Analysis of the Results of Civ VI GotM01

Landos

Chieftain
Joined
Aug 26, 2012
Messages
17
Location
Minnesota, USA
Hi Everyone. Congratulations on some insane turns. I also liked watching the 6otM thread get turned into an even crazier finish in the fastest science victory thread. I didn't get to play because of a newborn in the family, but I like the game and wanted to take a look at all of the agglomerated data to see if there was any strategic information to learn about how to play this game. I thought about posting this in the Strategy thread, but since it uses data from 6otM01, I thought that might be inappropriate.

Victory Speed vs # Cities
The first thing I did was to comb through all of the posts on the results topic, and pull out the turns to victory and the number of cities. I think the consensus is that expansion (Wide) is vastly more important in Civ6 compared to Civ5. If that's true, then broadly speaking & with lots of caveats there should be a trend correlating the number of cities and a faster victory. The data is actually more interesting:
Spoiler :
Turns per city.jpg


The user reported data has a huge variance, with almost no linear trend to speak of. Presumably this is due to sub-optimal play, but it's the first 6otM and everyone's still learning. Despite the poor clustering, there does seem to be an initial downward trend (faster victory) which then rises a bit. This is more obvious is one performs a bit of smoothing on the data. If we hypothesize that there is some # of cities which will yield a minimum victory time, we can fit this data to a parabola (pictured above) as an estimate. This would suggest that based on user play, a city count of ~26 final cities would yield the fastest victory. Obviously the winners of this challenge far surpassed the average play. Although it is interesting that not only did the winners tie on total turns, but they also had the same number of cities...


#of Early Cities
The second thing I looked at was early expansion. We all know that to win, we have to snowball an early advantage into a later blowout. So I went back and looked at the Opening Actions thread, and picked out the number of cities for each poster at turn 100. I then plotted their total victory time vs the number of starting cities, as shown below:
Spoiler :
Cities 100.jpg
Cities 100 trend.jpg

Here there is a very clear trend. More cities early yielded players a faster win time. The linear fit to this data was not too bad. One can make a silly extrapolation and ask the following question. According to this data, how many cities would someone need to have at turn 100 to win by turn 100? The trend yields 27 cities, which is obviously A) Pretty difficult by turn 100 and B) Making settlers instead of festivals clearly has diminishing returns. I do, however, find it interesting that the trend lines from both data sets points to similar "optimal" numbers. I also think it's interesting that our top 2 players differed significantly in their city count at turn 100. EliteTroops had 9 cities, and Trojan13 had 17 cities. I wonder if Elitetroops didn't go on a conquering spree right around turn 100, or turned on colonization and pumped a bunch of settlers right away.


First District Choice
Finally, I decided to look at whether a players choice of initial district (after all, earlier bonuses are better than late ones) significantly impacted their win time. Again, this data was taken from the Opening Actions post, and correlated with total victory time. There were only a few people who posted that information, so this data is quite limited. The result is below:
Spoiler :
District Choice.jpg


If you pick a campus, a holy site, a commercial hub, it doesn't seem to make (on average) a massive difference. Obviously, more data would be useful here, but I think this suggests that if you pick an early district to get a short term gain, that may be okay, and the long term consequences of that first district choice might be lower than I would have guessed.

Going through this data was fun for me. If you think it might be more appropriate in a different section, let me know or feel free to move it. Hopefully someone else at least finds it interesting.
Cheers,
 
Some nice stats. I've heard some people condemn holy districts as worthless, but I've found them to be quite useful.
 
Indeed. In my own play when going for an SV I wouldn't have tried for a holy site. I was surprised that one of the winners got a holy site first. I didn't check, but presumed that they went for jesuit education, but idk.
 
I also think it's interesting that our top 2 players differed significantly in their city count at turn 100. EliteTroops had 9 cities, and Trojan13 had 17 cities. I wonder if Elitetroops didn't go on a conquering spree right around turn 100, or turned on colonization and pumped a bunch of settlers right away.
I was in the middle of a war at turn 100, also had a settler about to settle in a couple of turns if I recall correctly. I think I reported my final city count wrong in the final actions thread, ended the game with 21 cities, not 19. Overall I was quite late with expansion in that game. Instead I had focused too much on early campuses, which cost a lot in the end. I haven't played in a while, and not at all with the latest patch, so I can't say what would be some kind of optimal amount of cities with the new rules.
 
What were the starting build orders overall? I've been doing scout > slinger > builder > settler, but it would be interesting to track what the first item or first four or five items were per a chart like above.
 
What were the starting build orders overall? I've been doing scout > slinger > builder > settler, but it would be interesting to track what the first item or first four or five items were per a chart like above.
Given that the more early cities you have the earlier you win, then I'd guess they probably shooted for military first. Probably a combination of Slingers, Warriors and Builders instead of Scouts.

Personally I've been mostly going for a 3x Slingers > Builder > Settler/Warrior/Slinger opening, so that I don't delay Archery for long. Though I might start with an earlier Builder and delay a bit Archery.
 
Actually, in that game I opened slinger->settler. Then I got a free builder from a hut and both the first cities built settlers at Early Empire before I built any more builders or went to war. That was just an experimental opening, mainly because all potential war targets were quite far away and on king difficulty they are very slow to build settlers and found a second city. On deity early war is much stronger as every AI will provide you with at least 3 cities. If they haven't founded all three yet, you get free settlers instead. I think Trojan did a more aggressive opening and went to war much earlier than I did, which shows in his city count by turn 100.
 
Given that the more early cities you have the earlier you win, then I'd guess they probably shooted for military first. Probably a combination of Slingers, Warriors and Builders instead of Scouts.

