weird children.
Share |
"I have created a slideshow of the Top 50 underboobs. You are out of business."

Image via Fark.

This future, which is a bleak future, is not even new. The fútbollers were reporting on Northwestern’s automated stats program—which have, in turn, become Yahoo's new and fascinating and horrifying robot-written fantasy football articles—back when New Meadowlands Stadium was indeed brand new. But if computers making news isn’t breaking news anymore, Yahoo’s kind of reporting is broken nonetheless. Fascinatingly and horrifyingly so, but broken all the same.

As with hearts, so it goes in sport: “Things get damaged,” sang Dave Gahan so preciously. Apropos more so, it was Built To Spill's Doug Martsch—or Yeats, if you're picky—who followed Yahoo down. “Things Fall Apart,” the IT center cannot hold.

We aren't the first site to note the inchoate poesie of Yahoo’s cyborg prose, or to indulge in that poesie ourselves. But, to leave aside the accidental literary triumph on display, I'd like to focus upon these recaps' innate readability. Which is sort of like "drinkablity," but for comprehension and the idea actually makes sense.

This would probably be a good place to mention that a L=A=N=G=U=A=G=E scholar I’m certainly not. I didn't even finish my undergrad. As a red-shirt, however, I did snow my coach for beer money to ‘research’ the aesthetic implications of text-to-speech algorithms—real ecstasy of influence stuff. And again, we're not on first. This guy went and—no pomo—beat me to it.

But we’ve got a much bigger problem on our hands than computers talking when we've moved on to letting them write. And I’m sore afraid that, after a few more seasons, there’ll not be a boat big enough to keep us sports writing humans afloat. It's probably best to be specific, here. Qualitative, even, to the extent possible. These post-game reports are, ostensibly, written by robots. And thanks to newsman “Charlie Skinner,” we all know how their half lives. There is no time to waste.

Like Paul ErdÅ‘s in ISBN 1-85702-829-5, I do know a thing or two (or 3, or 5, or 9, or 21) about numbers. And if you’ll notice my correction to this On-line Encyclopedia of Integer Sequences entry, I once had my very own electronic mailing address c/o Yahoo! Enough, certainly, to assert that—upon running their fantastic script through a variety of indices that, themselves, sound like Ivy League fantasies or boutique law firms (e.g. Flesch-Kincaid, Coleman-Liau, etc.)—it's crystal clear that Yahoo’s rubric appeals to our basest literary instincts.

I quote, verbatim, from the fantasy football robo-recap that Yahoo! sent to my new Google email address. Only the names have been changed, to protect against cyberpunks. You don't need to read the whole thing, although you're of course welcome to do so and offer any lineup tips you wish in the comments. But you'll get the idea of how these read pretty quickly.

As always, because I’d be remiss if I did not, I offer the SAMO caveat: Due to my strong personal convictions, I wish to stress that the following in no way endorses a belief in Yahoo!

[Chriscooleyhighharmony] picked up 32.22 points from Peyton Manning to edge [The Sheryl Yoast Infections] 134.47 - 132.77. Even though both teams hit their projections, [Chriscooleyhighharmony] put up 5.3% more than their projected 127.76 points to secure the win. Cam Newton led [The Sheryl Yoast Infections] with 23.62 points while Ahmad Bradshaw brought in 21.55. [Chriscooleyhighharmony] starts off the campaign with a 1-0 record, good enough for fifth place. [The Sheryl Yoast Infections] opens at 0-1 and in eighth place.

[Chriscooleyhighharmony] Smooth Moves

With 16.00 points, Matt Bryant had the highest score for a K in the league this week. Peyton Manning ranked seventh in the league in scoring with 32.22 points.[Chriscooleyhighharmony] had most of their starters beat their scoring projection (5 out of 8). By scoring 19.00 points, the Denver Broncos Defense beat their 13.98 projected points by 35.9%.

[The Sheryl Yoast Infections] Regret Tracker

Left Jeremy Maclin on the bench, where he scored 22.60 points and surpassed his scoring projection by 79.5%, the highest percentage of any player on the team. Did not start Jay Cutler, who led the team in scoring with 29.87 points. Only 3 of the 8 starters on [The Sheryl Yoast Infections] underachieved this week, but they still lost. Marques Colston scored only 77.0% of his 14.41-point projection, with 11.10 points. [The Sheryl Yoast Infections] ranked ninth in the league in starter/bench scoring ratio this week with 132.77 points from their starters and 118.44 total points left on the bench.  

What If [Chriscooleyhighharmony] would have beaten five other teams besides SmackYou InTheMouth this week. If [Chriscooleyhighharmony] had scored 44.12 more points, they would have beaten all teams in the league this week. [The Sheryl Yoast Infections] would have lost to five other teams besides [Chriscooleyhighharmony] this week. If [The Sheryl Yoast Infections] had scored 16.82 fewer points, they would have lost to all teams in the league this week.

