Friday, March 13, 2009

Unlocking Compubox

Punch stats provided by Compubox have become a staple of many boxing telecasts, but just how useful are these statistics in telling an accurate picture of what's taking place in the ring? Michael Nelson takes a look at the relevance of these oft-quoted numbers.

Shane Mosley smiled and shrugged his shoulders.

He had just thrown seven punches within a second and a half – so quickly that HBO commentator Jim Lampley incorrectly called it a four punch combination. Three of the seven landed. Or was it two? Is it possible none of them landed?

The difficulty in deciphering how accurate the blindingly fast combo was didn’t just lie with the speed it was delivered, but also in the shell-like posture of its target, in this case Ronald “Winky” Wright. Wright was a master at instantly reducing the surface area needed to land a scoring blow into infinitesimal space in between and around his forearms and gloves as soon as his opponent opens up. It’s an ability that has left a long history of foes befuddled and on the losing end of lopsided decisions, including Shane Mosley himself, eight months earlier.

It also leaves questions on how much guesswork goes into recording which blows land and which don’t.

“CompuBox numbers in round one [are] demonstrating exactly what Roy says about the problem of Shane getting to Wright. CompuBox found Wright landing 22 out of 45, and Shane… only 5 out of 59, all body connects – in other words, CompuBox didn’t see Mosley landing a single punch on Wright’s head in the first round,” Jim Lampley said near the beginning of the second round, with a hint of disbelief in his voice.

“I don’t believe that number,” Larry Merchant quickly interjected.

“That’s not true, he did land a few shots to the head, they just weren’t power shots,” Roy Jones agreed.

The CompuBox process of gathering punch statistics is a fairly simple one. Two operators are sitting side-by-side along ringside. Equipped with laptops, one operator carefully monitors the actions of a designated fighter while the other operator does the same with the opponent. They hit one of four keys whenever a punch is attempted by their assigned pugilist: jab connect, jab miss, power punch connect, and power punch miss. As the fight progresses, statistics are compiled for each fighter – how many jabs are thrown, how many are landed, how many power punches are thrown, how many are landed, and the percentages of punches landed out of punches thrown in each category.

photo © Ray Kasprowicz

It should be noted that any non-jab is considered a ‘power punch’, so a power punch isn’t necessarily a powerful blow. A grazing Paulie Malignaggi right hand is counted the same as a staggering Ricky Hatton left hook. In short, the CompuBox system measures the quantity of punches, not the quality of them.

The lead operator, generally the more experienced one, may note pertinent statistics or trends to the televising company’s production truck. The truck feeds the information into the headpiece of the lead commentator, who then relays the stats to the audience viewing the broadcast. The stats are also displayed during and after the fights.

An issue arises when these statistics are used as an authoritative source on what’s happening inside the ring. The operators at ringside pressing laptop keys, albeit trained in their craft, do not have the benefit of replay or slow motion. They’re dealing with live action, prone to compromised angles and perspectives that could prevent them from clearly seeing whether a punch lands or not, in addition to human bias and error.

More importantly, the definition of a landed punch is open to interpretation.

In this instance, Shane Mosley was throwing punches, often in rapid succession difficult for the eye to register, against a man who used his gloves and abnormally large forearms to get a piece of virtually every one of them. Within a matter of a second or two, the operator has to make a judgment call on whether a punch that caught both glove and face, or both elbow and abdomen, is considered a landed punch or a blocked punch.

When monitoring a Winky Wright opponent, a punch tracker is largely working within a gray area. Ten different people tracking landed blows on Wright might get ten different sets of statistics. And while Wright may be the most efficient practitioner of the shell defense, a large percentage of professional fighters use it to protect themselves. Even men whose defense revolves around head movement and distance cover up from time to time.

Still, it’s rare that you hear the type of objections Larry Merchant and Roy Jones threw out on air. Generally, punch stat numbers are used as concrete evidence that a fight is moving in a certain direction. Jim Lampley in particular seems to treat the numbers with special reverence, often using them to support the inclinations of HBO’s unofficial judge Harold Lederman.

The relationship between HBO and CompuBox is a close one. Owner and co-developer of CompuBox, Bob Canobbio, worked as a researcher for an HBO series called Boxing’s Best in the mid-80s. The man he worked under was then producer Ross Greenburg, now president of HBO Sports. After developing the CompuBox system with Logan Hobson, he sold the idea to Greenburg, and HBO started using the system to make their shows more interactive for the viewer. ESPN and other companies joined the fray after 1989.

A notable exception to the business of punch tracking is Showtime. One of the forgotten stories within the ongoing rivalry between HBO and Showtime is how HBO bought an exclusive deal with CompuBox after CompuBox worked a couple of shows with Showtime in the 80s. Since 1985, CompuBox has worked with HBO and only HBO on paid cable.

Ironically, the absence of punch statistics may offer Showtime viewers a clearer, more objective take on what’s transcribing between two combatants inside the squared ring. A reliance on what’s interpreted as a landed punch, along with the musings of an unofficial ringside judge, can easily blur the line between describing the action and dictating a story onto the audience. So when Max Kellerman asks “how can one fighter who’s out landing another fighter in every round seemingly two-to-one be losing a fight?” – in reference to the quantity of Chris John’s punches vs. the quality of Rocky Juarez’ punches in their draw last month – it isn’t difficult to see how faux controversy can be generated amongst boxing fans viewing the fight from their television sets.

That’s not to say CompuBox tracking is without its merits. Extensive video review of a fight can provide more accurate and comprehensive statistics that can be handy in analyzing a fighter’s strengths and weaknesses. An analysis of how active a fighter is, how many jabs, body punches, hooks, etc, are thrown, and how consistent a fighter’s output is throughout a bout can undoubtedly be useful information.

But saturating the television audience with punch statistics tracked live, and treating them as if they’re an infallible representation of events instead of an interpretative portrayal, is abusing the system for the sake of production values. More command should be yielded to the fans, especially ones who don’t know how punches are tabulated by CompuBox, in determining for themselves which blows are landed and which are even worth noting.

While it's easy to measure the quality of play through the use of live statistics in football, basketball, and especially baseball, professional boxing isn't a game. Stripped to its bones, the sweet science is about who is consistently hurting who the most.

That, you can't quantify. In this case, less is more.

e-mail Michael Nelson


dread said...

Nice piece. Compubox is horrible. I wonder how much bias is introduced just by having one guy for each fighter or how much different it would be if they alternated rounds?

Michael Nelson said...

Good point. The interviews I read with Bob Canobbio indicate that the operators follow the same fighter throughout, but it seems like it'd be little better if they alternated.