Week 261: Horses for Courses
The last thing I should write here about specialization as a terminal strategy.
[I mean it. I want to get done with the amount of stuff I write in the book against specialization, and why it’s counterproductive. So what is here should be more than what I end up putting in the book, and I will trim this down to size. But I wanted the catharsis of clearing it off the list, and I wanted to do it publicly.]
I am typing this out on a dead night at the Home Show, having composed the concept and first half during a dead morning at another home show. When my wife is doing these shows, I'm usually brought in to spell her on the slow nights in midweek. Tonight I got to tend to less than three visitors to the booth, which meant I knocked out a lot of text. Last Saturday when I took the morning shift at another show, I got another slow period which led to the passage that started, but did not finish this whole section.
Part of what I was doing on Saturday morning, was reviewing her inventory as we needed to be set for this week. This is probably counter to my peace of mind, but I do encourage her to have as wide a variety of products as possible. So when there is a sale of something that she had mostly given up on, I invariably respond with a bit of folk wisdom I picked up from an episode of Hu$tle, Robert Vaughn saying "horses for courses".
The fact that I'm using the setup of a con as formal advice should be telling. I know what it means, a saying for "different things for different tasks." But I've had it in my head that for a quiz bowl context, "horses for courses" is a trap and plan for failure.
In quiz bowl, the types of courses your team will be forced to run over exceeds the number of horses you have in your stable. That is if you have a player who is absolutely perfect for one particular category where they are absolutely dominant over the field at tournaments, they are still only going to help you maybe 25% of the time.
[To make the point I have to use a graph, which I didn't draw. I figure I'll have to for the book.]
If we make a graph of the position of people based on what they know in aggregate about a given subject and arrange them from left to right, we will see some portion of people who know nothing about it, and then a few people who know the subject exists, and then some that know only that the subject exists. We move right and see the people who know a little about it, maybe have read a little about it, or taken a course which touched on it, and then we get to the full course takers, hobbyists whom this is their jam, and those who major in it to the point of expertise.
If we were to simply take the knowledge and say how commonly known is the knowledge in the field, we might see something that we could convince ourselves was normally distributed. It's probably not, but the first idea you have in your head when you hear something taken in statistical averaging is a normal distribution. Changing the graph to what people know about a subject both is more accurate for our work and slips off that anchor. We can see that to know a lot about a subject requires moving across the whole graph, and not everyone is going to go all that far in accumulating knowledge. The graph, whatever form it takes skews highly to the left.
We have to impose some assumptions at this time, I don't think any of them are particularly egregious:
First assumption: There exists a sequence of knowledge that you accumulate in a subject. This probably has some statistical variance in aggregate over a wide enough subject, but it's true that there's certain things you need to know first among others in a field. So first principles or the most well-known events of a subject are going to be known by more people than deeper knowledge. I don't think we're breaking tremendous ground with deeper knowledge tracking with more obscure to the population taken as a whole.
Second assumption: Not everybody retains all their knowledge, and not everybody stays with a subject the whole way through. Not controversial, it's why survey courses and history podcasts exist.
Third assumption: Quiz bowl players are not an inaccurate sample of the general population, along these lines. They may retain more information, they may push themselves to go through a subject beyond the point of most, but they're obeying the same processes as listed in our previous assumptions. And if they are subject to the same processes, their distribution may scale or shift differently, but it's going to be roughly the same shape.
https://en.wikipedia.org/wiki/Gamma_distribution
What people know about a subject is probably best modeled not by a normal distribution, but by a gamma, chi-squared, or Rayleigh distribution. A normal distribution is our first impulse if we were to think about it as a distribution, but even from the start that's flawed, there can't be people with negative knowledge, or infinitely negative knowledge, no matter our opinion of humanity. Likewise there's no one with infinite knowledge of a subject, and the number of people with massive finite knowledge of the subject becomes vanishingly small.
Here's the other assumption which my reasoning hinges on: acquisition of knowledge in a subject is, at best, a rate constant process, and more likely the rate reduces at higher amounts of knowledge. Basically, to learn higher concepts in a subject takes longer than the basics. It could also be that higher concepts take longer or more effort to find a teacher, or to realize there's something there you don't know.