Game Notes [Chriscooleyhighharmony] did not cover the 2.51-point spread, winning by 1.70 points. The 1.70-point margin of victory was the smallest in the league this week. Both teams topped their projections, but [The Sheryl Yoast Infections] performed better versus expectations than [Chriscooleyhighharmony] and still lost. [The Sheryl Yoast Infections] got 55.60 points from their RBs, the third-highest combined scoring output from the RB position for any team this week. BenJarvus Green-Ellis exceeded his projected point total by 48.2% in the loss, scoring 21.00 points against a projected 14.17. Peyton Hillis failed to meet his projected point total, scoring only 7.15 points of his projected 14.48.

So. This raises two questions: one, how much worse is this, really, than the average at Bleacher Report, and two and more germane here is: how slow does Yahoo think we are? If my calculations, and the reigning metrics for assessing this sort of thing are correct, the answer is that this page should be easily understood by a ten-year-old. That is, it's written at a fifth grade level. 

I hold no baccalaureate degree. I was able to finish grade school, though. (FYI: I went on to be graduated from J.B. Beck Middle School—as valedictorian, no less.) So, à la Charlie Hustle, allow me now to crunch home the numbers:

Flesch-Kincaid Reading Ease

Calculated on a 0-100 scale, the higher the score, the dumber the text. As in golf, lower scores suggest that the text is complicated to understand, and thus more likely to turn off the average sports fan.

206.835 - 1.015 x (words/sentences) - 84.6 x (syllables/words)

The next five metrics help determine a text’s readability to student grade level (United States-only).

Flesch-Kincaid Grade Level

0.39 x (words/sentences) + 11.8 x (syllables/words) - 15.59

Gunning Fog Score

0.4 x [(words/sentences) + 100 x (complex words/words)]

SMOG Index

1.0430 x square root (30 x complex words/sentences) + 3.1291

Coleman-Liau Index

5.89 x (characters/words) - 0.3 x (sentences/words) - 15.8

Automated Readability Index (ARI)

4.71 x (characters/words) + 0.5 x (words/sentences) - 21.43

Lest you allege I’m skewing jingoistic, here's a handy-cum-dandy exchange table for The Classical’s international readership. Green is good on this one:

N.B. Coleman-Liau and ARI rely on counting characters, words and sentences. Meanwhile, Flesch-Kincaid Grade Level, Gunning Fog and SMOG consider number of syllables and poly-syllabic words (i.e. ≥ 3 syllables). As with doping v. 'roids, opinion varies as to which index yields the best hard results. Obviously, it’s difficult to automate the tallying of syllables, as the King's English does what it may.

Finally, for all the Cartesians in the house, here’s how Yahoo’s prompt scored on Fry’s graph:

But if it’s the area under the curve you’re looking for, look homeward to Alton L. Raygor’s estimate for a truer differential:  

Okay, but what does it all mean

It means war, mostly. You see, the war the House of McLuhan started never was a cold one, and has never quite stopped. On and off-the-field, video vanquished radio eons ago in Internet time; as tit begets tat, the net has more or less killed video tsars like Howie Cosell and Erin Andrews dead. In a hundred years time, the apparent glasnost championed by the present Sabermetric administration will have surrendered to one thousand more despots—each one less enlightened, more draconian than the last. This is a pessimistic sort of futurism, admittedly, but look at things this way and it's clear that you Yahoo! at your own peril.

Singularity U’s Ray Kurzweil has long proven that the exponential growth of Moore’s Law is a real, mathematical constant endemic to media, as well. Football will be no different. Moreover, since such takeovers are inherently hostile coups (R.I.P. ART), we can’t exactly arm a militia, much less field a team, to tête-à-tête a cultural line we can’t even fathom blitzing.

Right now, this sort of apocalyptic dread might seem a tad quackish, if not altogether alarmist. But take a look around the grounds. We’re awful close as it is. Cribbing from my own piece for Disinformation, Big Lehrer Style:

From insulin pumps to robotic limbs to the chips embedded in Parkinson’s patients, a nascent Singularity—namely, that man and machine will be indistinguishable no later than 2045—has arrived ahead of schedule. With IBM’s Watson having bested both Ken Jennings and Brad Rutter at Jeopardy!, this gauchely bipartisan Mr. Smith came to Washington earlier this year and quickly deposed a caucus that included former Rhodes Scholar Jim Himes (D-CT) and trained nuclear physicist Rush Holt (D-NJ).

And what does RGIII (QB-DC) have to do with this shit?

He's fuel, mostly. Yahoo's swapping out of human for non- scans as the latest update in John Henry’s bout with the steam hammer. Doing so on the topic of fantasy football—where we're already accustomed to slack, haphazard writing—can be seen as just the first note of a long, sad song.

And so, in the short run, the rotisserie football fan is now having a much more autistic read. In the long run...well, the long run does not look good, if it's a story told by the robo-scribes. We should hope that those underpaid to write about Sportin' Life will not go both-knees-down quietly, at least for a few more snaps. If we do, then, per that old Sondheim flight song, send in the Cylons. 


Share |

Comments

No comments yet. Login to post comments.