Old GOTM Scoring Discussions

[Yndy]

Guys,

I wish I could contribute to your discussion but my hard disk totally crashed Saturaday morning. I couldn't even format it.

Luckily it's during warranty period but now I can't really do anything to help ( except lamenting here).

Cheers,

Keep up the good work

 

[MSGT John Drew]

...and all I could think about was how much my pc would have skyrocketted had I joined in this debate. Damn GOTM took all my posting time!

peace!

 

[nbarclay]

I haven't tried to read through this whole thread, but I have a couple thoughts on the issue of milking and scoring.

The reason I have no interest at all in milking games is that it requires absolutely insane nation management decisions. A well-run civilization would want as many improvements in each city as possible, but milkers have to deliberately limit their building of anything that produces culture if they want the game to last until 2050. Worse, since science improvements produce culture, milking strategies seem likely to seriously compromise a nation's science as well. Further, making cities productive rather than merely happy is almost completely irrelevant to the success of milking. Thus, milking is NOT about building up a great civilization, but rather is merely about manipulating the scoring rules.

If other people get their fun out of manipulating the scoring rules to the greatest possible advantage, that's fine. But for me, Civilization games are about building a great civilization, and that transcends the manipulations of a rules mechanic. I'll bend my play style a little to try to score a little higher, but I won't smash it to a million pieces.

I enjoy the GOTMs as a shared experience, and I do compete fairly seriously (and in one case so far, successfully) for "fastest finish" medals. Speeding up the finish as much as possible does require a little bit of "rules mechanic" manipulation, but it's mostly about building up a nation that is militarily and/or scientifically advanced as quickly as possible. I can live with that, and even when I don't win medals, coming close tells me I'm a good player whatever the score chart says.

Which brings me to the ultimate test of any scoring system: Can players who are simply trying to build great civilizations expect to score at least almost as well as players who deliberately manipulate the scoring system to their advantage? Unless a scoring system can achieve that kind of transparency, it will be at least as much a measure of players' willingness to go out of their way to manipulate the scoring system as it is of their true ability to build great civilizations. I hope those of you who are working on ideas for alternative scoring systems keep that in mind.

Another point: an analysis of any proposed scoring system must take into consideration how the scoring system itself will affect people's playing styles. Even if a scoring system makes 100% perfect sense for balancing past games, it might not make sense when players start adjusting their playing styles to the new scoring criteria. (And self-calibrating systems that look at relative statistics rather than absolute statistics are even more vulnerable to fluctuations in playing style.) That, too, is something that those who work on alternative scoring systems have to watch for.

I'd love to see a scoring system good enough that those of us who focus on the more abstract concept of building a successful civilization can get scores in the same ballpark as what rules mechanics get. But I'm not holding my breath.

One last thing: in my book, it should be next to impossible for a game that lasts until 2050 to be considered a true success. Any nation worth its salt should normally be able to achieve some sort of victory long before then, and choosing not to do so diminishes that nation. That's not to say that success should be measured solely or even primarily in terms of fastest finish, because a well-balanced nation will almost inevitably not reach any particular goal as quickly as it could if it compromised everything else in favor of that one goal. But rewarding nations for deliberately refusing to capture that last foreign city after all but conquering the world, or for refusing to claim the little bit of extra land needed for domination, or for deliberately destroying their own centers of culture in order to avoid cultural victory, represents a gross perversion of the game's underlying concept and a serious flaw in the scoring system. So my view, in a well-designed scoring system, winning fast should certainly not be the only consideration, but milking until 2050 should not be a viable option at all.

Nathan

 

[Ribannah]

I've said it before: the only fair scoring system is by year of victory, rated separately for each type of victory.

 

[nbarclay]

Originally posted by Ribannah
I've said it before: the only fair scoring system is by year of victory, rated separately for each type of victory.

I don't buy that. A civ that wins a domination victory a few turns faster by going into a mode that builds nothing but military units in its core and suspending scientific research to get gold to rush temples should not automatically be judged better than one that keeps improving its cities and advancing in science while pursuing its goal of domination. Similarly, a civ that wins the space race a few turns faster because it does not pursue optional tech advances and associated wonders and/or does not take the time to engage in some modern-era conquest to expand its size is not clearly superior overall. A scoring system that takes into account the merits of building a nation that's balanced instead of using every available trick to speed up a particular victory condition could be at least as fair. (Consider, for example, the simple adjustment of counting each modern tech not reuqired for the space race as being equivalent to winning the space race four turns earlier.)

On the other hand, I do suspect that having completely separate winners for each victory condition is probably necessary to avoid comparing apples and oranges. Some maps are conducive to fast space launches but poor for early domination or conquest, and others are the opposite. It would take an essentially impossible level of sophistication in map analysis to determine whether the player who did best while pursuing the primary goal of conquering the world did "better" than the one who did best while pursuing the primary goal of winning the space race. Any scoring system that attempts such a comparison will inevitably rely heavily on approximation and guesswork. But even so, it's hard to imagine a scoring system much less sophisticated than the one Firaxis came up with.

