Earth challenge (huge size)

[Matrix]

Excellent work, gram123! [goodjob] Very nice! I use both Excel and SPSS. One comment I'd like to make is that I work with the squared scores rather than the scores themselves. Otherwise the high scores weigh in too much. By the way, you put down the wrong list for the civilizations.

But it's very interesting. I stick to the fact though that we shouldn't separate scores by victory type. It's like compensating the score of a football match by the system the teams use (you know, 4-4-2; 3-4-3; 4-2-3-1...). I always go for a domination victory if possible, but sometimes I just find that a space victory is available earlier than domination victory, so while fighting the space ship reaches Alpha Centauri before I obtain the domination victory. If people go for a space or cultural victory by choice, then they admit defeat towards those that do obtain a domination victory (at the same difficulty level with the same civ). Or they just don't know about the score correlation that gram123 has proven so elegantly here, which can be seen as a lack of skills. (After all, knowledge is power.)

SPSS simply calculates the arithmetic mean per civilization/difficulty/whatever. Combined with the variance an error margin can be calculated too and thus how significantly different the averages are between each other (by doing a one-way ANOVA test).

 

[WithTea]

Excellent work, gram123! [goodjob] Very nice!

Have to agree with that statement. Some nice extra info there gram123.

Just an FYI: The game itself has a bias on victory. I don't know every combination but I do know from first hand experience that even finishing on the same turn the game chooses domination over conquest and cultural (I found that out yesterday. Managed to pull a threefor.). Cultural over space and diplomatic. It's pretty straight forward that it will choose anything over time. I suspect but don't have first hand knowledge that it will go domination over space, diplomatic and religious.

 

[gram123]

Excellent work, gram123! [goodjob] Very nice! I use both Excel and SPSS. One comment I'd like to make is that I work with the squared scores rather than the scores themselves. Otherwise the high scores weigh in too much. By the way, you put down the wrong list for the civilizations.

Woups you are right here are the civs score list

Civs	Average	To index
India	406799	0,303936566
Zululand 278436	0,444055294
Rome	190898	0,647680758
Vikings	176099	0,702109932
Mali	168926	0,73192241
Mongolia 151916	0,813875326
China	115853	1,067219298
Egypt	81190	1,522845001
Maya   	70081	1,764259798
Inca   	69115	1,788902808
Khmer 	58070	2,129161033
Native 	48186	2,565913147
Aztec	37840	3,26747069
Russia 	39467 	3,13

Interesting that the system think as Aztec as the hardest start - i wouldn't have thought so. Also noticed that pretty much all the 'hard' start are in america. I think this is because of the same bias towards domination - it simply takes more time to get a domination victory when you have to invade another continent rather than just conquer the entire euroasia.... (notice that this list isn't weighted after victories - though it might change the picture quiet a bit - maybe i should try to make a list that do take this into account. I think this may give us a list more reflecting of the actual difficulty of the starting locations. )

Edit: ops noticed a miscalculation about Russia. Forgot to take the average of those just added the Russian win together. So actually Russia seems to be the hardest start - Quite the oppesite of what most players predicted before we started this. I think this might be because of too few players have played Russia.

And by the way - whats the reason for squaring the scores. My system multiply a factor to the initial scores - why should it makes any different if the scores is squared ?

 

[gram123]

But it's very interesting. I stick to the fact though that we shouldn't separate scores by victory type. It's like compensating the score of a football match by the system the teams use (you know, 4-4-2; 3-4-3; 4-2-3-1...). I always go for a domination victory if possible, but sometimes I just find that a space victory is available earlier than domination victory, so while fighting the space ship reaches Alpha Centauri before I obtain the domination victory. If people go for a space or cultural victory by choice, then they admit defeat towards those that do obtain a domination victory (at the same difficulty level with the same civ). Or they just don't know about the score correlation that gram123 has proven so elegantly here, which can be seen as a lack of skills. (After all, knowledge is power.)

