650 AI Game 4UC Stats / Analysis

[saamohod]

In my latest couple of games China is steamrolling hard. By midgame they are top score and their score is 50% higher than the next runner-up.

 

[hokath]

Would you have the stats on the score difference between the winner and 2nd place (at time of win), as a way to measure this?

Excellent idea

 

[Nightmare Dusk]

Just had a thought @L. Vern. While having the yields per era broken down is very nice, not knowing number of cities makes it an incomplete picture. You have any numbers on hand for like average number of cities by era as well?

 

[Nightmare Dusk]

Could i also get the exact numbers for the yields per turn by era graph? Rather then just eyeballing the numbers from the graph. Also, can you please explain what's the cause of MIsc. resources? Thank you for any help you can give me!

 

[Trumtram]

My statistics knowledge is a bit rusty, but 650 games is by NO means anything conclusive and should only be seen as a VERY small indication, or am I mistaken? It's awesome to get a glimpse into how different outcomes are. But the combination of 43 Civs in games of 8 plus the whole randomness of map generation and whatnot... Aren't there like more than 100.000.000 combinations possible by Civ choice alone? Again, rusty knowledge so I could be totally wrong with my calculation. Just want to caution everyone on getting to false conclusions like "Civ A needs a buff" or "Civ B is to strong".

 

[KungCheops]

My statistics knowledge is a bit rusty, but 650 games is by NO means anything conclusive and should only be seen as a VERY small indication, or am I mistaken? It's awesome to get a glimpse into how different outcomes are. But the combination of 43 Civs in games of 8 plus the whole randomness of map generation and whatnot... Aren't there like more than 100.000.000 combinations possible by Civ choice alone? Again, rusty knowledge so I could be totally wrong with my calculation. Just want to caution everyone on getting to false conclusions like "Civ A needs a buff" or "Civ B is to strong".

1) I think the sample size is fairly decent, buy I'm also not a statistician.
2) We don't have anything else, this is the closest we have to a data driven approach to game balance.

My 2 cents

 

[Trumtram]

Please, don't get me wrong, I love this stuff and appreciate it. We just have to interpret this data with caution and shouldn't base any modifications on it TOO strongly.

 

[Cebrudras]

You can do a rough evaluation of these samples sizes by picking what you would consider a problematic winrate and then putting some of the numbers from the first part into a basic binomial calculator.
Example: If Arabia was secretly OP with a 33% winrate, then the data we have, would have been the result of a terrible bad-luck streak with about 1:300 odds.

 

[Trumtram]

You can do a rough evaluation of these samples sizes by picking what you would consider a problematic winrate and then putting some of the numbers from the first part into a basic binomial calculator.
Example: If Arabia was secretly OP with a 33% winrate, then the data we have, would have been the result of a terrible bad-luck streak with about 1:300 odds.

Awesome, thanks for the explanation.

 

[Heinz_Guderian]

I am not a statistician either but I think a statistical analysis would show that there are differences in the category "civilisations" but less likely that you could prove that for example the Celts are significantly stronger than the Shoshone

 

[CppMaster]

We just have to interpret this data with caution and shouldn't base any modifications on it TOO strongly.

Don't worry, we can do small adjustments. If it'd be to much then when can always revert it. It won't have major consequences even if we make a mistake.

 

[civitasjohn]

I hope we can make the information era longer...

 

[Nightmare Dusk]

It needs more buildings first

 

[hokath]

I built a power plant once. It was awesome

Sadly information age is intentionally short because if you get there first it's because you are going to blow every great scientist you've been secreting away to win the game.
If you get there first and you don't win, what are you doing exactly? Probably you already won and you're doing Diplo victory or something because why not at that point.

 

[Nightmare Dusk]

Do people actually hoard great people to use at the end of the game? I either plant or bulb all of them when i get them.
And if people are, is it that much better then using them immediately? If so, i think we should change that cause I'm 99% certain ai doesn't do it. How about the amount a great person gives is based of when it was birthed, rather then the turn is bulb?

 

[Rustonreddit]

Do people actually hoard great people to use at the end of the game? I either plant or bulb all of them when i get them.
And if people are, is it that much better then using them immediately? If so, i think we should change that cause I'm 99% certain ai doesn't do it. How about the amount a great person gives is based of when it was birthed, rather then the turn is bulb?

The amaunt of yields you get is fix on born, so there is generally no reason to not use them immediately. I think he is refering to delaying the born of scientist by not working the slots, or the purchase of great people with faith, or stuff like that