Personally I've been mostly going for a 3x Slingers > Builder > Settler/Warrior/Slinger opening, so that I don't delay Archery for long. Though I might start with an earlier Builder and delay a bit Archery.
3 Slingers (+ rush Archery) > Builder > Settler is what I've been doing too on Deity. If you get tough neighbours it's required; if not it's still useful for a decent start. Scouts are absolutely useless, which is a shame.
 
@elitetroops clarifies nicely the fact that starting terrain has a big outcome on what you do.
The stars are great and good job, but they are limited in number and including ones that show victory turns but not number of cities skews figures. Not intentional and it provides sone idea, I am just saying one needs to be very careful with those graphs.

Did GOTM3 have better figures to work with?
 
GOTM 3 was prince - it does really not matter what u do at this lvl...
Just make sure u follow win condition at some point.
I started with builder into 3 slinger into holy site.
I m just a yields fan and early builder boosts yields .. - its pretty easy imo

btw discussing GOTM1 is pretty dumb as it was before patch 1 where all u should have been doing was choping all world for 1 turn settlers into overflow or even worse abuse the sell bug
 
Hi Everyone,
Sorry for the delays, I was getting in some holiday travel.
@elitetroops : Sorry I miscounted the total number of cities. It was actually surprisingly difficult to capture the correct numbers from many people's posts, and I'm sure I made more than 1 error.

@Victoria & @MyOtherName I completely agree that any conclusions here need to be taken with a grain of salt. The data is a statistical analysis of self reported data, but I'm open to the criticism that drawing patterns from it is a fool's errand. As for 6otM03, the opening actions thread has many fewer posts, and so analyzing build order becomes even sketchier. Based on the forums, it sounds like 6otM04 may have fewer people competing, so if people want to go back on 6otM05 and tell what their build orders were, I'd be happy to analyze it for any potential parallels with finishing time.
 
Hi Everyone. Congratulations on some insane turns. I also liked watching the 6otM thread get turned into an even crazier finish in the fastest science victory thread. I didn't get to play because of a newborn in the family, but I like the game and wanted to take a look at all of the agglomerated data to see if there was any strategic information to learn about how to play this game. I thought about posting this in the Strategy thread, but since it uses data from 6otM01, I thought that might be inappropriate.

Victory Speed vs # Cities
The first thing I did was to comb through all of the posts on the results topic, and pull out the turns to victory and the number of cities. I think the consensus is that expansion (Wide) is vastly more important in Civ6 compared to Civ5. If that's true, then broadly speaking & with lots of caveats there should be a trend correlating the number of cities and a faster victory. The data is actually more interesting:


The user reported data has a huge variance, with almost no linear trend to speak of. Presumably this is due to sub-optimal play, but it's the first 6otM and everyone's still learning. Despite the poor clustering, there does seem to be an initial downward trend (faster victory) which then rises a bit. This is more obvious is one performs a bit of smoothing on the data. If we hypothesize that there is some # of cities which will yield a minimum victory time, we can fit this data to a parabola (pictured above) as an estimate. This would suggest that based on user play, a city count of ~26 final cities would yield the fastest victory. Obviously the winners of this challenge far surpassed the average play. Although it is interesting that not only did the winners tie on total turns, but they also had the same number of cities...


#of Early Cities
The second thing I looked at was early expansion. We all know that to win, we have to snowball an early advantage into a later blowout. So I went back and looked at the Opening Actions thread, and picked out the number of cities for each poster at turn 100. I then plotted their total victory time vs the number of starting cities, as shown below:

Here there is a very clear trend. More cities early yielded players a faster win time. The linear fit to this data was not too bad. One can make a silly extrapolation and ask the following question. According to this data, how many cities would someone need to have at turn 100 to win by turn 100? The trend yields 27 cities, which is obviously A) Pretty difficult by turn 100 and B) Making settlers instead of festivals clearly has diminishing returns. I do, however, find it interesting that the trend lines from both data sets points to similar "optimal" numbers. I also think it's interesting that our top 2 players differed significantly in their city count at turn 100. EliteTroops had 9 cities, and Trojan13 had 17 cities. I wonder if Elitetroops didn't go on a conquering spree right around turn 100, or turned on colonization and pumped a bunch of settlers right away.


First District Choice
Finally, I decided to look at whether a players choice of initial district (after all, earlier bonuses are better than late ones) significantly impacted their win time. Again, this data was taken from the Opening Actions post, and correlated with total victory time. There were only a few people who posted that information, so this data is quite limited. The result is below:


If you pick a campus, a holy site, a commercial hub, it doesn't seem to make (on average) a massive difference. Obviously, more data would be useful here, but I think this suggests that if you pick an early district to get a short term gain, that may be okay, and the long term consequences of that first district choice might be lower than I would have guessed.

Going through this data was fun for me. If you think it might be more appropriate in a different section, let me know or feel free to move it. Hopefully someone else at least finds it interesting.
Cheers,

As a statistician, I commend analysis of the data. Unfortunately, the most important impacts in speed to finish are 3 factors that aren't very strategic:

1. Hut outcomes. Some huts really are far better than others.

2. Stealing settlers from the AI. Free cities are good, and the AI is (or was) terrible at protecting them.

3. Conquering the AI. The AI is completely inept, so conquering all the AIs early will boost speed dramatically.
 
@A+ombbomb you are right correct. If we had a larger sample size, I expect that the random factors would wash out. Quick, everyone get 4-5000 of your closest friends to play the same map in the name of statistical game theory
 
Back
Top Bottom