Hosting

rah · March 5, 2003, 09:26

Yesterday, SK posted some numbers in another thread

Quote:

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.

Quote:

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 · March 5, 2003, 10:11

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 · March 5, 2003, 10:21

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 · March 5, 2003, 10:31

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 · March 5, 2003, 10:34

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 · March 5, 2003, 10:37

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 · March 5, 2003, 10:47

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 · March 5, 2003, 10:48

Quote:

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 · March 5, 2003, 10:50

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 · March 5, 2003, 10:55

Quote:

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 · March 5, 2003, 10:58

Welcome to empirics; little worth testing is ever easy.

atawa · March 5, 2003, 11:09

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 · March 5, 2003, 11:56

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 · March 5, 2003, 12:15

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 · March 5, 2003, 13:24

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

Smash · March 5, 2003, 16:08

It would seem logical that certain color slots would get better terrains.When a game is started,the program starts plunking civs down starting with white,working down the list to purple.We've all seen when a map runs out of room and the purple civ is stuck on a pole.BUt we've all seem purple stuck in a corner which turns out to be paradise for civing.

But

,this game has a very screwy idea of what is good and what is not.So I think it is fairly impossible to test.

I will say this though,whenever I want to start in a river basin,I pick the Zulus.I have no concrete evidence,just experience of 1000s of games.Seem to get rivers often with them but maybe its just my imagination.

War4ever · March 5, 2003, 16:37

ironically i think we all see rivers with certain civs....

i am not sure if there is any way to test this tho

Alexander's Horse · March 5, 2003, 18:14

One thing that affects it is starting techs. If the host gets techs its usually because its settler has been placed on woods or plains i.e. no science/trade. You restart over and over till the host gets no techs. This usually means the host starts on a river, with a special or on grassland near water. In other words some sort of situation which allows beakers from the get go.

I usually host the Aussie games so I've seen this a lot. You never seem to get a tech starting on a river.

DrSpike · March 5, 2003, 18:18

Hmm, that's interesting; even if there is no a priori advantage there could quite well be one selecting for games that get played out: the ones where the host has no starting techs.

Good catch AH.

You should definitely test clean starts only Rah.

rah · March 5, 2003, 18:35

Having hosted as many as I have, I've started most games without a river in site. And it's always seems to be the river basin, special, huts starts, when you get techs and have to restart.

But I will limit myself to testing clean starts.

DrSpike · March 5, 2003, 18:38

Quote:

Originally posted by rah
And it's always seems to be the river basin, special, huts starts, when you get techs and have to restart.

Hehe it's human to remember those more than the others.

Alexander's Horse · March 5, 2003, 18:43

Quote:

Originally posted by rah
Having hosted as many as I have, I've started most games without a river in site. And it's always seems to be the river basin, special, huts starts, when you get techs and have to restart.

But I will limit myself to testing clean starts.

Hmmm, maybe game settings and colour choice also affect it. Its certainly not my exerience.

And hosting has given me some really crap starts more often than not.

rah · March 5, 2003, 18:54

Yeah, spike is right, we do remember the bads ones more. But I truely think (with the amount of games that I've hosted) I'd have noticed an advantage. I don't think so. But that's why we test.

On the surprise bonus, every experienced mp player knew deep in their hearts that one existed despite all the SP players that tested it saying that it didn't exist.

War4ever · March 5, 2003, 19:27

RAH, perhaps we can meet early on saturday and test this......duel settings which would be inline with what strat was talking about, if we do it over icq phone it will go quickly

Graag · March 5, 2003, 19:42

Interesting thread, something I've been thinking about for a long time.

I think a good way to test this would be to provide a form so interested apolyton civvers can note the specifics of their start positiion in each mp game they begin, and simply compile data for a while before trying to test hypotheses or come to conclusions.

it's certainly something that is worth testing, but you're gonna need plenty o' numbers.

Hydey · March 5, 2003, 20:12

Group testing is a good idea Graag.

MalevolentLight · March 5, 2003, 21:22

The real advantage of hosting comes from the fact that you have no lag, so when you're winning and the other guy is losing he'll also be annoyed by the lag hence making him really pissed off. I'd rather be losing without lag than with it.

Ming · March 5, 2003, 21:29

Quote:

Originally posted by MalevolentLight
The real advantage of hosting comes from the fact that you have no lag, so when you're winning and the other guy is losing he'll also be annoyed by the lag hence making him really pissed off. I'd rather be losing without lag than with it.

Good point...

StrategicKingMi · March 5, 2003, 22:20

The color of the civ is irrelevant. All that matters is the host. In 2 recent games with Kuja, we each hosted once....
In game one, I started on the end of a very thin pennisula, no rivers, no specials, he started on huge river system with gold and iron mine in the meat of the cont, he won obviously by 3600 bc . I only had 2 huts near me, he had 10.
In game 2, still going on, I was host, started in awsome, wide open terrign with a silk and in the meat of a cont, he started at the top end of a rough terrign portion of the cont, as much as I have revealed of it so far. As far as huts, I had many near me and he said he had few, or at least did not get citys, which I take as having few.
These 2 games prove out exactly what I said about host advantages, all be it only 2 games.
A host generally starts with extra settlers and techs if the other civ starts with resources in city radius. How many people have had a great start only to look at demographics and see a 2 next to the scie. and have to restart. So, maybe its just not that the host has a great start, but that the non host is almost assured of having a weaker start.

War4ever · March 6, 2003, 01:34

Quote:

Originally posted by StrategicKingMi
The color of the civ is irrelevant. All that matters is the host. In 2 recent games with Kuja, we each hosted once....
In game one, I started on the end of a very thin pennisula, no rivers, no specials, he started on huge river system with gold and iron mine in the meat of the cont, he won obviously by 3600 bc . I only had 2 huts near me, he had 10.
In game 2, still going on, I was host, started in awsome, wide open terrign with a silk and in the meat of a cont, he started at the top end of a rough terrign portion of the cont, as much as I have revealed of it so far. As far as huts, I had many near me and he said he had few, or at least did not get citys, which I take as having few.
These 2 games prove out exactly what I said about host advantages, all be it only 2 games.
A host generally starts with extra settlers and techs if the other civ starts with resources in city radius. How many people have had a great start only to look at demographics and see a 2 next to the scie. and have to restart. So, maybe its just not that the host has a great start, but that the non host is almost assured of having a weaker start.

my only problem with this statement, you can pop a ton of huts, but there is no guarrantee of cities/nomads....

its based on %'s or so they proved in the strategy forums....so in theory, you could pop tons of huts and NOT get cities or nomads