Nathan

 

[Cartouche Bee]

Cracker's, "Improving Your Opening Play Sequences" measures performance against what you actually accomplish. It's really the first attempt by anyone to actually evaluate the good vs. poor decisions on a grander scale and form some guidelines to improve ones play.

Evaluating all the tangible accomplishments in the civ is the real key to a better scoring system that can reward a broader range of play styles.

CB

 

[Aeson]

I tried the formula on the GOTM 10 results now that they are our. Here they are if anyone is interested.

I haven't done much of anything on the calculator or formula since last posting, didn't seem to be enough interest to warrant all the frustration of getting the decompression working right.

 

[Cartouche Bee]

Hi Aeson,

Alain's game was finished in 370BC! So that is why the base score is so high because the sheet has it as 370AD, that cuts 2220 points off the base score.

However, none of the "GOTM Police" ever saw this game so we may all wait along time to see that game.

CB

 

[Aeson]

Thanks for clearing that up CB! 370BC is an even more impressive date of course, and the base score fits much better.

 

[ainwood]

Aeson;

I like the intent of what you are trying to do. What would be useful is a comparison between the score your system would give someone at the point where they are starting to milk on a wholesale basis (eg. within around 90% of domination limit, and other civs suitably depleted), and what it gives them at the end of the milking process. Milking should be marginally higher, I would have thought - for a good milker.

I believe that your intent was to make them comparable - whilst recognising that milking in itself is a skill. Do you have any such comparisons available?

 

[Aeson]

I don't have many games where the saves are there to compare different victory conditions.... If anyone else has some laying around where they felt their effort was pretty consistant up to the domination limit, as well as throughout milking, feel free to share the results (need the tile counts as well).

Looking at my HOF Deity game, I would have scored 12655 through the formula if I had taken Domination at it's earliest available time (23541 score, 1100AD). If I had been playing specifically towards domination, my score probably would have been much lower, and the date a couple hundred years earlier. The formula score wouldn't have been much different most likely though. My score for milking was 10104 (63545 score, 2050AD). The point where I could have first launched would have scored 11202 (27214 score, 1290AD).

It does show an imbalance still, but the scores are obviously much closer than they are through the normal scoring system. There is a bias against milking in the formula mechanics in general, probably because my base numbers for 'best' dates are more like 'very good' dates, as Alain showed with the early conquest in GOTM 10 very well (beating the conquest prediction by 270 years). The 'best' milking scores are more a 'perfect' milking score. I think that out of all the GOTM's, only CB's dogpile influenced games exceeded the max milking score. So scores get judged against perfect, while dates are just being compared to very good.

That said, my HOF game is probably close to the most extreme discrepancy that would be shown through this formula between domination and milking. The map was such that I could expand out to the domination limit almost unfettered (8 civs on a huge map), so my domination date is going to be very good, and my base score as well, because I could focus almost entirely on building up population and territory to that limit. Also, it being a Huge/Deity game, the remaining Civs after I had reached domination were still very strong, and I had to drain my economy (postpone Hospitals and Marketplaces) for a while taking care of them so I could milk in peace. Maybe it shows a need to add a number of civs modifier to the MaxScore/MaxTurnScore modifiers. Because the fewer civs around (relative to map size), the easier it is to build out to the domination limit, and the harder it will be to take care of the remaining civs after that time. Each remaining civ will be bigger and stronger than on a more crowded map.

I don't think we can ever perfectly balance all the victory conditions on all maps. Some maps will just be more condusive to one victory type over another, even with the same settings. Still, 12655 (dom), 11202 (space), and 10104 (milked) are more comparable than 23541 (dom), 27214 (space), and 63545 (milked) wouldn't you say?

 

[EMan]

1. Set the Victory condition each month, just like you set the Civilization.

2. Increase the Early Finish Bonus so the "Trophy Hunters" don't have to Milk the game.

END OF SUBSTANCE!!!!!



"Jawboning" Begins:

WHO WANTS TO MILK, ANYWAY?
As far as I can see, a very small minority actually WANTS to milk a game.......top players, like Cartouche Bee, have admitted they only do it for the "Medal"............that's why I did it too!
BUT, I don't have time to play GOTM if it requires milking every month!

Matrix, I know you're Big on "Square Root + Bonus", although nobody else seems to support this idea.............BUT, since this is "YOUR dictatorship" , any substantial increase in the end-of-game bonus would put an end to milking..............which nobody REALLY wants to do in the first place!

 

[cgannon64]

I think we should combine the Civ3 Tourney scoring system here, and the RBCiv scoring system.

Maybe have a "goal" victory, which could net you more points if you get it.

The problem is that milking takes advantage of Civ's scoring on population, and early conquerers take advantage of the year-of-victory bonus.

I think we should take the RBCiv method, you know, with the "1 point for this, 5 points for that, 20 points for that."

