A Substitute for War

Basketball philosophy

Information Gain: The secret to a must-watch event

with 11 comments

Los Angeles Lakers Magic Johnson and Boston Ce...

Image via Wikipedia

TrueHoop makes an argument today in an article about the importance of parity in praise of uncertainty in sporting results that I think almost completely misses the mark:

Research suggests that more uncertain outcomes lead to more certain income, or … more pie.

There’s a reason that the TV deals for the NFL and the NCAA basketball tournament both dwarf the NBA’s. In just about every game of the NFL season, and in just about every game of the NCAA tournament, you simply must watch to know what’ll happen. It all matters. You wake up the morning of the game with almost no ability to pick any winners. That’s the kind of thrill-ride that leads to enraptured fans and huge TV income.

Don’t confuse luck with parity 

Really, in every game you don’t know what happens in football and college basketball, but it’s a given in the NBA? Pshaw. I’ve tackled the issue of certainty/uncertainty here before on several occasions. Here’s the table from when I compared March Madness to the various pro playoff systems:

The left number here shows the winning percentage of the top two teams in each “league” during the regular season. The right number shows their likelihood of being champion. In the regular season, college basketball and the NFL are the most predictable results around, and there’s nothing magical that changes this come playoff time.

However, the playoffs do see both of these sports take a hit in the certainty of final results of the playoffs relative to the NBA. What causes that? It’s the single elimination silly. The whole justification for having playoff series to begin with is that they make it more likely that the better team will win. (Note that’s ‘justification’ not reason which always had a lot to do with more games yielding more revenue.)

Let me say that a different way: It’s not parity, it’s luck. Yes parity would help matters, but the edge the other sports have is entirely based on choosing a playoff format with more randomness. Is one truly willing to say, that every sports league out there willing to adopt a strategy that minimizes actually merit?

It’s not as grim as all that though. Go look at the ratings of NBA finals over the past 30 years. An obvious trend emerges that has nothing to do with parity or luck: Ratings go up when the biggest stars are playing. I know that seems obvious, but consider that the biggest stars will tend to play on favorites, and thus if we up the luck in a playoff system, that would lower the chance of the biggest stars playing in the finals and lower ratings.

So if uncertainty in results and certainty that stars will be in the finals are both beneficial, but work against each other, what’s the unifying principle here?

Information Gain as hook: Yeah, that’s entropy (reduction), man

The key is Information Gain. Hopefully that term makes some intuitive sense to you. To help out a little bit more, you can think about it as what you learn by watching a particular event. If you’re satisfied with that, you can skip ahead past the rest of this paragraph – just know that this isn’t something that’s my idea, it’s something that has had tons of thought put into it in other contexts. For those wanting to explore more on this, Information Gain is a concept from Information Theory that refers to the reduction in Shannon Entropy as a result of an observation.

Here’s the concept applied to the facts above:

Fact 1: People prefer single elimination contests because they know one team will definitely lose everything as a result of the match. This yields much more powerful information about the team that will end up winning the title than a random Best-of-7 series game does, and thus you gain significantly more information in a single-elimination match up.

Fact 2: People tend to prefer watching games with the biggest stars at least partially because they see those games as more meaningful. Yes there is also a matter of emotional attachment to the stars (a tangential but not contradictory point), but there is also a core truth that if every year a different set of players are battling for the championship it begins to sink in that the fortunate players in any given year aren’t likely to be the standout players in subsequent years. Thus the urgency of the moment is diminished as it becomes clear that the watching this year’s event lead you to gain much information about of future results.

(What’s that you say, the NBA may need this, but college basketball is still doing fine? Don’t fool yourself. College basketball finals nowadays don’t get near the buzz of say, the Magic JohnsonLarry Bird finals back in the day. Doing fine is not the same as achieving ideal performance.)

Last I’ll note that while this thinking applies to sports, and can just as easily apply to a soap opera, or any other piece of sequential entertainment. It’s not just the uncertainty, else coin flipping would be a popular spectator sport.

What people need in order to really get reliably hooked into watching an event, even if they already have general interest, is the feeling that they are missing something important by not watching the event.

The NBA’s Playoff Situation

Let’s go back to the NBA here. With the 7-game series system that they have, combined with the relatively luck-independent realities of the sport (much less luck than baseball, soccer or hockey), this means that watching any particular playoff game is not likely to reduce much entropy. Early in any playoff series, I not only know that both teams have additional chances to win, but that the superior team is likely so confident in their abilities that losing the game doesn’t make them or their fanbase sweat. That’s the bad.

The good part is that you don’t win an NBA championship by luck because of this lack of randomness. People may not watch every game of the playoffs, but they do know that what is happening is quite meaningful, and this certainly helps revenue to some degree.

If the NBA is not satisfied with this, what should they do? Well it’s all well and good to say they should increase parity, but that’s actually not very easy to do in a sport where the value of superstars is as tremendous as it is in basketball. I’d say its looking to play around with luck that could conceivably help them.

Consider that in a Best-of-7 series, the further along you get in the series (presuming the series remains competitive) the greater the perceived Information Gain. That translates into better TV ratings, but how big of a deal is that?

Well, the average rating for the NBA finals last year was a 10.6. By contrast the NCAA finals this year got a 11.7. However, Game 7 of the finals earned an 18.2 share. The difference between being beaten by the NCAA, and beating the NCAA salary, really is just a matter of having more at stake in a particular game.

