Hosting

[rah]

Yesterday, SK posted some numbers in another thread

Originally posted by StrategicKingMi
Rah, I determine the starting position advantage in the following ways:
1) pennisula or no pennisula- the host almost never starts on a thin pennisula, almost always in the meat of the cont
2) river system- the host has about a 30% chance of starting on the largest river system on the mapand non-host less then 20%
3) nearby huts- host generaly has 2-5 more huts in same radius as non-host
4) centralized location- expansion in all directions ( harder to defend but a big advantage none the less)
less corruption etc.
5) better resources- about 30% of the time host starts with a resource in starting city radius (20% for non-host)
I was even talking about going first.

This was greeted with some skeptism by our favorite MP troll. Eye's called out SK on his numbers. Since we're all familiar his style, I need not post his reponse.

SK fired back.

Originally posted by StrategicKingMi
Second, i did not make up thoose numbers, I have about 100 saved games in my computers vastly bloated memory, they bare this out. I further set up hundreds and hundreds of games and have anyalized the start positions of both players. I examine every game I play and reveal the starts at the end of the game.
And no, it has to do with who hosts. The computer gives the host advantages such as techs and settlers to make up for bad starts. Now, when they have a better start, they have no extras. Most players of course play "clean" starts so it is therefore logical that most starts by the host will be good and play bares this out in reality.

RAH

[rah]

Now that the stage is set, my cusiosity got the better of my, so I thought I'd do a little research on my own. My premise was to prove advantage or no advantage for the host in MP games, in terms of civ placement. I didn't know exactly how to quantify it so I thought I would do some preliminary research to see if I could get my hands around it and see if any general patterns existed that I might then try to test for. The are so many different variables and possible definitions.

I'll say upfront that my primary interests are for games similiar to what I usually play so please don't rag me for how I went about the PRELIMINARY review.
Since I didn't have four players to help and I needed to use the cheat menu, for the sake of time, I tested in SP.
World settings were the same as standard RAH rules with two exceptions.
1. 1x 1x since 2x 1x is not an option in SP
2. I used a map size of 40x50 instead of 39x49

I did 20 starts with 4 civs (taking white every time because it was faster) and charted the coordinates and if there was a river square in the opening screen. I also charted the coordinates and river for the other civs in the color order. (yes I know 20 isn't enough but this was preliminary research)

While I haven't analyized this in detail yet, there was no obvious pattern. WHITE did not mostly get placed in the middle of the map, vs other civs. For the presence of a river in the opening screen was also seemingly random in number of occurances.
1st civ 2nd 3rd 4th
9 8 8 7

In general, I could not see any significant terrain advantage for white (ME). I don't know if MP has a different algorithm for placing civs, (I doubt it) but it needs to be tested.

ANY comments or opinions on what is important. What consititutes better land. We know the software has a different def of what constitutes a good start then we do.

I will continue to analyize my original data and do a few more observations, to come up with some concepts to test. Any help with the definitions would be appreciated.

RAH

[DrSpike]

I thought about this a bit when Strate posted in the other thread. My thoughts at the time were that most of the propositions are an absolute b1tch to test.

If you are determined to try (which I applaud) sample sizes are going to have to be in the 100s.

[Ming]

Keep it up. It will be interesting to see how it comes out.

If you play clean starts... ie, no free techs, I've never really noticed any advantage to being the host in terms of land, huts, or rivers. Since I do host a lot, and I've gotten tons of bad starting positions, I currently doubt that any such advantage exists... but it would be nice to know

[rah]

Spike,
Maybe not. Statistically I agree with you, but we're dealing with reality here. I don't care if there's a 3% improvement. A slight difference is really immaterial in terms of gaming. If there is a dramatic difference, it should be obvious with considerably less. (nothing was obvious so far, so I have to question the conditions that I researched )

WHAT DEFINES A GOOD STARTING POSITION and how can it be quantified? Help!!!

RAH

[rah]

I'll post my coordinate results (I only have them at home since I forgot to e-mail them to work) after I've looked a little more and cleaned them up a bit. I'm trying to figure out the best way to lay out the data to be used by analytic software (SAS)

[Ming]

We know that the AI considers food to be important... Did you notice a higher percentage of food specials for the Host? Granted, in many of the games we play, trade specials are considered more valuable than food specials... but a higher percentage of food specials for the host would indicate that the computer is trying to give the host an advantage...

[DrSpike]

Originally posted by rah

WHAT DEFINES A GOOD STARTING POSITION and how can it be quantified? Help!!!

RAH

Hehe, anyone's guess is probably fine, so just pick something. This is what makes it so hard to test......whatever metric you choose is a very imperfect measure of the underlying nebulous concept. The usual reaction when carrying out such tests is to completely overkill the sample size.

I agree stark differences should be apparent with 20 repititions, but some of Strate's number suggest he thinks that some of the advantages, whilst there, are quite small in magnitude. If this is the case for the particular element you test then your task is unenviable, and sample sizes have to be big to yield meaningful results.

It's a tough one.......if he has equal confidence about his statements then pick ones with the neatest metrics, like your location test using coordinates.

[rah]

Yeah, I considered that, but then the question is how close does it have to be to be considered good. I was concentrating on what you would see on the opening screen, but there are those starts where it looks like crap but 3 squares away there is a perfect location. It's those definition things that I was trying to get a feel for in my preliminary analysis. (which was an obvious failure due to the complexity of the issues)

[rah]

Originally posted by DrSpike
I agree stark differences should be apparent with 20 repititions, but some of Strate's number suggest he thinks that some of the advantages, whilst there, are quite small in magnitude. If this is the case for the particular element you test then your task is unenviable, and sample sizes have to be big to yield meaningful results.

It's a tough one.......if he has equal confidence about his statements then pick ones with the neatest metrics, like your location test using coordinates.

Yep, why can't it be as easy as the sneak attack bonus. We didn't do anywhere near enough observations, but I don't think anyone thinks it doesn't exist anymore.

Maybe when I fire up SAS, the coordinates will be revealing. I may have to go to a larger map size to see it though. And there lies the problem. I don't care what happens on a mega map, I'm more concerned in how it effects the 100's of games that I play.

RAH

[DrSpike]

Welcome to empirics; little worth testing is ever easy.

[atawa]

Rah: You might want to change your research a bit to 2 civs and clean starts.

This was what Strate was talking about.

Good effort though

[rah]

The real effort has yet to begun.

Yes, I should consider CLEAN starts only. Noted, thx.

But as I said, I'm more concerned about our normal games, and if the results turn out to be more generic, fine.

So I guess it will be easiest to test with four people, but when we have 4, we're playing.

This requires more thought. Keep those GREAT suggestions and ideas coming. It will probably take a bit for me to set up again, so there is some time to incorporate more thoughts from others.

RAH

[War4ever]

yeah i think you should test it for duels first, as this was strates claim........

as for your relults......the fact that the fourth civ only got rivers around 35% of the time compared to whites 45% shows a difference right there. Granted its SP mode and only 20 starts....but still it raises a question, is the color you choose more likely to produce certain terrain or closeness of huts

i still think choosing db has some advantages with the right color mix, but of course i can't prove this.

does taking purple in a four man game give me the fourth best spot for placement, or does it still look at the 7th best spot on a map?

[rah]

7 to 9 in 20 obs, means nothing. Your question about purple being 4th or 7th crossed my mind when I was looking for possible patterns, so I'll keep that question in mind.


RAH