I still dont agree with this point of view. I think the score system do not refect that domination is almost always the easiest win. And the system as it is dosen't give the other wins a change to compare it scores, because the finish date could never compare to whats possible for a domination/conquest win. But thats just one thing

The very nature of a domination win is set to give the highest score, because the score is calculated by population and land area - which is exactly what you are trying to maximize for a domination victory. Therefore a domination win and the same date as a conquest victory will return a higher score for a domination victory simply because you choose to get the (maybe crappy) city that you conquered. Also the score system is crashed because it dosn't apply to the most effektive win for some of the other victory types. IE: if you want a high score for a diplo victory - you should actually just do it the same way as a domination victory - then just before you reach the domination land limit build the UN and vote your self - this will return a higher score than actually do a diplo win - where you get the AI to vote for you. Simply because you got more land and pop.
Same goes for culture and space. You get rewarded for having a lot of city with at lot of pop, eventhough the most effective cultural win might not have pop of 60% of the world. So if you want to maximize your cultural win - spam cities not culture

 

[Matrix]

Well, I think you're right about everything there, except that you turn everything around. The goal in this challenge is to get the highest score as possible. You get that by getting as much land and population as possible. Subsequently a domination victory is easiest if you are successful in gaining that. If you are not, all that is left for you is a space race or cultural victory.

I think victory type is unimportant, despite the high correlation. But a domination victory is not guarantee for a high score, a high score game simply results in a domination victory.

 

[Matrix]

Well, I think you're right about everything there, except that you turn everything around. The goal in this challenge is to get the highest score as possible. You get that by getting as much land and population as possible. Subsequently a domination victory is easiest if you are successful in gaining that. If you are not, all that is left for you is a space race or cultural victory.

I think victory type is unimportant, despite the high correlation. But a domination victory is not guarantee for a high score, a high score game simply results in a domination victory.

 

[gram123]

Well, I think you're right about everything there, except that you turn everything around. The goal in this challenge is to get the highest score as possible. You get that by getting as much land and population as possible. Subsequently a domination victory is easiest if you are successful in gaining that. If you are not, all that is left for you is a space race or cultural victory.

I think victory type is unimportant, despite the high correlation. But a domination victory is not guarantee for a high score, a high score game simply results in a domination victory.

Hmm but then its not like comparing football systems as you speak of - because they are all trying to get the same victory. Then its like giving a speed iceskater scores on a 0-6 icedancing scoresystem. Obviously the speed skater has not done as good as the icedancer because he was not trying to make a perfectly timed pirouette.

 

[gram123]

Here are the scores weighted for wintype. The big difference is that the weigthed consider Mali easier, Egypt easier. Native and Aztec a lot easier. Vikings alot harder. Mmongolias as easier. Both agree that the easiest is India, Rome and Zulu, While Russia and Khmer seems hardest.

[B][I]Weigthed for Win[/I][/B]
India	0,473187472
Rome	0,796159933
Mali	0,810085202
Zululand0,876850537
Inca    0,897709437
Egypt	0,918002129
Native 	0,926574297
Aztec  	1,012121212
Vikings	1,093088695
China	1,114675425
Mongolia 1,267092057
Maya	1,708873893
Khmer	1,834093966
Russia	2,267237216


[B][I]Not Weithed for Win[/I][/B]
India	0,303936566
Zululand 0,444055294
Rome	0,647680758
Vikings	0,702109932
Mali	0,73192241
Mongolia 0,813875326
China	1,067219298
Egypt	1,522845001
Maya	1,764259798
Inca	1,788902808
Khmer	2,129161033
Native	2,565913147
Russia	3,132744994
Aztec	3,26747069

 

[illram]

As discussed earlier in this thread (and other places i suspect) it seems that the scores in civ4 is biased towards domination victories. Now that we do have a lot of data on the same map, i thought i would be interesting to try and control for victorytype and see how the scores compares then. Im not quite as sophisticated as Matrix so that i use SPSS for this kind of stuff (eventhough i actually did use SPSS for school purpose not so long ago)

Basicly this is a excel sheet, where i have found the average scores of the different victory types and civilizations this compiles into a index where 1.0 would be the average. In this way both civilizations and victorytypes gets a so called difficult factor which then is multiplied to the scores of the different players.