Of course, there is also the matter to consider that a Best-of-7 series means there will be a lot more games. The goal is not to maximize the revenue on a per game basis, it’s to maximize total profit, so there’s a downside to reducing the number of games.

Still if the NBA truly comes to the conclusion they are getting their collective butt kicked by March Madness, it makes sense to try to move more in that direction. That would mean looking to reduce the number of games in some of the series (I don’t think anyone would cry if we went back to a Best-of-5 first round), and getting more serious about properly seeding the playoff teams both initially and after every round to maximize the chances of having a series go deep.

11 Responses

Subscribe to comments with RSS.

  1. I would go as far to say the “What’s the outcome?” question drives not only sports but almost every form of popular entertainment – movies, television, fiction novels, video games

    In sports for the reasons you mentioned it’s a balanced between a) How unpredictable is the outcome and b) How much does this outcome mean. For that reason, Roger Federer playing the 120th best player in the world in the 1st round of Wimbledon and the 64th and 65th best players squaring off are equally “flawed” from an entertainment standpoint. In the first case you know what will happen, in the second it appears to be 50-50, but since the outcome doesn’t really mean or effect anything, nobody cares

    Likewise in forms of fiction (movies/television/novels), the way to hook people is to set up something for them to “find out” by the end of it. But as with sports, this outcome has to matter – so it’s important the characters and plot matters to the consumer. Which is why dramas can still end up much more beloved than whodunits with stock characters, even if the latter have a clear hook of finding out an answer by the end of it. Of course the dramas usually have a love triangle somewhere which is an unstoppable device for making viewers keep at it for ‘the outcome’

    And of course in video games the ‘what’s the next level’ and ‘what happens when I level up’ is by far the strongest device for keeping people playing

    Julien Rodger

    April 20, 2011 at 3:28 pm

    • Good thoughts Julien.

      Re: Federer vs 120th. I do have to admit that that match still probably gets more viewers than the 60th vs 61st match on a typical day. I only mentioned the impact of stars briefly, but by no means would it be right to imply that closeness of the match is all that matters.

      That said, part of the reason for the interest in the Federer-120th match is that in watching Federer you are likely to be gaining information on someone who will be relevant at the end of the fortnight. Whereas in the 60-61 match, whoever wins, still probably loses a day or two later.

      Worth noting a key counterexample though: The Isner-Mahut match from ’10 Wimbledon which lasted for 11 hours. If something happens that is truly singular in a match, that trumps the stars.

      Matt Johnson

      April 22, 2011 at 11:15 am

  2. I’m just excited there’s a post on a sports blog discussing entropy.


    April 20, 2011 at 4:20 pm

  3. Phil Birnbaum over at Sabermetric Research wrote some posts about uncertainty of outcome too. I don’t recall them in detail and I haven’t got the time to read them now but if I remember it correctly it was good writings as usual.




    April 21, 2011 at 8:32 am

  4. Hm, I skimmed them briefly and I think I mixed them up with some stuff about competitive leagues and salary caps. All those years and all those posts, it’s all a blur to me!


    April 21, 2011 at 9:01 am

    • Welcome Mogilny!

      Those are excellent articles, and are relevant to the point I’m making.

      Is it clear though that uncertainty of outcome is only one facet of what I’m saying with Information Gain?

      Matt Johnson

      April 22, 2011 at 11:24 am

      • Absolutely.


        April 26, 2011 at 7:49 am

  5. Long series don’t help with macro-narratives (especially with a clear superiority established early on), but they’re gold for micro-narratives, given that individual player performances have a high degree of variance. You get individual failure, redemption, one-on-one-rivalries and quite frankly a lot more opportunity for wild variance like bench-clearing brawls and suspensions.

    It’s a superstar’s league, as you note, and the NBA will ensure that there’s enough room for the limelight to settle on each of them before the result is known…


    April 21, 2011 at 4:11 pm

    • Well, you’d think so, wouldn’t you? But yet people seem to be able to make a morality play out of a coin flip. The Bulls didn’t exactly dominate the Jazz in the finals, yet those results are certainly the difference between Malone being talked about as a choker (as he often is) and him being talked about as something more admirable.

      Matt Johnson

      April 22, 2011 at 11:27 am

      • Well… 6/24 and all that.

        People absolutely retro-fit their narratives to the eventual outcome, but here’s the rub…

        In a seven game series, there is a lot of information to be absorbed and picked over, ranging from complex derived metrics to simple box score statistics to highlights, to personal rivalries, body language and (for that matter) locker-room interviews. That makes a lot of data that can be analysed and used to contextualise the eventual result.

        An example… Lebron is widely thought to have choked away last year’s eliminator with a relatively sub-par final game performance. But if you look at his performance in the other LOSING games of that series, he was shooting something like 44 / 8 / 8 @ a 68% TS clip. His performances in losing efforts were inhuman, the team just wasn’t good enough to capitalise on these performances. But in the popular imagination, it’s his loss and counts exclusively on his legacy.

        So looking at all that information, what would that tell you about Lebron as a player and the Cavs as a team. And did that narrative end up as the popular perception of that series?

        Humans are bad analysts, and they’re especially bad analysts of aggregated knowledge. Whether you can parse that into a call for shorter NBA playoff series, I’m not entirely sure, of course. 😉


        April 23, 2011 at 6:22 am

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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: