Discussion in 'Civ4 Strategy Articles' started by VirusMonster, Jan 28, 2008.
o x x x o
x x x x x
x x x x x
x x x x x
o x x x o
It does look like a big, fat, cross, doesn't it?
And I forgot to congratulate you on the game. How long did the game take?
I finished in 2 sessions; I had the PC open and went eating, chatting, doing other stuff probably more than 50% of the time. civ4 timer shows 25 hours..
but probably the learning curve took longer =)
alright, I see the saves are up in the HOF page =) now this game officially has the highest score in the HOF tables.
5 more saves added... some minor edits to the text...
and thank you Thunderfall for including this writeup in the war academy
Congratulations; that's one heck on an achievement, especially with the new BtS AI!
The first EQM congratulates me I am honoured.
Nice work. I was wondering if you learned anything about scoring differences in BTS. I think BTS got a huge score bump, but I'm not sure. Your game is a great achievement and your writeup too. I hate to be the one to rain on your parade, cus I don't know for sure, but 500,000 on BTS might only be like 300,000 on vanilla. Some investigation needs to be done.
I like the rain I will investigate whether the scoring got a bump in BTS, but you are the first one I heard this from. On the contrary, the AI is poprushing units much more effectively in BTS, and inital rushes fail more than they did in Vanilla(well not this game, but in general).
and you are one of the best players in the forum too, so thank you for your kind words as well I enjoyed studying your deity cultural victory games.
edit: I found this thread on scoring discussion between Vanilla/Warlords and BTS. BTS vs. Vanilla/Warlords Discussion
My conclusion of the discussion in this thread is that BTS and Vanilla/Warlords scoring systems are hard to compare, yet probably more or less equal. According to the thread, the math behind scoring in both BTS and Vanilla/warlords seems the same. Base score calculations (land, pop, tech, wonder) seem the same as well.
BTS is harder to score, because of the better AI, defensive AI poprushing, espionage spending, the new tech requirements, slightly lower traderoute profits according to UnconqueredSun (see his post at traderoute profits are less in BTS, and exponential inflation (was linear in Vanilla). On the other hand, BTS has a few more turns added toward the end of the game, so total number of turns in a BTS game is more than a Vanilla game. Consequently, when you do the scoring calculation (current turn/total number of turns), a slightly(by how much?) lower value is returned in BTS. This value, according to the discussion in the thread, seems negligible, but nevertheless increases the final score by a little amount compared to a Vanilla game. Here are some graphs:
date vs. number of turns graphs on different games speeds for Vanilla/Warlords vs. BTS
and the original scoring system explained post:
Civ4 scoring explained
Let's not forget the random events... I think random events could both increase or decrease the final score in BTS, so I am undecided on their effect to the final score. For example, EliteSwords mission could be very strong if you get it while preparing your initial Praetorian rush, but in this game, I don't think any quest or random event affected the final score much, well I got the lucky Greed quest it gave me 6 free unpromoted Praetorians Probably saved 20 turns for producing that many units.
I also feel the city revolt mission could be increasing the speed the cities are captured compared to Vanilla where you had to produce 10-20 cats just so you could reduce the city defences in 4-5 turns. I still use the siege for the colloteral damage, but I wait less time reducing city defenses.
gtg now, take care...
If it matters, I find random events to be a positive effect, making the game easier. (Too small to effect score enough for this discussion though.) The fact is, almost any bad event can be bought off with like 10 gold. There is very little risk.
I looked at that formula too and I think I found the big difference. The currentTurn/Maxturn is the big change. (In other words, how quickly you finish the game) is exponential! BTS moved the max turn out as much as 300 turns. That probably gives the huge bump in score I perceived.
If my calculations are correct, a score of 500,000 on BTS would get roughly 346,000 in vanilla just due to this one factor. I haven't studied the formula that closely, so don't quote me on those numbers....yet.
When it rains, it pours.
My next rough estimate sets 500,000 on BTS to be just over 250,000. It all depends on what turn to end on.
Let me do your exact game, finish on turn 546.
Turn/MaxTurn is .364 BTS and .455 Vanilla
The short answer is that causes your score to almost double the Vanilla score on a game ending around 546 turns. An earlier finish will have less effect. Late game scores would be 4 times greater in BTS.
The increased # of tech and # of wonders does lower your BTS score a bit, so I'd estimate 500,000 BTS is right around 300,000 in vanilla at turn 500-600. Hey, right between my first 2 rough estimates.
The slight timeline changes also work against you, making the BTS score even more inflated...another long story.
edit: Let me add a plug for separating BTS HOF completely. Score is just another reason they are too different to combine in one set of tables. Just like Space Race had to be split into Space Colony. BTS scores will bury the greatest Civ IV vanilla scores. The best games in that category will be forgotten. Same goes for Conquest, but it's Vanilla that buries BTS. I can make a case for culture too. We are not honoring the all the right games in the HOF...my opinion. Treat BTS like CIV 5. You would never try to combine Civ 3 and 4.
Take a look at some examples: (maybe Kamino is the greatest score hound ever, but you be the judge):
1 Kamino 443464 760 AD
2 WastinTime 249260 680 AD
I remember my only goal that game was high score, so its not like I just got domination quickly. I was milkin' it. I even finished earlier than Kamino, but he got 443,000!
1 Kamino 479838 400 AD
2 Bozso77 261266 1286 AD
3 randomlyaware 251583 410 AD
Guess who played BTS?
ok, I have just read your post... you could be right, but I need to dig out the math first, because your rain is washing all my charisma here on this thread jk... I am also curious what my finish date corresponds to in Vanilla, since apart from the highest score, I also seem to have the earliest finish date for Huge Immortal Domination.
I also wondered how Kamino got those high scores... one of his Immortal Tiny submissions almost reached 500k... it was something like 479k... ah you already mentioned that game in your post. I opened up his saves, and my idea is he is killing all enemy units with quechuas and leaves them down to 1 city.. Then he milks the whole map without crossing to domination. Did you apply a more or less similar tactic? Since with BTS 3.13 installed, I can't open your vanilla saves, can you load up your game rivaling the 479K score by Kamino and tell me the land, population, tech, and wonder ratios at final turn?
Also, can you tell me for my future analysis how you know that my game ended at turn 546? Where did you get this number? and where did you get the max turn numbers for vanilla and Bts? I want to open Matlab and plug this scoring formula in for Vanilla and BTS
In the links I brought up, there is probably sufficient information to bring this confusion to an end...
now reading the edit to your last post... I am not sure why space coloy and space race got split btw... Can you enlighten me and other readers of this thread?
I agree with your observations on quests, they only seem to make the game easier except the few negative 10-20 gold ones. On the other hand, the AI gets them too, no?!
And don't underestimate that most Vanilla scores , especialyl those super quick BC large and huge domination games by Incans are all achieved through AI not putting more than 2-3 archers per city :-/ I am not trying to say what you say on the scoring system could not be true, rather I have felt while trying to beat previous scores in the deity tables on BTS( and trust me I tried many times) I usually end up facing a lot more defenders than in Vanilla, so I could not dominate as early or maybe(almost always) fail.... The defending AI poprushes much more effectively and it greatly delays your domination finish date... however, I am not sure whether this is a strong enough reason to balance out finish turn/max turn getting lower in BTS. I started to agree with you that BTS scoring should be thought as seperate than Vanilla scoring...
I also think there is some truth to HOF not honoring the right games, and tomorrow, I will crank out the math Take care for now, it rained where I live today, I love rain
I did the same tactic for a high score as you described.
To get turn 546, I look at your "player log" tab on the game submission.
Total turns show at the top of the screen in-game (if you turn them on in the options.) 1200 vanilla, 1500 BTS for marathon.
Space Race/Colony are split because no one thinks a Space Colony game will ever beat the old Race. It changed too much.
Beating the Immortal BTS AI on Huge was great work. Your writeup should focus on that achievement and give advice towards that goal. Your game doesn't really do much to milk the score, so the title about 500K is misleading. I think 500K will end up being just an average score (for a huge domination). I think someone with a lot of time on their hands will come along with 800K+, probably 1 million.
First, congratulations to VirusMonster for this nice game.
I am happy there is another guy who compete for high score.
It will be a challenge for me to score all the immortal line.
I agree with you on the settings except :
1) Low sea level,
2) Despite the smarter AI, I think Quechua rush is still, and by far, the best strategy. AI are harder to conquer, but easy to cripple. One Q is enough handle an immortal AI. He is the best worker stealer (Q with Woodsman II is a nightmare for AI )
@ WastinTime :
It is obvious things are different in BTS.
You can't compare BTS and Vanilla/Warlors.
Scoring is easier.
Conquer/Domination date is more difficult for big maps on high levels.
AI is smarter.
Sorry for my poor english (I am french)
Oooo.... mr. Kamino arrived as well I love the French... Thank you for your kind words mister.
I am cranking out some numbers to calculate the effective score for BTS and Vanilla at the moment... it will take some time to avoid any mistakes, but I am on track...
nice new tip for quechua rushers btw everyone do woodsman II from now on I also agree with you that low sea levels would probably have been better for higher scores.
As promised, I plugged in the whole scoring formula into Matlab and am confirming that WastinTime is right I think I should make a seperate article for this topic and post in the HOF forum.
Here are my particular findings:
% Matlab code for scoring BTS vs. Vanilla/Warlords
% You can copy paste this code straight into Matlab Student edition, and it will do the math and get the numbers out.
% data from the 500k game
% not sure about initial pop,land, tech, and wonder values so I calculated for 2 possible scenarios
techscore0=(techfactor*techrawscore)/(techinitialscore0 * power(floor(techmaxscore/techinitialscore0),(turnonwin/maxBTSturns)))
% adjustedscore0 = 53515
% adjustedscore1 = 519280 so let's assume this score is almost correct, because it is very close to the actual 502553 in game score...
% now with maxturns=1200 for vanilla and warlords, and the adjustedscore1 becomes: 279740
%Conclusion: WastinTime is right unless the 519280 score randomly appeared so close the the actual 502553 score
Although I am getting a very close number, I can't replicate the exact 502553 ingame score at Matlab, since I am not sure about the initial population, land, tech, and wonder scores.
They could be 0, 0, 2 (for 2 ancient era starting techs), 0 or 1(for capital initial population), 21(for initial land squares), 2(for 2 initial ancient era techs), and 5 (due to palace national wonder). I calculated for both set of values and found that the 1, 21, 2, 5 combination gets a very close score to the in game final score of 502 553.
Another reason I can't replicate the exact in game score could be rounding errors or lack of information about the actual scoring formula. I did check score calculation formulas for both BTS and Vanilla/Warlords in the XML and Python files. There does not seem to be any change except the max turns getting 1500 in BTS.
so in conclusion =) WastinTime is right While BTS games might be somewhat harder to win due to various reasons mentioned in previous posts, BTS scoring system is significantly more advantageous than Vanilla or Warlords.
I must, however , defend that this Huge Immortal domination game ended at turn 546, while the only other submission for Huge Immortal Domination, despite being played on Vanilla huh , ended at turn 673. For large Immortal submission, the best score is 259381, and it still is lower than the normalized score for this game, ie assuming maxturns for BTS is 1200 So as WastinTime suggested, this game is still very good for Huge Immortal Domination, both for Vanilla and BTS. I hope it will stay that way for a long time.
Well done, 5-star writeup!
Thank you bro I have just solved the whole dilemma on spy math and updated the article accordingly. Make sure to read the recently updated Espionage Economy vs. Manual Research section of the article, then go build some spies Updated contents are:
technology cost explained
espionage base cost explained
arithmetic of the espionage mission cost modifiers explained
espionage mission cost modifiers explained
espionage point spending modifier explained
to conversion ratios compared between EE and manual research
most effective way for generating
spy detection analysis with links to game mechanic articles
mission success rate analysis with links to game mechanic articles
spy production and maintenance costs analysis for the average Medieval scenario
pure EE vs. hybrid EE
Nicely done, minor quibble is that beakers actually get multiplied by 20% just from having one prereqs(and 20% more for each additional optional prereq) so beakers are a bit higher than you show in your analysis. Additionaly burrecracy capital in a small empire often carry most of the weight of research(with an academy having 75% base research bonus) giving between (0.5*1.75+1.25*0.5)*1.2=1.74 without other know modifier and 2.262 beaker per commerce in the mid game. Getting those espionage multipliers doesn't always come free, although sometimes it do... Obviously spy spceialists still rock but not as hard as you say, only need a 5+ commerce title to match it if you get favourable conditions... They still by far outpreform scientists by their raw output but in the long run academies are sometimes very good, contradictory to popular belif though you are right in that if you are not going to use the great scientist for some specific military advantage(liberalism slingshot for something, for example) or for academies, you are better off running spies.. Unfortunatly there is no way to run alot of spies early in the game and as such using spies for great people is nowhere near as fast as if you make scientists/merchants/artists as these are way easier to get alot of slots for. This part of your analysis is a bit too optimistic in the light of the EE. The analysis of the cost of spies is rather good, except for the loss of a cottage doesn't actually translate into lost cottage turns but into lost town turns as it will take longer for the cottage to reach town etc. so "Say this citizen was working a cottage with +3 commerce, then the total lost commerce for 12 turns becomes: 12*6*3=216." I am guessing this is a riverside hamlet or a village, but lets say it is village, then you only lose 1 more commerce as the village will still likely grow to town before you get printing press 12*6*4=288 so a minor difference, but the main gist of it still holds true. I belive against the AI EE can still be exploited but it isn't as totaly dominating as you say here, if there were more spy slots available earlier in the game it might have been different but as it is, that isn't the case and as such normal research can often do what you want better when you factor in the delay in getting the techs as well as the AI's not actually researching what you want. In adition due to the rather high risk factor EE have very high variance and as such it could be very good for trying to milking the game as good results from stealing is indeed favourable compared to research. High variance means you will have some great results and some bad reults, with normal research you don't have the same variance and even if normal research would on average outpreform EE, in the lucky case EE will win as normal research don't have the oportunity to get lucky.
I am sorry for the lack of structure in the post but hope it is somewhat understandable. Nice game anyways. I certanly don't have the patice a huge marathon game would need so big props for that.
Separate names with a comma.