    [U]Player            [/U]  [U]Civ  [/U]  [U]Points[/U]  [U]Date   [/U]  [U]Result     [/U]  [U]Difficulty[/U] [U]Score[/U]
1	Kadazzle	Egypt	142225	1300	Cultural	Deity	599782
2	Gustave5436	Inca	94879	1718	Space Race	Monarch	547944
3	WithTea		Native	64025	1855	Cultural	Monarch	454939
4	Archon_Wing	Aztec	37840	1903	Space Race	Monarch	399155
5	gram123		Inca	52151	1778	Cultural	Emperor	258352
6	gram123		Egypt	57305	1752	Cultural	Immortal 241663
7	Rom_Addy	Rome	564671	730	Domination win	Noble	234913
8	Archon_Wing 	Native	32347	1883	Cultural	Noble	229846
9	QuantumPion 	Egypt	45361	1900	Space Race	Monarch	223006
10	illram		Egypt	192424	1712	Domination win	Monarch	188220
11	balatian	Khmer	137299	1810	Domination win	Emperor	187770
12	zbgayumn	Inca	37220	1768	Cultural	Monarch	184385
13	Bjeurn		Khmer	25768	1942	Space Race	Prince	177120
14	evil_spock	Inca	146644	1847	Domination win	Monarch	168500
15	gram123		Egypt	34704	1265	Cultural	Settler	146351
16	Brenador	Inca	29537	1768	Cultural	Noble	146324
17	m00n		Maya	119019	1878	Domination win	Emperor	134874
18	west45		Mali	264564	1748	Domination win	Emperor	124379
19	Elbow_Jobertski	Maya	21143	1932	Space Race	Warlord	120423
20	Archon_Wing	China	60638	1818	Diplomatic	Monarch	119488
21	DesertPfox      China	171069	1814	Domination win	Emperor	117267
22	illram		Rome	55083	1851	Space Race	Monarch	115175
23	dilettante	Russia	33855	1923	Space Race	Warlord	114132
24	monkybone	Mongolia213852	1710	Domination win	Emperor	111795
25	Elbow_Jobertski	Inca	22367	1808	Cultural	Warlord	110804
26	DigitalCraft	Egypt	22499	1867	Space Race	Warlord	110611
27	zbgayumn	Egypt	110919	1738	Domination win	Warlord	108495
28	grenobles	Vikings	231560	1774	Domination win	Emperor	104428
29	gram123		Mali	73289	1857	Diplomatic	Noble	99044
30	AbsoluteZero	Zululand292290	1605	Domination win	Immortal 83368
31	monkybone	Vikings	180154	1816	Domination win	Emperor	81245
32	Kadazzle	India	406799	1655	Domination win	Deity	79417
33	gram123		Rome 	174347	1555	Domination win	Noble	72531
34	CaF		Zululand237385	1605	Domination win	Emperor	67708
35	dilettante	Khmer	11144	1936	Cultural	Noble	65707
36	alhimik		Rome	152288	1824	Domination win	Monarch	63354
37	Bugg123		Zululand428130	1315	Conquest	Immortal 54903
38	Morholt		Vikings	116584	1864	Domination win	Monarch	52577
39	stevoh		Mongolia 89981	1887	Domination	Monarch	47039
40	Makahlua	Russia	66967	1890	Domination	Noble	44917
41	Ororo           Zululand 155940	1814	Domination win	Emperor	44478
42	topsecret	Egypt	8260	1963	Space Race	Chieftain 40608
43	dcmort93	Rome	8102	1968	Cultural	Chieftain 14532
44	Mellus		Russia	17580	1982	Domination	Noble	11792

Here are the scores of the different victory types. It seems there are too little data to create a fair score fore Conquest and Diplo. Both at which only a few scores were made and all are fairly high scores, therefore they are punished as the system think as these victories as easy to get victory. Witch might or might not be true.

Spoiler :

[u]Win Type      [/u] [u]Average [/u] [u] Index (1) [/u] 
Domination 	192492	0,642317439
Conquest	428130	0,288793336
Cultural	44647	2,769247058
Space		38298	3,228339304
Diplomatic	66963	1,846395289

And here are the scores of the different civs. I think this might be the same formula Matrix is using when he weight the civs scores. So here are the numbers on a index to 1

Spoiler :

[u]Dif[/u] 	[u]Average[/u] 	[u]To Index[/u] 
Deity	274512	0,450403228
Immortal360210	0,343247247
Emperor	176299	0,701313567
Prince	25768	4,798241653
Monarch	88762	1,392949283
Noble	121235	1,019844401
Warlord	42156	2,932899971
Chieftain 8181	15,11320021
Settler	34704	3,562733141

Note that Prince is getting way to high scores because only one player played this difficulty and compared to other players he scores lower than the average. Noble is getting relatively low scores because Rome_addy proved that it is possible to make a very high score on this level. And finally settler is getting a fairly low scores compared to the other low difficulties - thats because the only game played on this level made a better score the other low difficulty games.

