Huge World High Score Metric

[Svar]

I'm trying to develop a huge world high score metric to evaluate which of several worlds would potentially yield the highest score based on the state of the game and the condition of the known world at 10 AD.

As background, I developed one for my fastest finish 20 K culture games and used it to pick my last submission. It worked great and I finished much earlier than my goal for the highest score in that catagory (1585 AD for Regent Standard world). The metric is much easier for a fastest finish game since it is based on culture score in 1 city and by 10 AD you know what culture improvements are available for the rest of the game.

I'm playing a huge Pangea, warm, wet, 60% water world as the Mayans on Warlord level with barbarians set to restless and 8 AI opponents set to least aggressive. So far I have searched over 11000 maps and selected 12 maps that have been played to 1000 BC with an additional 29 maps that will potentially be played to the same date. I can generate good maps faster than I can play to 1000 BC (about 1 hour). I'm wondering whether it is important to get a great start on average size domonation limit or a good start on a high domination limit.

Many of the maps still to be started have 3 cows in the starting position and my best start so far was from one of these. I'm still learning how to build multiple settler factories early. I have decided to eliminate worlds that have lots of jungle, mountains or marsh close to the starting position as it seems to inhibit early growth.

Also does anyone know of a utility that will count the various types of terrain tiles for any known world. It will really help in setting up my metric.

 

[Takeo]

Mapstat can give you the terrain counts. under File- click on
'view terrain counts'. i usually will look at this to see if the grassland count is high.

 

[Svar]

Originally posted by Takeo
Mapstat can give you the terrain counts. under File- click on
'view terrain counts'. i usually will look at this to see if the grassland count is high.

Thanks for the input. I'm playing C3C version 1.15 and get strange results with Mapstat, CRpMapStat doesn't seem to have the same option.

 

[superslug]

Hmmm, I wonder if we can sweet talk Dianthus one more time. I'll post a request in the CRpMapStat thread...

 

[superslug]

I posted a request in Dianthus' thread. He responded wondering if that would be legal. I asked in the rules thread and still awaiting a response from Aeson.

In general regards to finding 'good worlds', the factors can get really dynamic. Obviously, grassland is god, but jungle/marsh and some forest are grassland waiting to happen. If you're playing an Ag civ, desert is practically plains. Also, if you're an Ag civ, you could do as Aeson once suggested and pick nonAg AI deliberately to set the map conditions worse and rely on your extra food to help you out over them.

 

[Svar]

Originally posted by superslug
I posted a request in Dianthus' thread. He responded wondering if that would be legal. I asked in the rules thread and still awaiting a response from Aeson.

In general regards to finding 'good worlds', the factors can get really dynamic. Obviously, grassland is god, but jungle/marsh and some forest are grassland waiting to happen. If you're playing an Ag civ, desert is practically plains. Also, if you're an Ag civ, you could do as Aeson once suggested and pick nonAg AI deliberately to set the map conditions worse and rely on your extra food to help you out over them.

Thanks for pointing me to Dianthus' thread, I think he can give me what I need.

After searching 12500 maps I have decided enough is enough. I have played 20 that have been saved at 1000 BC and still have 27 to start. Not all 27 will be saved but over half of them probably will. The minium number of cows in the starting position for these 27 worlds is 2 but 6 have 3 cows and 1 has 4 cows. More than 2 cows doesn't guarantee success though, 2 previous 3 cow starts were abandoned because the start locations were on a narrow peninsulas.

Even 2 cow starts can have great potential. My last saved start was a 2 cow start but there are about 10 cows within a 10 tile radius of the first city. I already have 2 4-turn settler factories running and should have 2 more before 500 BC. The best location has 3 cows very close together and only 6 tiles from the start city but I only found it 5 turns before the 1000 BC save because it is near the coast and my initial move is inland to find a suitable location for the future palace that can build a great core of cities. The 1000 BC save isn't as good as some of the other saves but after playing to 10 AD and comparing it against other good starts that get played to 10 AD I can use this tool I'm trying to develop.

