REVIEW:

The Visualization of Quantitative Data, 2nd Edition, Graphics Press, 2001
Edward R. Tufte, Author.

People ask me, after I explain the layers of thought that went into the play aids I design, what kind of books I’d recommend.  The first book is this one, by Edward R. Tufte. It concerns itself with the accurate display of statistical data in its examples.  It offers more to the game designer than simple rules of thumb regarding the intermixing of graphics, tabular data and text. The lessons he presents are quite useful.

Among the more notable of them, with direct applicability to a wargame designer (especially in the creation of play maps and reference aids, or making user interface controls for computer games):

* Make your displays wide rather than tall. The human eye sees this as a horizon; moreover this is the ideal format for presenting data over time (something germane to thrust charts for AV:T). A ratio of 1.5:1 width to height is considered close to ideal.

* Eschew unnecessary ornamentation such as “three-deifying” bar graphs or chart elements, or using the moire effect to create the illusion of motion. (Eye catchers is another way of describing them)

* The data itself should draw the eye in, and ideally use the data itself as the representation of chart data where possible. (The Minard maps of Napoleon en route to Moscow, and the Ayres tabular graph of units deployed to Europe in WWI are primary examples in the book; on the AV:T thrust chart, I use thrust arrows for displacement effects.)

* If using graphs or graph elements as a play aid, remove as many boundary lines as is possible, and for the ones that remain, use a lighter shade of gray than your data representation objects. Use shades of gray to represent levels of intensity rather than differentiation in color, as the human eye reads from light to dark as a hierarchical phenomena, but does not do so for colors.

Beyond layout guidelines, it coaxes the game designer into seeing games, particularly spatial games, as multivariable representations of data, with some of the variables set aside for user input.

Wargames can be visualized, in the terminology of Tufte’s dissertation, as graphical representations of small multiples — a series of graphics representing the interactions of several variables, indexed over changes in another variable. While his example of the hourly distribution of hydrocarbon emissions over the LA basin, the same principle applies to the movement of troops over terrain from a starting point to an ending point.

The primary difference is that the indexing variables are put into the hands of the players of the game. Representation of one statistical result in absolute fidelity is less important in a wargame than presentation of all the possibilities inherent in the data set, and letting the players explore the potentials combinatorially. Where ambiguity is a flaw in any sort of
statistical graphical display, ambiguity (within bounds) is the essence of game design. In fact, the ideal of the wargame not only puts different variables in the hands of each player, but also restricts the communication of data regarding which variables were
selected between opponents. Columbia Games’ block games are an excellent example of this.

Ambiguity in user-controlled variables is vastly different from loss of fidelity in constants used to measure effects. Random elements and differentiation of intention and misdirection to gain an asymmetrical statistical advantage over an opponent is desirable. Saying that a trireme is a good model for a space ship is not.

As the principles of hagiographic and statistical maps apply to the clear visualization of statistical data, they also apply to the presentation of probalistic displays in CRTs and logarithmic scales for maps. To wit, the aim is to present the most information in the least amount of space with the least amount of ink and clutter as is possible.

Tufte rightly cites the Minard Figurative Maps (the ones representing 6 data axes covering Napoleon’s Moscow campaign, and Hannibal’s Transalpine Campaign) as bellweathers of excellence in statical display. They use cartographic features, chronological events and
line width and color to represent troop strengths and positions over time, as well as a second graph which highlights temperature in the case of the Moscow campaigns.

What Tufte appears to have overlooked, due to the focus of his academic specialty, is that the Minard figurative maps have their linear descendants in modern publishing with the wargames industry. One can even make an argument that wargames present more complexity more readily than the Minard map, because above and beyond encouraging comparison of data elements and sets, they encourage experimentation with the data sets.

The Minard maps, when placed in a military-historical context,  are abundantly clear progenitors to the scale maps and statistical models of warfare which both the French and Prussians used (and mis-used) in the 1870s as part of the Franco-Prussian war. The data analysis tools in the map can be directly traced to the methods of reconing miles of territory taken in casualty densities used by the war planners of World War I. Perhaps even more germane is the use of very similar maps (Messier’s maps of the later phases of the Franco Prussion War) to give immediate and viseral representations of the logistical needs of troops in the field.

in essence, the Minard maps, after loosening some variables to interpretation and “what if” experiementation, became the basis of wargaming as we know it.

(And for the record, if I ever write a historical wargame based on the Moscow campaign, I intend to use the Minard map as my box cover…)

The principle of a good statistical graphic, and the principle of a good wargame are the same. It is not complication of the simple which is the aim of the designer, or even the simplification of the complex (which results in a lossy data compression scheme), but the
clear presentation of multiple variables over a spatial-temporal context, using words, numbers, graphs, tables and cartographic representations. Furthermore, it MUST tell a narrative with the data, and provide avenues to analyze the data not only as it historically occured, but as it might have been, through manipulation of carefully controlled user inputs.

A wargame which complicates the simple tends to be unpublishable. There are examples however, which prove this point, and the point that complexity is in the eye of the beholder. Wargames that oversimplify the complex are much more common, and result in wargames which either arbitrarily prevent a player from attempting a strategy or tactic, or which channel themselves into one strategy so neatly that they become, for all intents and purposes, statistical displays with counters and fiddly bits.  There’s not enough ambiguity to make the game fun any longer.

The creation of a wargame is to give a visual access to the subtle and the difficult, with the added burden of putting complexities and indexing variables in the hands of the player.

In short, the art of wargame design is the revelation of the complex.

I recommend this book highly. My primary complaint is that many of the examples of
poor graphics could use a layman’s discussion of the subjects covered
by them, to make it easier to see where the error lies. It’s one that I pull down from the shelf and read about once a year.  He’s written three additional volumes, Envisioning Information, Visual Explanations, Images and Quantites, Evidence and Narrative, and Beautiful Evidence.  If there’s strong responses to this review, I’ll review the other three in future posts.