I also tried to test for difficulty but the result didn't make sense. It appears that because people mostly play on a difficulty that matches there skills, the result will not be comparable. That is: Bacause all the good players play on higher dif, and getting higher scores, therefore the system think as higher dif as easier to get high scores. The result being the lower the dif. Eventhough i actually think this might be right, because it seems easier to get higher scores on higher dif. IE. I played two games with Egypt, both cultural victories and eventhough i finished 700years earlier on settler i still get higher scores in my immoral game. So this system could actually be used to level out the bias the game has for higher dif, but that would require that both good and bad players played on both high and low difficulties.

Spoiler :

[u]Dif[/u] 	[u]Average[/u] 	[u]To Index[/u] 
Deity	274512	0,450403228
Immortal360210	0,343247247
Emperor	176299	0,701313567
Prince	25768	4,798241653
Monarch	88762	1,392949283
Noble	121235	1,019844401
Warlord	42156	2,932899971
Chieftain 8181	15,11320021
Settler	34704	3,562733141

Interesting work! Thanks for doing that. Only correction (and this may have been my fault) my domination win was with China and not Egypt; I submitted the wrong civ on matrix's website, oops.

 

[Tirinn]

Would love to post ss but I forgot to take them

Anyway:

Civilization: Aztec

Difficulty: Settler(My first full game of civ4)

Victory: Space Race

Finish Date: 1907

Final Score: 7833

Edit: Ty Olaf

 

[Optical]

Normalised.

 

[dalamb]

I notice you didn't turn on No Events; much as I like events, you should probably have disabled them if you want comparability between different peoples' games. Also, I've seen arguments that you did exactly the right thing by disabling tech trading, since apparently that makes games highly variable too. I don't know enough to have my own opinion (though I suppose ignorance doesn't have to stop me from having an opinion).

How important was Choose Religions to you? I really like seeing the normal ones so I can judge tech progress a bit, but maybe you wanted to eliminate that?

Does it make a big difference that the Americas are reachable pre-Astronomy via both Siberia and Iceland? ISTM that with a couple of Nav I galleys (or triremes to be safer) the Vikings could get a serious amount of commerce from intercontinental trade, long before anyone else could.

Is there somebody reading this thread who knows a good statistician or experiment designer? I haven't done any since the early 1980's. ISTR you don't necessarily need huge amounts of data to get a decent Analysis of Variance -- that is, percentage of the differences in score attributable to each different variable. I suspect the player would be the biggest contributor to variance, rather than the civ (that is, 50% or more of the differences in scores might be attributable to player, with lesser percentages for things like choice of civ). I can't guess how much data you'd need to distinguish effect of player from effect of difficulty.

 

[dalamb]

It seems to me the Viking start location is a little weak; did most people who played them SIP or did you move? I made 2 dotmaps, one SIP, one involving settling the capital on turn 2 (spoiler warning: resource locations for copper etc shown via signposts):

Spoiler SIP :

Spoiler move to river :

In either case I think I'd send the 2nd settler and the warrior around the Baltic to the maroon location near the lake; it could be a good GP farm or commerce city (maybe both, at different times in the game?).

 

[Yoshiegg737]

I gave up after a game with the Khmer. I was doing well until Shaka suddenly attacked me and captured all my cities except my island ones and razed them. So I just quit.

Further proving my inability to handle sudden attacks

 

[Archon_Wing]

Yea you really have to watch out, since Khmer's start seems pretty lousy, and Shaka becomes very strong on this map-- Ramsees is probably one of the worst AI leaders out there, and Mansa is Mansa...

 

[Yoshiegg737]

I got a few Great Spies from the start that allowed me to see the insides of China, India, and Mongolia. Cathy was friendly, I pushed her into wars with Rome to distract her...

I never even bothered with Africa, thinking "they aren't going to send an army halfway across the world..."

