What is the metaverse for?

from Stardock Forums
In all the threads here about how to score games, there's been a few posts trying to concentrate on the player's overall score, but they've been swept away most of the time by game-scoring ideas. I think it's worth having a proper discussion on what the goals of the metaverse are - even if a lot of them conflict. Then, we might be able to figure out a new method of eroding scores, and a better player scoring system than

(sum of all degraded game scores) / sqrt (number of games)


Assumption: We have just been given a perfect system to score a single game, inspired by GOD. Everybody is happy with this system - it reflects the player's skill, their speed, and their dedication (i.e. the real-world time spent on it). It cannot be milked. It will not change. No-one questions the game-scoring system in this thread. :)

Okay, here's a few ideas. I didn't make them all up, some are ripped off from other posts, some conflict with each other, and I don't agree with all of them.

- The devs want as many games submitted as possible, that means more data for their machines to churn through and learn from. Even losing games should be submitted, since they let the devs see the AI winning. People shouldn't be punished for submitting these.

- A newbie with natural talent for the game should be able to churn out some tough games and immediately see that he's quite good compared to the average player who's been submitting average games for a while.

- An average player should be able to submit average games and see
some recognition, not just with long-service medals, but in
the score.

- Scores should degrade to some extent, so you can't have someone submit a stack of top games, hibernate, and still be ranked no. 1.

- Someone who's submitted a bunch of games should have to stay active to keep their rank, but not feel disadvantaged compared to newbies.

- The final score is like a chess rating, and you can use it to compare player skill.

- The final score is just an abstract number.

- The top-ranked player is:
a) the person with the most high-difficulty experience of anyone
b) the person who's beaten the toughest scenario in history
c) the person who plays 24/7 and submits normal gigantic games like
clockwork
d) the newbie genius, who "sees" everything about the game, rocks
up, and can post crippling wins (but not maso... yet...)
right off the bat, with no learning games tying him down
e) the person who's showed the most skill in the last day/week/month
f) some combination of the above

Chuck in whatever ideas you want, and we'll see where it goes. If we get a better idea of what the metaverse is for, then we can find a better way to score it for that - but not before.
548 views 5 replies
Reply #1 from Stardock Forums Top
A quick thought. One of the things that would interest me more is seeing different tiers of rankings for all types of players. An example would be to have a 'novice' ranking board, which would include difficulties from Cakewalk to Beginner, then have an 'intermediate' ranking board, which would include difficulties from Normal to Tough, then a 'master' ranking board for difficulties Painful to Masochistic. Each would act as seperate entities and players could move from one to the other as they please, but would be restricted to placement only on one of them. Rewards could be given to top ranking players of each, or simply something that feels like there is a 'progressive' feeling towards submitting. Heh, just an idea to throw out there while I download the O.R.B v1.00-v1.04 58MEG Patch on my twiddly 28.8k connection. Hah.

~SDC~
Reply #2 from Stardock Forums Top
I think that scores should be calculated differently into the final player score depending on this score difficulty etc. Also, crippling/masochistic games scores should depend on speed, and all others like now in 1.05 (population, techs, wonders).

I see that devs don't read 'Lecture 8: Play Balance' written by by Zileas.
1) Modularize, Modularize, MODULARIZE!
End of quote

It's impossible to get intended result using only one formula for everything.

--
d) the newbie genius, who "sees" everything about the game, rocks up, and can post crippling wins (but not maso... yet...) right off the bat, with no learning games tying him down
End of quote

That's me :) Well i played 1st game with intelligent comps (not completed it, i just wanted to see how economy works), but all others was with 5 Incredible AI.

--
Metaverse score will never be a reflection of a skill because AI is a bad opponent even in this game.

~SDC~
Multiplicity - Control multiple PCs using one keyboard and mouse
Reply #3 from Stardock Forums Top
There should also be 'Hall of Fame'
-Longest #1 spot held.
-First game ever played by version and by non gal civ staff
-Most Planets ever conquered, with PQ over x
-Highest PQ planet ever.
-Most useful forum post-er
-Largest Empire Ever.
Reply #4 from Stardock Forums Top
A very interesting thread, Quack. I think there are i) individual gains and ii) group gains to be had. The individual wants to know how 'good' they are, in a relatively objective sense - both in terms of whether they are better than the game's AI, and better than the other humans that might be playing the game. Players know they are better than the AI sometimes, IMHO, by learning the numerical or strategic 'cheats', that is, the tactics that exploit loopholes discovered in the games scoring (eg pop milking in GalCiv), or in figuring out the AI never attacks in such-and-such a situation. They discover they are better than humans by playing against them, but this is not an option for us, unlike chess.

The group option is strong in this game - the Metaverse allows us to band together, add scores and go head to head as tribes. How should we really compare, say, 100 Apolyton members who end up on top, with half a dozen White Councilors who manage the same total score, and just happen to be a small group of advanced players? Do we just say 'tough luck' - if you can get the numbers together, win the group title any way you can? I think the work of Pontius and Gerakken in flying over many threads recruiting members like mad contributed greatly to their top spot, and good luck to them. I'm not for a minute complaining, just trying to tease out the issues. Do we somehow 'weight' the GROUP scores according to number of team members, experience and difficulty levels attempted?

You raised the chess system, and I'd be interested if anyone could expand on how this operates. Computer chess programs can offer you an approximate 'grade' based on your efforts in standard matches against the computer - how is this done? If I rate Staffa as a 'grandmaster', and I see he wins a tiny maso game in 250 turns by cult victory, and gets 500 metaverse points for that - well that gives me a guide - Gal Civ allows me to replicate that (except for his exact starting position and proximity to rich planets). I wonder if the chess rating system has parallels we can use here?
Reply #5 from Stardock Forums Top
Well, the hack they used to take was presenting a bunch of scenarios, and seeing your next move. The computer knows that in that position, newbies have made blunders, average players play average, vets play a good move, and grandmaster can see 10 moves ahead and see a devastating move. It's memorized the moves made by different rated players in the past. Give the player a bunch of those, and get an approximate rating. Nowadays, I think they can do it by giving the game AI a chess rating and rating the player from match results.



For two player games, the ratings are supposed to give an idea of skill, and are modified based on the difference. E.g. if a grandmaster plays a newbie, the grandmaster is expected to win - his score will go up a little bit, the newbie's will go down a bit. However, if the newbie wins, his score will go up a lot, and the gm will get hammered. Even a rare newbie win will keep the scores at around the same level overall.



It's possible to give different game scenarios in GalCiv different fake ratings and take it from there, and that could also be tweaked to have a difference limit. e.g. a newbie rocks up and starts beating easy games again and again. His rating goes up, then slows down, and finally flatlines, since he's rated way too high for that game. He can't bump his rating any more unless he upps the difficulty. The player's actual score could be set at the score of the game just considered too easy to count for him, or the score of the game he's expected to win half the time, or anything else.



The main problem is it punishes losses the same way. I think the devs want all the data they can get, and people won't submit losses to get dragged down. However, low-scoring wins will still bump up newbies, but not vets. If losses aren't counted, then the only way scores can go down is deflation.



About deflation, I'm tossing up deflating the score, vs keeping the same score but deflating how much attention is paid to it. I'll get back to ya :)



re: different ranking boards - they could have a most improved list, and a list of the highest-scoring games ever. Also, people could compare games more if they had extra info, like year won and final population, not just the game score.


[Message Edited]