RogerBW's Blog

I've been playing quite a lot of Rallyman GT lately, and I've been thinking about how racing games should be scored.

For a single race with a nominally even start, finishing order is the obvious way to go. But how do you combine these scores when dealing with a race series?

The usual system seems to be something like what Formula 1 does in the real world: a set number of points for the winner, and an exponential decay of points thereafter (so the gap from 1st to 2nd place is greater than the gap from 2nd to 3rd; in other words, over a race series a 1st plus a 3rd place finish is worth more than two 2nds). This is the canonical approach in Rallyman GT's Championship supplement, and what BoardGameArena does; it's easy to add up a driver's points from a series of races to determine an overall winner.

(How do you resolve ties? The nature of the game is such that an individual race can never be a tie, but it's entirely possible for one driver's places to be 1,2,3 while another's are 2,3,1, giving them the same number of points. I tend to tie-break on the most recent race result.)

A different point-per-place formula might be linear (6 points winner, 5 points 2nd, 4 points 3rd, etc.), which would encourage steady play rather than going for the win, or even with inverted intervals, such as 25, 23, 21, 18, 15, 8, which makes fighting for the last few places more important than the first.

But starts aren't even. One way round that is to interweave the racers, perhaps using something like the line-ending formula of a sestina, so that everyone gets a start at the front of the grid and one at the back, and doesn't always end up next to the same other racers.

1 2 3 4 5 6 - start position
A B C D E F - race 1
F A E B D C
C F D A B E
E C B F A D
D E A C F B
B D F E C A

If you are evening out starts in this way, you could score people on places gained or lost rather than on finishing order: if you got from sixth place to second, that's worth more than starting in first and simply staying there. (Some people have argued that Rallyman GT has a runaway leader problem; I don't see it myself, but here's a possible answer to it.)

Or you could take the finishing order of one race and use it as the starting order of the next, effectively concatenating them. In a real race, you want the fastest drivers at the front to reduce dangerous conflicts; in the game, those conflicts are some of the fun. One group I'm involved with uses the current point standings, reversed, as the starting order, so whoever's doing best so far will have to fight through the entire pack to get another win.

There also seem to be competing league models: Americans tend to start from the assumption that there'll be a number of nominally-equal individual groups having their own competition, each of which puts its winner forward to a final race. (Which runs into the stack ranking problem: you want to be in a group full of bad racers to maximise your chance of being the best in it.) The European model has ranked divisions and no final: the winner of the top division is the overall winner. (Which means that if you didn't start in the top division, you can't win; and you need to seed the divisions in some way to get broadly equal skill categories, as well as finding some way to add new players in later seasons.) There are probably other models we haven't tried yet.

Rallyman GT also generates a time for your race, based on the gear in which you ended your turn – but while adding up the race times works in rallying, here I think it would pull the racer in a different direction from actually wanting to win the race, because the two are disjoint enough that you might come first in the race but half-way down the pack in total time. (Once Rallyman Dirt is available, that'll be another matter; in rally-type racing, and in that game, the total time matters far more than the finishing order.) Flamme Rouge can potentially do this better, giving a time effectively tied to finishing order, and one of the things on which a rider is scored at the end of a series is the total time they've taken.

There probably isn't a single best system for this, but it's fun to explore the possibilities.