By phrasing our graph as percent of people knowing vs the aggregate percentage of knowledge in the subject, we can also measure time to acquire knowledge as distance across the graph. So the area under the graph represents what percentage of people know less about the subject than a particular person at that position on the graph. Here's where the analysis is key, and the graph would make it visible. Moving to acquire the same amount of knowledge at the beginning (say from 0 to 10%) and in the middle (say from 40% to 50%) has very different results. A little knowledge at the beginning puts you ahead of a much larger population. Even if it's shorter there's the matter of investing the study time in every step prior to that point.
Specialization is the idea that one can and should spend their efforts in one category, for our analogy above, the idea that one should make themselves the horse that absolutely dominates that one particular course. Of course there's lots of courses, even in just one tournament, and a coach can only enter so many horses in the race.
There are enough categories that could be specialized in that the pie is split into too many pieces. And most commitments to specialization occur in categories that aren't frequent enough to justify specialization.
Consider the expectation that a specialist covers science (a frequent idea), and in doing that they become dominant locally in science, so they lock it down for their team. Most science questions (save science bowl) make up no more than 20% of the subject matter. So if the player is expected to lock down 20% of the total round, they are relying on their teammates to cover the remaining 80%. With a four-player team, this is assigning more responsibility to the remaining players, and you've misallocated your resources, if your player who has the dedication to studiously cover a subject stops there, they are leaving more of the distribution to be covered to players who they feel are less dedicated.
Specialization is the idea that once one finds a subject worthy of specialization, they pursue it exclusively and continue to pursue it to mastery. When someone decides on specialization they do not typically intend to return from it. This is bad in two dimensions, both because mastery is almost always not necessary, and because mastery is ill-defined and not really testable.
Specialization to mastery sets an impossibly high standard to end the pursuit. Every positive reinforcement in getting questions could be percieved as a false positive. It is difficult to believe you have achieved your goal, until you reach the top level, and play against the top competition. But if you're specialized only in one thing, and your team isn't able to get you to the top competition, you'll never feel you've proven yourself, further pushing a wedge between you and your team.
Specialization also lulls you into a kind of unprofitable butterfly chasing. When you find something out past your bounds of difficulty in your specialty, and you miss it, you feel an obligation to go learn it, whether or not it could be useful in the future.
Specialization is emotionally satisfying. It fulfills the cathartic need of having a plan, enabling you to take control of the situation. It provides you something to focus on. It allows you to feel like there is a specific role during play. But it also provides a lot of negative reinforcements during play. If you feel that your teammates have let you down by not taking it as seriously as you do, or they're not pulling their weight, or you can imbue this specialization with a sense of moral superiority. You can feel like your specialization makes you essential to the team. None of these are true, all of these are traps.
Specialization also isolates the player from their teammates. If you plan to focus on the category to the exclusion of other subjects, you're not going to want to do the same things during practice, you're not going to listen to other questions with the same focus in practice, you're not going to want to even be part of practice. That may be the goal of the specialist, but it doesn't help their teammates to get better, and it's easier and more effective to have every member of the team working to improve.
The specialist's efforts may actually damage other members of the team's efforts to grow. It is always better to have more than one person who is trained in a category, and the specialist may see that as a threat or the rest of the team may see the specialist as sufficient, and thus any effort in the category by other people is unnecessary. As a coach, you must understand that a specialist is both temporary and not essential to your team. They will leave, and if their knowledge is not transferred to other players, it will be lost. A specialist may be loath to give up their knowledge to their teammates even if it improves the team.
By and large, specialization is a drain on coaching resources, or coaching oversight. If you are helping the specialist find the knowledge they need you are not giving the same attention to the rest of the team.
If there is one technique which can separate the positive aspects of specialization from the negative, it's this:
Learn when to stop specializing, and return the knowledge to the team.
Specialization without end is obsession, specialization with an end is a boon.