 

[m00n]

Yeah I tried another game as the Khmer after my first one ended in a situation where China had no other land target than me and backstabbed me three times takeing away in the back end what I captured at the front. So this time I killed China first then India. We'll see.

It's like compensating the score of a football match by the system the teams use (you know, 4-4-2; 3-4-3; 4-2-3-1...). I always go for a domination victory if possible, but sometimes I just find that a space victory is available earlier than domination victory, so while fighting the space ship reaches Alpha Centauri before I obtain the domination victory. If people go for a space or cultural victory by choice, then they admit defeat towards those that do obtain a domination victory (at the same difficulty level with the same civ).

If I combine the two examples, you're saying playing football with a 4-4-2 system instead of 3-4-3 is admitting defeat towards those that are able to play 4-4-2. I think it's plain to see that this is not the case. Because in football, the system you play does not determine whether you can only win 1:0 or anything between 1:0 and 15:0. In Civ 4 it does. And that's the problem.

 

[Florian K]

If I combine the two examples, you're saying playing football with a 4-4-2 system instead of 3-4-3 is admitting defeat towards those that are able to play 4-4-2. I think it's plain to see that this is not the case. Because in football, the system you play does not determine whether you can only win 1:0 or anything between 1:0 and 15:0. In Civ 4 it does. And that's the problem.

I agree with Matrix that the scores should not be ranked by victory type. Since we are using football analogies:

If you are playing against Andorra or San Marino (currently tied for last in the FIFA rankings) in World Cup qualifying you are almost certainly going to win (much as in Civilization, as almost nobody finishes, let alone even posts their losses). The question is by how much you are going to win (ie.your score). If you play a 3-4-3 formation (ie. go for a domination victory) you are much much more likely to win 10-0 (or some similar score) than if you play a 1-5-4 formation (ie. go for another victory type). There is a chance you will win by a greater margin with the 1-5-4 formation, just as some cultural or space race victories score better than some domination victories, but if you want to maximize your chances of getting a high score you would play a 3-4-3 formation (ie go for a domination victory).

Woups you are right here are the civs score list

Civs	Average	To index
India	406799	0,303936566
Zululand 278436	0,444055294
Rome	190898	0,647680758
Vikings	176099	0,702109932
Mali	168926	0,73192241
Mongolia 151916	0,813875326
China	115853	1,067219298
Egypt	81190	1,522845001
Maya   	70081	1,764259798
Inca   	69115	1,788902808
Khmer 	58070	2,129161033
Native 	48186	2,565913147
Aztec	37840	3,26747069
Russia 	39467 	3,13

Interesting that the system think as Aztec as the hardest start - i wouldn't have thought so. Also noticed that pretty much all the 'hard' start are in america. I think this is because of the same bias towards domination - it simply takes more time to get a domination victory when you have to invade another continent rather than just conquer the entire euroasia.... (notice that this list isn't weighted after victories - though it might change the picture quiet a bit - maybe i should try to make a list that do take this into account. I think this may give us a list more reflecting of the actual difficulty of the starting locations. )

Edit: ops noticed a miscalculation about Russia. Forgot to take the average of those just added the Russian win together. So actually Russia seems to be the hardest start - Quite the oppesite of what most players predicted before we started this. I think this might be because of too few players have played Russia.

And by the way - whats the reason for squaring the scores. My system multiply a factor to the initial scores - why should it makes any different if the scores is squared ?

Interesting stuff. I wonder if Matrix would consider using some combination of his current system and the average scores (or the square roots of average scores) to make the current rankings more accurate. Currently the rankings are completely based on the correlation between scores achieved by the same player playing different civilizations. There has only been one loss reported in the current standings (with the Vikings). Due to this one loss the three other games played with the Vikings have had huge scores boosts resulting in the players of those three games being ranked in the top four. Currently the easiest way to get near the top of the player rankings is to win any game with the Vikings and not to play a game with anyone else. I would think it would be better to encourage people to play multiple games with multiple civs.

 

[Smwat]

Spoiler :

Immortal Khmer
Space Race
113965 (Peaceful Game)

 

[Matrix]

Nice one! What was the date?

In general, I would prefer if people would use the form I created to submit their result, so that I won't overlook a result. The link is also in the first post.