Designing a scoring system

[VoiceOfUnreason]

I'm trying to complete an outline for a game proposal, and am having trouble working out how to score it without rewarding players for delaying the victory (aka milking).

So presumably I need a secondary source of score, which provides a pressure against prolonging the game. Once I've got a good option there, it's my responsibility to experiment with the balance of the scoring algorithm until it seems to encourage the kinds of choices I hope will be made. (side question: is it effective to run "regular" games through the scoring algorithm when tuning, or have folks found that the variants are too specific

What kinds of choices have sponsors found to be effective in the past?

I've considered fastest finish bonuses (bonus points for finishing first, or bonus points for how many turns ahead of the player who finished second), but I don't want to encourage players who estimate that they are out of the running to abandon finishing as quickly as they can, so that alone doesn't seem sufficient.

Therefore, I think I want a scoring rule that rewards the turn of finish. So I could do something turn based, or something year based (which has a steeper scale in the earlier years - an interesting possibility).

Since I do want people experimenting within the primary scoring framework, perhaps I could rig things so that the primary scoring framework only applies up to a particular date.

In short, I'm trying to avoid re-inventing the wheel here. Does anybody have suggestions, so that I can shortcut the work of reviewing mumble years of epics trying to figure out if the scoring actually promoted the behavior the sponsors intended.

[Wyatan]

Well, one idea I'd played with was to have the tally date = fastest finish date.
Definitely removes milking as an option, and requires players to aim at achieving the perfect balance between scoring and speed.

Main issue would be technical : since the turn of scoring is unknown until after reporting, scoring might be tricky.
Solutions include :

• Have an "event-driven" scoring system (wonders built, tech X discovered, capital cities captured, etc...) so that simply recording their occurence date enables the score to be computed for any given turn
  
• Have a "static" scoring system : the score for a given turn can be computed entirely from the situation on that turn, without any other factors interfering (total population, number of cities with state religion, etc...). Requires players to autosave every turn and keep them.
  
• Write a mod that computes and logs score
  

[T-hawk]

Voice, I'd be happy to review your proposed scoring system in detail via private message, if you'd like.

Scoring by fastest victory has been commonly used as a simple yet meaningful basis of comparison in our games. However, I'd recommend against it. In Civ 4, so much of the game already depends on building economy. Scoring Civ 4 by fastest finish really does reduce the game to a series of known powerful moves. In Civ 3, building a research economy was much less mutually exclusive with other lines of development (military, cash, culture), so there was actually more room for variety underneath the "fastest finish" umbrella.

Going back to your inital questions... We've long tried to keep milking out of our score systems. As Wyatan says, that essentially means all the scoring goals should be under time constraints. You'd score either based on conditions on specific dates, or on specific events (wonders, etc) that are naturally time-limited.


(side question: is it effective to run "regular" games through the scoring algorithm when tuning, or have folks found that the variants are too specific

Depends on the variant, but I'd say that any scoring variant worth playing will naturally skew game plans enough that a "regular" game scored by that variant won't really have any meaning. Particularly if the variant involves goals that are clearly not pursued in most "regular" game plans, such as culture or permanent alliance.


What kinds of choices have sponsors found to be effective in the past?

Civ 3 Epic 29, which I sponsored, garnered lots of approval for its scoring system. Simple to describe and understand, yet pleasantly intriguing to execute, with a few bonuses to add spice. Epic 25 was similar, and contained an anti-milking provision. If you want a small subset of interestingly scored games to research, also take a look at Civ 3 Epics 3, 14, 22, and 36.

[Raven_Lore]

how about..... Every year over ________<Insert Year>_______ loses ___<Insert point>___ per year.

[TheArchduke]

Civ 3 Epic 29, which I sponsored, garnered lots of approval for its scoring system. Simple to describe and understand, yet pleasantly intriguing to execute, with a few bonuses to add spice. Epic 25 was similar, and contained an anti-milking provision. If you want a small subset of interestingly scored games to research, also take a look at Civ 3 Epics 3, 14, 22, and 36.

Especially the Pax Americana sounds like a lot of fun. I seldom have modern wars in CiV IV and if I do, only as a mopup for domination..

[Compromise]

I just have a couple of suggestions for disincentivizing milking, and I don't think they're unique. (And I don't think that's a word!)

1) As mentioned above, limit the timeframe during which points can be scored or change the scoring for different timeframes (e.g. +1 for each war declared between 1AD and 1000AD, -2 for each war started before 1AD and -5 for each war started after 1000AD)

2) To reward fastest finisher, give points based on the number of other participants whose finish date you beat. This can be scaled many ways, e.g.:

a) fastest gets 10 pts, 2nd fastest 8 pts, etc.

b) 1 point for each person whose time you beat--you don't even know how many points that could be worth until reporting day!

c) percentile: slowest finisher is 1, fastest finisher is N (so N=number of participants); finish bonus is +MaxPoints*(p/N) where p is the player's finish date ranking

d) Scoring penalty of (y-Y)/100 points where y is your year of finish and Y is the year that the fastest finisher finished; that's -1pt for each century you delay your finish

Looking forward to seeing your scenario!

[Swiss Pauli]

I've considered fastest finish bonuses (bonus points for finishing first, or bonus points for how many turns ahead of the player who finished second), but I don't want to encourage players who estimate that they are out of the running to abandon finishing as quickly as they can, so that alone doesn't seem sufficient.

I was going to write about late game rewards, but I think it all depends on how the absolute fastest finish would be achieved. If a military victory is possible, then later finishes (Diplo, Culture, Space) should be rewarded in a way that makes them viable alternative strategies.

[Raven_Lore]

I'm guessing Time Victory is out

[zakalwe]

I would prefer scoring methods that implicitly penalize late finishes over those that explicitly specify complex formulas involving turn numbers, dates etc. Don't know how feasible that is, but here are some random ideas:

- Penalty for each spaceship part built by an opponent.
- Penalty tied to the global population.
- Penalty tied to the maximum culture of an opponent's city.
- Penalty tied to inflation (doesn't that increase over time?)
- Etc.

Of course most of these would work poorly if the player can just conquer everyone else and then milk the game, so they probably need to be combined with anti-conquering penalties to make much sense. Depends on your variant.

[Raven_Lore]

Just my 300 cents but aren't those hard to tell/ take a lot of time to tell?

What is maximum culture?

I think the Penalty for each spaceship part built by an opponent. is pretty good, unless they don't build any