You may noticed that players such as Sullla and Sirian have left GOTM for RBCiv. Maybe that is a sign of something?

CG

 

[Jabah]

Originally posted by Aeson
I tried the formula on the GOTM 10 results now that they are our. Here they are if anyone is interested.

I haven't done much of anything on the calculator or formula since last posting, didn't seem to be enough interest to warrant all the frustration of getting the decompression working right.

Sorry if this is a repeat but the thread is a little too long to check

At least for GOTM 10, where lots of us decide to have a cultural win (OCC for the lowest score challenge), I have the feeling that the 'scoring date' considered 'very good' or 'perfect' was completely unreacheable by far.

I probably have much less information than you about how a game can be won by culture, but 1500 in general seems very, very hard and 1375 on GOTM10 seems, let me think, not easy .

I still like your works anyway.

Jabah

 

[Aeson]

The culture date used is based off of 100k, because I just didn't want to go through all the games and check which was which (and the games weren't posted anyways). 100k isn't too difficult to reach, just requires a dedication to ICS and early temples inbetween settlers for the most part.

I was able to peacefully settle a good bit of land before cutting off my expansion in my game, with about room for 50 cities (if I had gone ICS) by the end of the BC's. That was while dedicating all my resources towards libraries and techs for a spaceship win. I don't think ~1400AD 100k culture would be that difficult on this map. Consider that the tournament game 1-3 (cultural) had less available land to peacefully settle, yet qitai won in 1295AD and JuicyCivNewbie hit in 1375AD. The domination limit on that game was 1477, while the domination limit on GOTM 10 was 1539. So in the room available, it certainly is possible.

The 20k single city culture victory takes until about 1800 to hit, and that date doesn't vary much according to game settings. There are only so many cultural improvements to build, and only so many 1000 year doublings available. I agree that anything much below 1800 is an unrealistic goal. The 20k cultural date for GOTM 10 would have been 1700AD, if you want to see how the OCC wins you know about did, just change the cultural date (in the 'Game' row) to that in the spreadsheet. Then just ignore the new scores for any culture games with a base score over a few hundred (which would be 100k games). Looks like Ribbannah (1740AD, 190 base score = 9053, 6th place) did very well in that regard. The rest of you were probably trying to wait until as late as possible to trigger culture, so as to lower your score, so naturally didn't do as well with the formula.

 

[Jabah]

Yes, you are right, I was more thinking in term of 20k culture victory (OCC or not) which are more 1800/1900's.

I never thought of peaceful ICS early 100k victory, that looks weird but of course should work to trigger an early win (if there is room for such a strategy).

In accordance to this, the cultural victories should probably be separated in 20k and 100k victories... but that is more work

Thanks for sharing all these feedbacks.

Jabah

 

[Aeson]

I'm a little embarrased to show this update...the formula (namely the Emperor/Small/Conquest, and Emperor/Small/Domination modifiers) didn't do so well. I think this has to do with turn/date ratios, and the vast difference that shows up in those two victory conditions depending on if there are galley crossings or not. The modifiers obviously should be a bit lower in those categories, especially the Small (and Tiny) modifiers. The occurance of Galley crossings would be more likely the smaller the map.

I guessed from screenshots that it's a Continents/70%.

I excluded the one game which didn't have a listed date, and I'm just guessing that the 540AD Cultural victory was really a Domination, could have been a Conquest though. 540AD Cultural victory wouldn't have gotten past the GOTM police I think.

There really isn't too much difference between the formula results and the actual GOTM results this month. Spaceship and Cultural games did a little better with the formula, but Domination and Conquest dominated (or rather completely conquered) the top 23. This is still slightly better than the top 39 as in the official results though.

The files in the .zip are an Excel file, and an .html version for those who don't have a .xls viewer.

 

[Kemal]

Aeson,

there seems to be a problem showing the frenchzulu's game in the table. I presume he ended 5th with your formula but there is no game listed at 5th place.

I really like your formula, as once again I ranked higher with this calculation than on the actual result list ( although it's only a 1 place improvement this time).

 

[Aeson]

It shows up for me. I think it has to do with the background color that I used (a dark grey) to signify the official silver medal game. If you are viewing the files with some sort of text color override (IE has one), then the white text I used for that listing will be replaced, and possibly not visible because of the background. To see the results, highlight the area, and the text should be visible then.

Here is the listing for #5 thefrenchzulu:

5 thefrenchzulu 10286 150 AD Conquest Victory 786 1451 12258 13709

There is also a problem with BillChin's game. A score of 1308 shouldn't be possible at a date of 1340AD on Emperor. The bonus for that date would be 3550, which makes his base score -2242. Anyone know the real score/date for BillChin's game?

 

[Kemal]

Yes, you were right, thanks for making that clear (literally ).

I guess the score listed for billchin's game is the unmodified (before bonus calculations) score, perhaps Billchin or someone of the gotm staff can shed some light on this, and how this score ended up in the results list.