FvA Opening Database

Forum rules
This forum is limited to topics relating to the game Diplomacy only. Other posts or topics will be relocated to the correct forum category or deleted. Please be respectful and follow our normal site rules at http://www.webdiplomacy.net/rules.php.

Post a reply

Confirmation code
Enter the code exactly as it appears. All letters are case insensitive.
Smilies
:points: :-D :eyeroll: :neutral: :nmr: :razz: :raging: :-) ;) :( :sick: :o :? 8-) :x :shock: :lol: :cry: :evil: :?: :smirk: :!:
View more smilies
BBCode is ON
[img] is ON
[flash] is OFF
[url] is OFF
Smilies are ON
Topic review
   

If you wish to attach one or more files enter the details below.

Expand view Topic review: FvA Opening Database

Re: FvA Opening Database

by RoganJosh » Sat Nov 07, 2020 3:07 pm

Another aspect there is predictability. An isolated unit can only move, and is usually stuck with only one or two realistic options (read: easy to predict). Units in a cluster that can support/convoy in different ways can give you a dozen of realistic options (read: almost impossible to predict).

Of course, you also have to rush to Scandinavia and Tunis...

I think the Belgian (ENG/Pic/Bur) is the best, but people where playing it too much. Even if it's the best, playing it every time allows your opponent to optimize their response.

Re: FvA Opening Database

by President Eden » Sat Nov 07, 2020 1:21 am

RoganJosh wrote: ↑
Sat Nov 07, 2020 12:59 am
 I don't mind sharing it, but let me check in with our president.
You have my permission
RoganJosh wrote: ↑
Sat Nov 07, 2020 1:11 am
President Eden wrote: ↑
Fri Nov 06, 2020 4:23 pm
 This suggests to me that while the difference in abstract power level between ENG/MAO is essentially zero, moving to ENG synergizes better with moving to Pic/Bur.
One of the informal principles I try to follow, is that you should think of your units like the pawns in a chess game. It's all about the structure. An isolated unit is weak. I think that's why ENG/Pie is weak. The fleet in ENG is definitely isolated. The army in Piedmont can also end up being somewhat isolated, if you're not careful.

The MAO fleet can always get a backup fleet build in Mar.
Definitely agree. The nature of supports and "tied strength" units not dislodging each other means that you really want to concentrate force and keep your units together. I seem to remember the "northern" opening for France (ENG/Pic/Bur) being considered the best opening, since all three units can fight together to win the battle over Germany. I don't know whether it's true that this opening is the best anymore - it's been a couple years since I looked at the format in detail - but I'm sure its strength comes from concentrating everything on a key front.

Re: FvA Opening Database

by RoganJosh » Sat Nov 07, 2020 1:11 am

President Eden wrote: ↑
Fri Nov 06, 2020 4:23 pm
 This suggests to me that while the difference in abstract power level between ENG/MAO is essentially zero, moving to ENG synergizes better with moving to Pic/Bur.
One of the informal principles I try to follow, is that you should think of your units like the pawns in a chess game. It's all about the structure. An isolated unit is weak. I think that's why ENG/Pie is weak. The fleet in ENG is definitely isolated. The army in Piedmont can also end up being somewhat isolated, if you're not careful.

The MAO fleet can always get a backup fleet build in Mar.

Re: FvA Opening Database

by RoganJosh » Sat Nov 07, 2020 12:59 am

Thanks all!

I don't mind sharing it, but let me check in with our president. I don't know how willing they are to just provide dumps of the database either, this was en exercise for me, so I never asked.

Re: FvA Opening Database

by Peregrine Falcon » Fri Nov 06, 2020 11:35 pm

RoganJosh wrote: ↑
Fri Nov 06, 2020 4:20 pm
jasnah wrote: ↑
Fri Nov 06, 2020 3:51 pm
 I think there is a slight error on the Winter 1901 page.
Thanks, fixed it, and added colors.
DougJoe wrote: ↑
Fri Nov 06, 2020 4:06 pm
 Did you get a dump from the DB or something like that?
I was convinced that I should learn some BeautifulSoup...
This is excellent. Thanks for putting it together!

I had actually been thinking of doing some similar analyses, but writing a scraper for webDip was a barrier to actually starting it. Would you be willing to publish the code on GitHub or something?

Re: FvA Opening Database

by teccles » Fri Nov 06, 2020 9:46 pm

This is amazing.

