The Troubles with CompuBox
This article was published by Boxinginsider.com on July 8th, 2017: http://www.boxinginsider.com/columns/the-troubles-with-compubox/
Boxing is a notoriously difficult sport to score. Although the brief a fighter must follow is simple enough – hit and hurt your opponent more than he does you – deciding who completes this task more successfully can be a complicated affair. Witness the myriad of disputed decisions that litter boxing history as evidence of the above. In the wake of Manny Pacquiao’s defeat to unfavoured Australian Jeff Horn last weekend, another contentious result can be added to that list.
Much was made in the controversy that followed the Filipino’s defeat of his apparent dominance according to the CompuBox punch statistics. Despite being unanimously declared the winner on the three judges’ cards, Horn allegedly landed just 92 punches of 625 thrown (15%) compared to 182 of 573 (32%) for the Pacman. While these stats seem to provide objective support for the idea that the judges’ scores did not accurately reflect the action in the ring, there are several reasons why we should handle the punch data with extreme caution.
First of all, while the name “CompuBox” might evoke images of a supercomputer programmed for the specific purpose of calculating the winner of boxing matches, the truth is somewhat more prosaic. In reality, CompuBox means two guys sitting at ringside with a laptop, each with the job of watching one boxer and four keys to press that record the punches that boxer attempts. The four buttons correspond to jabs thrown, jabs landed, power punches thrown and power punches landed. At the end of each round the numbers collected on the laptop are compared, and hey-presto – there’s your CompuBox punch stats.
So while the name sounds technical enough, we have to remember that there is a considerable amount of room for human error. For instance, recording when a single punch lands or misses sounds fairly straight forward, but telling if a blow grazed the gloves or connected with the side of the face can often be tricky even with the benefit of slow motion replays. In real-time as the action unfolds, it is far from an easy task. Consider also a body shot thrown on the blind side of the CompuBox operator: it could have been blocked by an elbow or it could have been a fight-changing rib-cruncher and they may be none the wiser.
The potential for human error only increases when we take into account combination punching. One-two-three-four – rat-a-tat-tat! – fired in a flurry of blurring gloves and grunts in less than a split second. Could you be sure that you’d accurately gauge, in real-time and without revision, exactly how many landed cleanly? Of course, a focused CompuBox operator would no doubt perform the task much better than the average viewer, but over the course of a twelve-round fight that would still leave a significant margin for error.
Even supposing, generously, that such errors could be reduced through careful training to a fairly negligible amount, the types of punches recorded are still liable to present a misleading picture of how a fight is unfolding. That’s because, as noted, CompuBox operators are faced with a choice of only two types of punches: “jabs” and “power punches”. The problem with this description is that the actual power behind a particular punch has no bearing on it being categorized as such. A jab is simply a straight punch thrown with the fighter’s lead hand; everything else is considered a “power” shot, by CompuBox definition.
In the words of one the co-founders of CompuBox, Bob Canobbio, “We call a non-jab a power punch for lack of a better description… we call it power punch because it sounds better than non-jab.” While this might be a great recipe for a catchy, TV-friendly sound bite, it’s a terrible method for recording which fighter is actually landing the most damaging blows.
Ignoring these practical limitations for a moment though, a more fundamental problem exists: the difference between tallying punches and scoring rounds. For while it’s a convenient formula to regard the fighter who lands the most as the winner (as the old amateur, Olympic-style scoring system used to do), professional boxing depends on much more than this. Specifically, the four scoring criteria are: (1) accurate punching; (2) effective aggression; (3) defence; and (4) ring generalship. What CompuBox seeks to provide is an objective measure of the first criteria. What it does not provide is an accurate reflection of all of the criteria judges are adhering to. In this sense, it can be a dangerously flawed tool in assessing who won a professional boxing match, even assuming the punch stats are 100% accurate.
Of course, it’s easy to dismiss the importance of “effectiveness”, “defence” and “ring generalship” as wordy and intangible. On the contrary, they are in fact very real and concrete: they are the difference between a shot crashing into an arm, or it being rolled over a shoulder; they are the difference between a fighter pushing his opponent back into a corner, or being lulled into a trap; they are the reason why Willie Pep and Pernell Whitaker star in the Boxing Hall of Fame, and Eric Esch (aka “Butterbean”) starred in Jackass: The Movie.
Consider, for example, the following exchange: Fighter A throws a fast series of looping punches as Fighter B backs into a corner. Let’s imagine that three punches out of a six punch combination land – but none of the blows does damage, and are either lacking in force or partially deflected by Fighter B. Then – boom! – Fighter B unloads with a well-timed, accurate jab-shot that rocks the head back of Fighter A, before nimbly skipping off to the other side of the ring with his opponent ambling after him.
Now, according to a CompuBox reading, in the above exchange Fighter A out-landed Fighter B by 3 power punches compared to 1 jab – an impressive statistic. He also showed more “aggression”, throwing 6 punches to his opponent’s 1. Any astute ringside observer would know to discard the significance of the punch stats in this instance though. Not only did Fighter B render the incoming attack ineffective with his superior defensive skills, he also controlled the flow of the action by enticing his opponent to throw punches only to achieve the goal of countering him, which he did accurately and effectively with a more damaging blow (that was not recorded as such), before lulling his foe towards another trap.
This is called ring generalship: making your opponent fight your fight and dictating the pace, range and terms at which the action takes place. Too often, punch data simply does not pay heed to these nuances. Multiply the above exchange by a few times per round over the course of a twelve-round fight, and you start to get a very good idea of just how skewed any reading of a fight based purely on CompuBox stats could become.
One final problem is that the final punch statistics for any given fight are never revised. That means however many punches are recorded on the night, in real-time, remains the “go-to” data for that match forever more. This is a real shame. While it is obviously useful during the live airing of a fight to provide viewers with a measurable guide to the action unfolding in front of them, there’s no reason why these figures could not be scrutinized and corrected utilizing available technology in the weeks that follow.
Consider how much more meaningful these numbers would be if a panel of observers reviewed fights and re-tallied the punch stats using slow motion and different camera angles to assess them more precisely. It would then be possible, for example, to extend the number of categories of punches to provide a fairer reflection of who is the more effective aggressor (i.e. who was actually landing the “power punches”). As a starting suggestion, “Damaging Blows” could be added to include those that land with greater visible effect or force, snapping the head back, producing a noticeable facial reaction or clearly hurting the opponent. This kind of revision might not be practical for every single fight, of course, but it would certainly be a welcome addition for replays of the biggest PPV contests.
All in all then, considering the potential impact of human error on the numbers recorded, the somewhat spurious categorization of the kinds of punches thrown, and the notoriously subjective nature of boxing’s four-point scoring criteria, what we are left with is a system containing significant flaws. The wider point here though is not that CompuBox is a completely useless tool or that it should be abandoned. The point is rather a cautionary one: while CompuBox stats can provide valuable insight into the activity unfolding in the ring and a fascinating guide to understanding the ebbs and flows of a particular contest, we should remember that it is ultimately just that – a guide.
As the CompuBox website itself clearly states: “CompuBox stats in no way, shape or form, determine a winner of a fight. The stats are used to enhance a telecast, show the estimated barometer of activity by both fighters and paint a picture of the activity on a round-by-round basis.”
We would do well to remind ourselves of these words more often in the aftermath of a controversial decision. The troubles with CompuBox only arise when the numbers are cited as incontrovertible “proof” that a fighter was dominant or used to justify cries of “robbery” – without putting them in context of the wider judging criteria, or considering that they could just be plain wrong.