Ranking Road Runners in Serial Competition

Lincolnshire Serial Road Running Championships, a perpetual cycle of twenty-three races held each year - informally called the

Tour de Lincolnshire

Road running is a spectator sport in which the spectators are the competitors. The primary concern is as always the practical proof of one's own developing physical powers, but the runners are so interested in each other's achievements that the complete statement of results is keenly awaited after each race, and the performance of everyone debated. It is mutually respectful and social, with close finishers shaking hands at the end, and many a marriage arranged.

However, there has always been something missing. A haphazard succession of unconnected one-off events doesn't tell you much about rivalry among individuals, and practically nothing about the clubs. We would all like to be systematically ranked, like they do it in other sports, and follow our progress up and down the leagues. Inevitably this requires the aggregation of points, though any attempt to do that on a sufficient scale is apt to be a technical nightmare. You need a thoroughly logically stringent formula for assigning points to each performance that truly reflects its additive worth in the minds of the competitors.

Briefly, the additive value of your performance is its proportion of the winner's speed, obtained by dividing the winner's time by your own. This is a fraction which we express over a denominator of 100,000, so that each winner (male or female) scores 100,000 points. In 2006 the average male scored 66592, and the average female 66640, both nearly two thirds of winner's speed. We use these big numbers to avoid decimal points.

To appreciate the system's simple elegance best go straight to my web-site at http://homepage.ntlworld.com/ronald.hindley/Serial_Competition/lincs.series/ranking%20lists/ where you see how twenty-three representative races of any distances are linked by the points aggregation principle. As each race is run its last year's performances are removed from the database and this year's substituted, so that ranking is over a perpetual cycle, with every stage having the same importance as any other. Individuals can be ranked by several criteria of excellence, depending on how many performances are taken into account - your best single, best two, best three, best four, or all you did.. Everybody who runs but one of the races listed in http://homepage.ntlworld.com/ronald.hindley/Serial_Competition/lincs.series/racelist.htm then has a county rank, from the slimmest and fastest to the fattest and slowest without exception, and can begin to climb the ladder.

Clubs are ranked in the same way. Assuming that a competitive club will produce at least three performances per race the number taken into account is 69. The operator of the program has a menu option to split big clubs vertically, sharing out performances among the divisions, starting at the top and dealing them like cards to notional team captains. The output is club ranks, summary, and shorter summary.

This system is logically stringent upon one premise, that all wins are of equal value. Its only error is in random variation of the quality of the winners, which tends naturally to self-cancellation in the competitive situation. It has been enthusiastically adopted by the Columbia Track Club, Missouri, USA - see http:/ctc.coin/serial/results/NAMESALL.HTM, and I am building a database of twenty-five races in Lancashire. Such databases, all in the same format, can be readily selected, shuffled, merged and filtered ad lib, offering the prospect for the first time of a world-wide structure of competition. See overleaf for an interesting, though less rigorous, treatment of the same information.

Timelists

Before the advent of the system described overleaf in Page 1 it has been customary to use another that is plausible enough provided you are prepared to scrub round a destructive fault and do without some flexibility. As I said, the only error in my ranking system comes from random variation in the quality of the winner, which has a natural tendency to self-cancellation over a long series of races, as runners cannot in practice hope to pursue a clever strategy of avoiding fast winners. I could, incidentally, have used as a standard the average time of all the performances in each race, which some people might prefer, but that would make it impossible for runners to calculate their own scores, and there would be a rounding problem to solve. Not only that, some of the reality of the challenge would be lost.

The basic premise of the Timelist system is not that all winners are equal, but that equally measured courses are of equal difficulty, so that equal times run over them are assigned the same value. This introduces a permanent non-random error due to known big differences among races over the same distance. The premise is plainly not true, as all runners know, and you will notice it when you examine my Timelists closely - one of them in Lancashire is quite shocking. But given the faulty premise they look superficially like honest ranking lists, and are certainly worthy of perusal. They are limited in utility because one can only take one distance at a time, it being impossible to relate performances over different distances as in the real ranking lists. The objection to using them for ranking is that runners naturally look for easy courses, distorting the competition scene to the annoyance of some tourism interests, and this behaviour is re-inforced. There is also the temptation to race organisers to chip bits off the course when laying the start and finish points. Do judges regularly bother to check the marks from the measurer's report?

All the same here they are for Lincolnshire, over the past twelve months, at http://homepage.ntlworld.com/ronald.hindley/Serial_Competition/lincs.series/timelists/ . To make them I sort my file "lincbase" into time order, and apply a little program that churns them out as CSV files, which you can look at with Microsoft Excel. The whole lot flies off the belt in less than two seconds.

For a harder discussion of some aspects you may like to read http://homepage.ntlworld.com/ronald.hindley/Serial_Competition/ranking_system.htm

This file is called advent.htm because it is intended to presage a positive agenda for the world-wide sport whilst aspirant governing bodies jockey for the power to "permit" it to happen. The principle is so strong that its eventual universal adoption is inexorable, but its present weakness is that it takes too much skill to analyse files of results received in many formats and map them into a standard database. With some planning, and the full co-operation of race organisers, it should be possible to reduce that to an easy routine performed by anyone, and it is certainly very much easier now than it was when I started assembling the Tour de Lincolnshire in 2003. Given the database the ranking program that processes it can be run by a child.

Ronald Hindley, 269, Dysart Road, Grantham, Lincolnshire NG31 7LP.
Telephone 01476 402875.
E-mail address ronald.hindley@ntlworld.com
To page one

The paper work is the teamwork.