I will probably only play 10 of the 1000 BC saves to 10 AD.

 

[Svar]

Dianthus has just updated his CRpMapStat utility as stated here .

 

[Svar]

I'm still working on getting enough games to 10 AD to develop my evaluation tool but I now know what to look for in a starting position and the 1000 BC save. I want a minimum of 2 cows with a river start that is relatively open (minimum number of hills/mountains and no jungle or marsh) for the start.

For the 1000 BC save I want at least 18 cities 2 of which are 4-turn settler factories with the potential of two more well before 500 BC. I want relatively open terrain for most of the current cultural territory and the same for the potential 10 AD save. Having lots of forest is OK because I can chop it in 3 turns. Rivers are nice but not manditory if enough fresh water is avalailable or most of the terrain is grass after chopping any forest.

I played some 1000 BC saves last night to 10 AD and came to this conclusion when comparing the very first one with the last one. The very first one had most of the above criteria but had only 17 cities, 16 workers, 17 military, 4 cows and 1 wheat inside the cultural territory at the 1000 BC save. It also only had 1 luxury plus iron and horses were a long way away. It only had 1 settler factory (first city) but the future capital was 2 turns away from building the Pyramids. After a sneak attack from a distant Civ (1 city lost), a war with a close neighbor (1 city gained) and finally getting iron and a second luxury hooked up I saved at 10 AD. The 10 AD save totals were 72 cities, 76 workers, 60 military, 5 cows and 4 wheat inside the cultural borders.

The last game played had a much better 1000 BC save position but didn't achieve anywhere near the same totals at 10 AD primarily because of the number of hills around the future capital. My goal for the 10 AD save had been 60 cities and about the same number of workers and military based on analysis of the potential for a game that hasn't been played past the 1000 BC save but meets all the criteria previously mentioned. It looks like a well played game with the above criteria has the potential to get 80 cities by 10 AD.

 

[Svar]

It's strange that once I decided what was needed for a good 1000 BC save, the very next world that was started meets all the criteria. It started as a 2 cow river start and had more jungle close than I would have liked but I played it anyway. As I explored north and west I discovered that the jungle was only on the south and we are on the northeast corner of the Pangea continent. More cows, wheat, and game showed up as I continued to explore. By the 1000 BC save I had 19 cities (4 settler factories), 19 workers, and 24 military. The 24 military were a mistake, it was lunchtime and I was playing too fast. About 4 of the warriors were meant to be workers but I hit the wrong button over and over again. The real suprise was the food rich tiles. As of 1000 BC I had 6 cows, 2 game, and 2 wheat inside the cultural borders. Within my immediate grasp (10 AD) are an additional 7 cows and 3 more wheat.

This game will played to 10 AD but first I will play the remaining 19 starting positions (5 with 3 cows and the 4 cow monster).

 

[Svar]

I'm down to my last 5 starting games and wanted to try to play one of the 1000 BC saves. I picked one that required me to move the capital quite a distance but had a SGL with which I could rush the Pyramids whenever I wanted. Here is the result.



I counted 74 cities from the F1 screen. This map was a challange to expand in because the starting city was down by the two settler factories and I only built 3 instead of the 4 that is my goal. I was really using this map for practice because I have at least two more that have SGLs to rush the Pyramids. The original starting position was really confined and I had to expand straight north to get out of the jungle but the new palace is just on the other side of all that jungle so my cities cultural borders had an hourglass shape for most of the time. All this transit that was required really slowed down the movement of settlers from the south to their new city locations north of the jungle.

With this test I think I will manage at least some maps that have over 80 cities by the 10 AD save.

 

[superslug]

Your setting of goals and methodical elimination of maps should make the most of Mapfinder! I salute your discipline and patience. I'm sure it'll pay off!

