Gunboat Stats

Use this forum to discuss Diplomacy strategy.
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.
Message
Author
RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Gunboat Stats

#1 Post by RoganJosh » Thu Dec 31, 2020 7:56 am

In case anybody is interested:

https://public.tableau.com/profile/jens ... e/Overview

Based on 15k games from webDip. There's plenty to click around and explore. Let me know if something is missing, or if something isn't working. This is just a preliminary version.

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Re: Gunboat Stats

#2 Post by RoganJosh » Thu Dec 31, 2020 8:09 am

One highlight:

If you rank powers by average share of the pot for all games, then you get:

Turkey .196
France .183
Germany .153
Italy .131
England .128
Russia .108
Austria .100

And if you filter by games who's pot size was at least 701 (i.e., at least a bet of 101) and that was played during the last 2 years:

Germany .216
France .213
England .135
Italy .126
Austria .110
Turkey .107
Russia .094

Turkey has not been performing good in high pot games lately... In fact, it's about 2 years since the last Turkish solo in a gunboat game with 101+ pot...

teccles
Posts: 237
Joined: Mon Mar 25, 2019 8:10 pm
Contact:

Re: Gunboat Stats

#3 Post by teccles » Thu Dec 31, 2020 9:13 am

Very nice. I've already found several surprises in the opening stats, which I need to think about. Perhaps the most notable are the success of Russias who move to Finland (maybe they get Sweden more?) and the anti-French DEN, MUN, RUH opening for Germany (though this seems to have died out).

A quibble: the 'average return' column is quite surprising (and I initially assumed it was 'percentage of pot'). I'd maybe remove it or replace with 'average pot size' if you want to convey that?

jasnah
Posts: 507
Joined: Tue Jun 02, 2020 8:13 pm
Contact:

Re: Gunboat Stats

#4 Post by jasnah » Thu Dec 31, 2020 11:17 am

teccles wrote:
Thu Dec 31, 2020 9:13 am
Very nice. I've already found several surprises in the opening stats, which I need to think about. Perhaps the most notable are the success of Russias who move to Finland (maybe they get Sweden more?) and the anti-French DEN, MUN, RUH opening for Germany (though this seems to have died out).

A quibble: the 'average return' column is quite surprising (and I initially assumed it was 'percentage of pot'). I'd maybe remove it or replace with 'average pot size' if you want to convey that?
Anti-French? All it seems to me to do is increase the chance that France will get Belgium, Germany will get bounced out of Holland or both.

teccles
Posts: 237
Joined: Mon Mar 25, 2019 8:10 pm
Contact:

Re: Gunboat Stats

#5 Post by teccles » Thu Dec 31, 2020 11:42 am

Indeed. My initial thought was that Germany was gunning for Burgundy, but that's so much worse than Holland that it can't be right.

I guess this opening makes sense in a meta where France would sometimes go for Ruhr in Autumn? It discourages that move. Otherwise I can't make much sense of it.

Yonni
Silver Donator
Silver Donator
Posts: 408
Joined: Thu Oct 19, 2017 6:55 pm
Contact:

Re: Gunboat Stats

#6 Post by Yonni » Thu Dec 31, 2020 12:49 pm

Very cool job!

For comparison, here's analysis that Tru Ninja did a while ago for full press games. I was going to say a few years ago but, holy hell, it's almost a decade old now.

https://onedrive.live.com/view.aspx?cid ... 5D9F%21176

jay65536
Posts: 59
Joined: Mon Dec 16, 2019 7:36 pm
Contact:

Re: Gunboat Stats

#7 Post by jay65536 » Thu Dec 31, 2020 1:53 pm

teccles wrote:
Thu Dec 31, 2020 11:42 am
Indeed. My initial thought was that Germany was gunning for Burgundy, but that's so much worse than Holland that it can't be right.

I guess this opening makes sense in a meta where France would sometimes go for Ruhr in Autumn? It discourages that move. Otherwise I can't make much sense of it.
I don’t play gunboat, but I can speak to this opening, as I have seen it played and used to think it was rubbish but recently I figured out what the point of it is.

Many German players want to keep an army in Munich. Moving Ber-Kie and Mun-Ruh, which is “standard”, leaves Munich open if France goes to Burgundy (as many times happens). Rather than risk it, they just leave Munich at home.

In press games, this is often accomplished by bouncing in Bur (or less often somewhere else like Tyr or Sil) and keeping the Ber-Kie move intact. But there is one small problem with this: it allows England to bounce Holland if she wishes. And if that happens, not only do you not get the build, but Kiel is occupied and you can’t build a fleet there.