Re: FvA Opening Database

by Squigs44 » Fri Nov 06, 2020 8:13 pm

RoganJosh wrote: ↑
Fri Nov 06, 2020 4:20 pm
jasnah wrote: ↑
Fri Nov 06, 2020 3:51 pm
 I think there is a slight error on the Winter 1901 page.
Thanks, fixed it, and added colors.
DougJoe wrote: ↑
Fri Nov 06, 2020 4:06 pm
 Did you get a dump from the DB or something like that?
I was convinced that I should learn some BeautifulSoup...
Beautiful Soup is a wonderful tool. Tableau as well. Nice work

Re: FvA Opening Database

by DougJoe » Fri Nov 06, 2020 7:31 pm

RoganJosh wrote: ↑
Fri Nov 06, 2020 4:20 pm
DougJoe wrote: ↑
Fri Nov 06, 2020 4:06 pm
 Did you get a dump from the DB or something like that?
I was convinced that I should learn some BeautifulSoup...
I had to lookup what BeautifulSoup is, but I get it now. Pretty cool.

Re: FvA Opening Database

by jasnah » Fri Nov 06, 2020 4:33 pm

President Eden wrote: ↑
Fri Nov 06, 2020 4:23 pm
 Based on taking the highest average score for each moves, the "best openings" appear to be:

Austria:
A Vie -> Tyl
A Bud -> Tri
F Gre -> Alb

France:
A Par -> Pic
A Mar -> Bur
F Bre -> ENG -or- MAO

Interesting note: Despite the difference between ENG/MAO being negligible in average score [~0.001], moving to ENG while also moving to Pic/Bur increases average score by 0.018. This suggests to me that while the difference in abstract power level between ENG/MAO is essentially zero, moving to ENG synergizes better with moving to Pic/Bur. This makes sense to me because moving the fleet north allows it to influence the fight over Germany that those armies will conduct in 1902-04.
While it's certainly interesting to note these stats I wouldn't put very much stock into a bald evaluation of opening strength. Most people I think, consider the openings to be a rock-paper-scissors where Vie-Tyr Bud-Tri Tri-Alb beats Par-Pic Mar-Bur Bre-ENG beats Vie-Tyr Bud-Gal Tri-Ven beats Par-Bur Mar-Pie Bre-MAO beats again the first opening, and so on, with no dominant combination. This is reflected in the stats and the handicap from losing the rock-paper-scissors can be pretty large.

Anecdotally I also would speculate that the Italian (Vie-Tyr Bud-Tri Tri-Alb) is played disproportionately by more experienced players compared to other openings and this is partly related to its historical success.

Re: FvA Opening Database

by President Eden » Fri Nov 06, 2020 4:23 pm

Based on taking the highest average score for each moves, the "best openings" appear to be:

Austria:
A Vie -> Tyl
A Bud -> Tri
F Gre -> Alb

France:
A Par -> Pic
A Mar -> Bur
F Bre -> ENG -or- MAO

Interesting note: Despite the difference between ENG/MAO being negligible in average score [~0.001], moving to ENG while also moving to Pic/Bur increases average score by 0.018. This suggests to me that while the difference in abstract power level between ENG/MAO is essentially zero, moving to ENG synergizes better with moving to Pic/Bur. This makes sense to me because moving the fleet north allows it to influence the fight over Germany that those armies will conduct in 1902-04.

Re: FvA Opening Database

by RoganJosh » Fri Nov 06, 2020 4:20 pm

jasnah wrote: ↑
Fri Nov 06, 2020 3:51 pm
 I think there is a slight error on the Winter 1901 page.
Thanks, fixed it, and added colors.
DougJoe wrote: ↑
Fri Nov 06, 2020 4:06 pm
 Did you get a dump from the DB or something like that?
I was convinced that I should learn some BeautifulSoup...

Re: FvA Opening Database

by DougJoe » Fri Nov 06, 2020 4:06 pm

This is really cool. How did you pull the moves from so many games? Did you get a dump from the DB or something like that?

Re: FvA Opening Database

by jasnah » Fri Nov 06, 2020 3:51 pm

I think there is a slight error on the Winter 1901 page. The description on the left says that stats are given in the order:

n games
Austria's mean score
France's mean score

But in the actual table, France's score is listed above Austria's.

Re: FvA Opening Database

by jasnah » Fri Nov 06, 2020 3:44 pm

Omg, fantastic. I've really wanted to see something like this for a while - great work and thanks for putting it together!

FvA Opening Database

by RoganJosh » Fri Nov 06, 2020 3:34 pm

I was convinced that I should learn some Tableau. So I put together an opening database for FvA:

https://public.tableau.com/profile/jens ... ublish=yes

It stops at builds 1901. For two reasons: 1) even for the most common combinations of openings, there are just too few games to make any statistical analysis past 01, and 2) I fear the Tableau format isn't malleable enough to push it further. It's fancy and all, but it's not built for this.

What are the takeaways? For inexperienced players, just looking at what orders are played the most will help you avoid most mistakes in 1901. For more experienced players, I'd guess the approx. payoff matrices for the builds in 1901 will be the most interesting.

Let me know if there are any improvements you'd like to see. I'm not promising anything, but I might fancy a challange.

Top