 

[Svar]

I knew that 80 cities by 10 AD was possible but when it rains it pours. I played one of my 1000 BC saves that had a SGL to build the Pyramids. Since I wanted to practice with an early GA (950BC) I picked one that had no iron in sight. I figured that if I messed it up it was no great loss.

By the 10 AD save I had found some iron and managed to finally build 6 MI, I also had 35 horsemen because they were the best offensive unit I could build after my requisite 20 Javelin Throwers. I also have 6 Ancient Cavalry because I hooked up ivory as my second luxury and built the Statue of Zeus after the Pyramids.

Toward the end of this session I was sure I could get 80 cities but when I counted I had 91. I also have 8 settlers, some on the way to far off luxuries. To improve the land there are 120 workers. I have iron and horses hooked up along with 5 luxuries. I'm about 11 turns from hooking up the 6th luxury.

There are 3 more 1000 BC saves with SGLs that I will definitly play then all I need to do is figure out which of the remaining 16 1000 BC saves to play for my final evaluation.

 

[Svar]

In between playing promising games from 1000 BC to 10 AD I work on metrics for my evaluation tool. I came up with one that might suprise people. I looked at the top 5 games in the Monarch level except for fret's which I couldn't download for some reason. I used Dianthus's CRpMapStat utility to analyze the last turns to see what the terrain types were.

Interestingly enough if all you do is look at the total grassland tiles on the final turn there is an almost exact correlation to amount of grassland and ranking. EMan had 2534 and is first, Darkness had 2416 and is second, fret had 2355 and is third, Drazek had 2373 and is fourth, finally superslug had 2232 and is fifth.

Since everybody looks for high domination limit I tried to incorporate that into the equation by just adding it. EMan had a total of 6143, Darkness had 6101, fret had 5,930, so far so good, Drazek had 6207, oops, finally supesslug had 6178 another oops.

Domination limit has to figure into the equation somehow so I decided to use 1/2 value of domination limit for the calculation. EMan had 4338.5, Darkness had 4258.5, fret had 4142.5, Drazek had 4290, and superslug had 4205. Not perfect but much better, player skill has to enter into it. In fact when you look at my sixth place entry you get 2495 for grassland and 4357 for my new map metric. I told everybody that the map I played on was a great map it was just my playing skill that kept me down.


Edited to include fret's game which he so expeditiously posted. Thank you fret. Now the problem is that fret's game forced me to take a closer look at the terrain so I will do another evaluation and post an update to this.

 

[fret]

Originally posted by Svar
I looked at the top 5 games in the Monarch level except for fret's which I couldn't download for some reason.

Thats weird!

I got a great start in that Monarch game, probably my best ever. Had about 80 cities at 10ad IIRC. The dom limit was only 3500 ish though.

Here ya go ...

 

[Svar]

OK, I got the Darkness and Drazek games switched, they both played the Ottomans.

Now if you look at the total grassland tiles on the final turn there is an almost exact correlation to amount of grassland and ranking. EMan had 2534 and is first, Darkness had 2373 and is second, fret had 2355 and is third, Drazek had 2416 and is fourth, finally superslug had 2232 and is fifth.

Since everybody looks for high domination limit I tried to incorporate that into the equation by just adding it. EMan had a total of 6143, Darkness had 6207, fret had 5,930, Drazek had 6101, finally supesslug had 6178. That data is kind of scattered.

Domination limit has to figure into the equation somehow so I decided to use 1/2 value of domination limit for the calculation. EMan had 4338.5, Darkness had 4290, fret had 4142.5, Drazek had 4258.5, and superslug had 4205.

When fret posted his game he mentioned he had a great start. I don't have the 10 AD saves from EMan or Drazek but the 10 AD saves from both Darkness and fret look impressive. Darkness had a 10 AD save that included 79 cities and 786 tiles inside his borders while fret had 81 cities and 801 tiles inside his borders. I think this is why fret was able to outscore Drazek even though his map metric that I'm now using was 4142.5 versus 4258.5. Other than fret's great start the metric works.