Consider that most players will never face the very top players of their chosen specialization. So a specialist doesn't need be the very best in the world, merely the very best in their league, or their event. And where on that axis will their opponents land? Probably 60-70% of the knowledge of the category that appears frequently, maybe 50% of all that could appear. Will the remaining knowledge change the outcome of a single game? Not in that field.
In most cases, the local group of opponents will not contain a large amount of players dominant in all fields. That is, most opponents you will face will tend to be on the left end of any particular subject.
The other problem is that really dominant players tend to be dominant not just in one specialization but in a number of fields. So specialization in one field and to the abandonment of all others actually gives the advantage to that kind of opponent.
All that time that would be spent getting to completeness could be used to specialize in two or three more fields faster, honed by the techniques developed the first time. That time could be used to teach a younger teammate what they need to know about the category or write a study guide for the future generations of the team. That time could be used to teach the study techniques that enabled a player to learn all that. If the specialization can be turned off, it can be turned outward.
If you must specialize...
Specialization, if relevant to your chosen field of study outside of quiz bowl, is fine, but if you're expecting it to improve your game, realize it's probably less likely to improve your game than learning a little about a number of subjects you know nothing about now, or learning a little about something that croses a number of subjects. When you realize there's a hole, an absolute lack of knowledge in a field in your skill set, it's easier to fix that than it is to work on moving at a higher level in a subject you already dominate.
[In the book I go after the idea that a team's best course of action is to identify holes in their knowledge, and reduce their size and scope until there's a large hole to fill. I do not feel it necessary to elaborate on it in this section.]
If you're going to specialize, you need to stop at a certain point and pick up something new. IF you successfully develop the skill to specialize, you've honed the skill and it doesn't make sense to see the end of it as a terminal state.
If you're going to specialize, you should choose a category which bleeds over into other fields. That increases the types of questions that you can answer because the answers you learned as part of the specialization are clues in other fields. This effectively multiplies the amount of the distribution covered by the specialization, possibly making it worth the resource allocation. Examples of these include:
Geography - Any time a writer wants to create a clue for a cultural, political, language, scientific, literary or artistic answer, they can ask the question beginning with "Where?" Geography may under 10% of the official distribution, but the knowledge gained in knowing where things are can slide to categories comprising over 50% of the distribution.
Literature- Adaptations of literature leak into Fine Arts both visual and performance, Film, and Television.
Mythology - Adaptations of mythology and use of names flow into all the literature connections listed above, astronomy, business branding.
Summary
1. There are too many courses that you can go through in the course of a career to have any faith in any one of your horses.
2. Specialization, as most frequently seen is a final act, the last thing to do. When it is in that context it is an act of frustration (against their teammates), an act of desperation (to achieve something against the world of quiz bowl), and an act of selfishness (to do without end).
3. Specialization separates the player from the team, and the coach. When you are fighting through a subject that no one else on the team can help you with, you are going to find it harder and a more lonely experience. It increases the chance that if you get stuck you ragequit alone, rather than work with your teammates to overcome the obstacle. Some also use it as an excuse to ragequit.
4. Specialization is rarely going to cover enough of the distribution to make it worth more than 1/n of the questions your team will face (where n is the number of players.) In most cases the player who chose specialization because they believe they can help their team because their teammates do less actually is leaving more to their teammates to cover.
5. It's more valuable and quicker to learn the what gets you from zero to 20 percent of a subject than it is to learn an equivalent increase in the middle or the end.
6. The problem with that is if you learn the studying required of specialization, you should not stop. You've now primed yourself to do something better, and you're seeing this as liberating you from having to do more with the talent you've developed. Paradoxically the best thing you can do with specialization is to stop it and take up another subject.
7. As a coach, you can't help someone dedicating themselves to specialization without removing instruction time from the rest of the team. If a player feels they have to specialize to get better, it's a slight indictment of them feeling unsatisfied by quiz bowl now. They were satisfied with quiz bowl before, so what has changed for them? What can you as a coach do to give them more satisfaction?
8. Even when successful, when a specialist dominates the local field in that category, their impulse is to continue with that category, when the more effective strategy is to switch. Once at the top of the local area's teams additional knowledge won't result in more wins, removing weaknesses in categories might.