Huge World High Score Metric

[Svar]

Just a quick update to illustrate my final map metric for my high score evaluation tool.



This screenshot is the metric with the top three games currently in Warlord with my top candidate for my about to start HOF try at this level (11506 is the map number from MapFinder).

First an explaination of how this works. I add the domination limit to the total grassland potential (grassland+jungle+marsh+.58 forest) plus 1.5 times the floodplains plus .5 times the total plains potential(plains+desert+.36 forest). Finally I subtract the mountains and .5 times the hills and tundra. This is based on an agricultural civ, if you want to use this metric and aren't playing an agricultural civ don't add the desert. Coast tiles are neutral since there is no food surplus.

When you look at the numbers, Drazek's game was played on a Archipelago map and the rest are on Pangeas. The metric for Drazek's game is much lower than it could be because of the unusual high tumber of tundra tiles (320), just look at EMan's tundra tiles (8). An average number should be about 40.

When you look at Darkness's total grassland potential number you again see a very high number but his metric isn't much higher than EMan's because of the low number of potential plains tiles and higher than normal number of mountains and tundra.

If you look at my games numbers the they are almost average except for potential grassland which is about 150 tiles higher than normal.

The only refinement that may be made to this tool is the percentage of forest tiles that have underlying grassland or plains. My database isn't very large yet and I haven't checked these warlord games yet, it's late.

 

[Svar]

I finished my 12th game to 10 AD tonight and have no more good prospects without running MapFinder again. This process screened over 15,000 maps. This is a screenshot of the tool as it now stands.



The first chart is the condition of the maps at the 1000 BC save, the second chart is the condition of the maps at the 10 AD save with final score in column 15, and the third chart is the chart that calculates the map metric. The map metric needs to be imported to the second chart to get the final score.

The high scoring game and the one I will continue is highlited in the second chart, number 11506.

There needs to be plenty of explaination to show how this tool works. First there is more information on the second chart than I actually use to calculate the final score. The final score is the sumation of the map metric, the start metric and the hidden luxury metric which is the number of luxuries hooked up at the 10 AD save multiplied by 200. You can see the weighting of the three metrics as 6500 for map, 1500 for the start, and 1000 for the luxuries for a total of about 9000.

The reason that some luxuries are 1/2 rather than whole numbers is a luxury that can be hooked up in less than 20 turns is counted as 1/2.

The start metric is the sum of 3 times the number of cities, 2 times the number of settlers and the number of workers multiplied by 5. From this number, the number of techs to IA is multiplied by 5 and subtracted. Also the number of workers is reduced by the military unit cost for those games where there is a cost.

The start metric and the luxury metric are subject to further change in the future after I see how this works. I like the map metric.

Now some observations, the game with the best map metric wins but the game with the best start metric and luxury metric is a very close second. The only reason the second place game didn't win is the huge difference in map metrics, 6640 versus 6294. If you look at all the worlds with map metrics over 6500 only two 11506 and 14975 got starts over 1500. The only reason 14975 didn't score higher was the 3 luxuries hooked up by 10 AD. Now look at map number14135, it had the highest potential grassland of all these maps but got by far the worst start and it was the very last game played. You could blame it on the late GA but the winner hasn't had its GA yet. The fastest start also had a very early GA along with three other games with the help of a SGL but 8363 didn't have as early a GA as those 4 games but still outscored two of them with a worse map metric.

In conclusion nothing jumps out on what constitutes a good start other than the number of luxuries hooked up by 10 AD and that might be the weighting I assign to them. The good start conditions will require more study on my part and I have lots of saves and data to study but now I'm going to get some sleep and start on this warlord HOF try tomorrow.

 

[EMan]

I haven't commented on your work Svar because:
A. It's a bit heavy-going for me
B. I haven't had time to study it

BUT, I wrote this post just to say:
You're doing some Great Pioneering Work which will be helpful to all of us HOFers!

Keep up the Good Work

 

[Svar]

Thank you EMan. I'm really just doing it to make me a better player but it helps the thinking process if I report the results on this forum.