 

[Aeson]

I had lost fret's Monarch save somehow. I've uploaded it again and the link is working now.

 

[Drazek]

Originally posted by Svar
[When fret posted his game he mentioned he had a great start. I don't have the 10 AD saves from EMan or Drazek but the 10 AD saves from both Darkness and fret look impressive. [...] I think this is why fret was able to outscore Drazek even though his map metric that I'm now using was 4142.5 versus 4258.5.

My nearest save is 90 CE, I had about 40 cities. It was my first milking attempt and I didn't even have settler flood going! This makes it interesting. Fret had double number of cities but the final score was only 158 points higher. We could perhaps check how fast the conquest phase was done, but seems like it's essential to have a good map.

 

[fret]

Conquest phase was finished in 1060AD on my game.

At 920AD it was 18 tiles from the domination limit. Not sure exactly when I hit it, thats the nearest save I have.

Also, it might be worth mentioning that 10 cities at 10AD had come from beating the crap out of the Persians (always a pleasure ), the Egyptians and the Chinese. Having a couple of neighbours close by making early expansion by conquest easier might make a differance.

Could the number of luxuries hooked up by 10AD make a differance as well- I had 5 at 10AD.

 

[Svar]

The number of luxuries hooked up is definitly an factor in my evaluation tool. I currently have 3 10 AD saves with 5 hooked up and it makes expansion much easier. I had to abandon a game last night because I only had 1 hooked up by 500 BC and my cities couldn't keep their people happy. I looked around and hooking up 2 more luxuries was possible by 10 AD but would take too much work.

The three main metrics right now are fast start, the map metric and number of luxuries hooked up with the potential of more within about 20 turns. There will be others but they will play a minor role.

I have 7 10 AD saves now and the new map metric has me excited about 4 1000 BC saves and 6 4000 BC starts. The 4000 BC starts can be played in about 1 1/2 hours each but the 1000 BC saves take 3 to 4 hours to get to 10 AD.

 

[Svar]

I have 9 games successfully played out to 10 AD and am refining my high score metric evaluation tool. It took much longer to finish my 9th game because even though I had a SGL and triggered my GA in 950 BC for the slingshot effect I get, there were so many mountains so close to me that I really didn't want to play it. Because of this world I have modified my map metric but I needed this game completed to 10 AD as a check for my evaluation tool.

My new map metric is the number of potential grassland tiles plus domination limit minus the marginal food tiles (mountains, hills and tundra). This last game had a new map metric of 5393 with a domination limit of 3826 (my highest) but I have several games that have new map metrics of over 5500.

Even though there were mountains blocking my expansion in this last game and I didn't want to play past 10 AD, I still did my best for a good start. I must be getting better with this phase of the warlord games because I ended with 85 cities (my second best to date), 2 settlers, 157 (my best to date) workers, and 44 military units. My quick start metric gives points for cities, settlers, and workers along with the number of techs needed to ented the IA. Because of this game I decided to subtract the maintainance cost of military units from the number of workers as I saw a down side to the number of workers this early in the game.

As it now stands the last world ranks third in my high score metric evaluation tool so I think I'm ready to finally finish 3 very high grassland starts plus about 4 1000 BC saves. It doesn't sound like I'm making much progress but I did play 3 of the 1000 BC saves and started 2 of the promising 4000 BC starts. Both of the promising starts yielded 3 settler factories from a single cow river start and of the two 1000 BC saves played to 10 AD saves currently rank 1st and 3rd.

If it weren't for the map metric I wouldn't have even played the 1st ranked game because even though it had 4 settler factories at 1000 BC there wasn't a logical location for the new capital where I normally locate the Pyramids. In playing this game I changed location for the Pyramids so it isn't complete yet and at 10 AD it is 7 turns away from triggering my GA. My new capital location will be triggered after my after the end of the GA but I will have to move it again because I'm sure the area around the new capital will be abandoned during the milking phase.