So this opening is meant to mitigate that problem. It allows you to keep an army in Munich (which is often good) while guaranteeing that if you get bounced from Holland you can retaliate by using your one build for F Kie.

jasnah
Posts: 507
Joined: Tue Jun 02, 2020 8:13 pm
Contact:

Re: Gunboat Stats

#8 Post by jasnah » Thu Dec 31, 2020 2:23 pm

jay65536 wrote:
Thu Dec 31, 2020 1:53 pm
Many German players want to keep an army in Munich. Moving Ber-Kie and Mun-Ruh, which is “standard”, leaves Munich open if France goes to Burgundy (as many times happens). Rather than risk it, they just leave Munich at home.

...

So this opening is meant to mitigate that problem. It allows you to keep an army in Munich (which is often good) while guaranteeing that if you get bounced from Holland you can retaliate by using your one build for F Kie.
Pfft. Who cares if France goes to Burgundy? He'll have to use it to cover Marseille in autumn anyway. Go to Ruhr and Kiel. They're the best orders. /endrant

tr1285
Posts: 59
Joined: Tue Oct 23, 2018 8:25 pm
Location: NJ, USA
Contact:

Re: Gunboat Stats

#9 Post by tr1285 » Thu Dec 31, 2020 3:09 pm

Great stats!

A few things I found, though I'm not sure if all of these numbers are statistically significant?

In PPSC games (all prior to 2017), Turkey was the best power (.23 avg share)

In 2017-2018, France and Turkey were about tied for the best (.19) with Germany 3rd (.16)

In the last 2 years, France, Turkey, and Germany are all about even (.17 avg share).

One thing I couldn't find is a breakdown of how often the different powers make it into draws. I think that would be interesting too.

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Re: Gunboat Stats

#10 Post by RoganJosh » Thu Dec 31, 2020 3:16 pm

teccles wrote:
Thu Dec 31, 2020 9:13 am
Perhaps the most notable are the success of Russias who move to Finland (maybe they get Sweden more?)
This was surprising to me too. I always thought Fin was weak because you can't threaten BAL if G decides to bounce.
Yonni wrote:
Thu Dec 31, 2020 12:49 pm
For comparison, here's analysis that Tru Ninja did a while ago for full press games. I was going to say a few years ago but, holy hell, it's almost a decade old now.
He definitely put a lot of effort into the analysis.. I merely made the data available. But reading through that text, I realized I completely forgot to add filter for phase length and turn.

Also, I trashed all games that ended in A 1905 or earlier. It doesn't say so anywhere in the viz now.

@jay and jas
Given the opportunity, does France really want to capture Mun in 1901? In most cases, you won't be able to keep it in 1902. And all eyes will be on you.

jay65536
Posts: 59
Joined: Mon Dec 16, 2019 7:36 pm
Contact:

Re: Gunboat Stats

#11 Post by jay65536 » Thu Dec 31, 2020 5:19 pm

Again, I don’t play gunboat, and I personally have never played this opening. I’m just pointing out that it’s not strictly inferior to a different opening—it’s for German players who want to keep Munich occupied and Kiel open. Whether you want to do that in gunboat is beyond my ability to answer with any confidence.

Aristocrat
Posts: 998
Joined: Sun Jul 19, 2020 6:09 pm
Contact:

Re: Gunboat Stats

#12 Post by Aristocrat » Thu Dec 31, 2020 5:40 pm

I would guess the Finland opening scores highly because (1) it is the only way to communicate to Germany that you are pro-them and (2) the only people who would think about that angle instead of just moving Bothnia (which is strictly superior in press) are probably more skilled than usual, so their higher scores come from better play overall and not necessarily the Finland move.

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Re: Gunboat Stats

#13 Post by RoganJosh » Thu Dec 31, 2020 5:59 pm

I added a panel where you consider only one opening of one power, and where you can compare how it fares against the standard openings of a second power of your choice.

leon1122
Posts: 437
Joined: Fri Sep 29, 2017 3:43 pm
Contact:

Re: Gunboat Stats

#14 Post by leon1122 » Fri Jan 01, 2021 4:11 am

Is there any way to find which games had a particular opening? For instance, the highest pot opening for Germany is Ber-Sil, Mun-Ruhr, Kie-Bal. Only 5 games were played with this opening, and I'd like to see how Germany followed up on it.

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Re: Gunboat Stats

#15 Post by RoganJosh » Fri Jan 01, 2021 2:43 pm

leon1122 wrote:
Fri Jan 01, 2021 4:11 am
Is there any way to find which games had a particular opening? For instance, the highest pot opening for Germany is Ber-Sil, Mun-Ruhr, Kie-Bal. Only 5 games were played with this opening, and I'd like to see how Germany followed up on it.
Yes - the tab called "Compare Openings". Maybe it need another name. Set Germany to that opening, and keep the other powers to (All), and it'll give you the statistics and the Game ID's for those games. (They are 64634, 79592, 87822, 236866, 245045)

There's a second way to do it. In 'All Openings' tab, if you find that opening in the list of German openings. If you click the 5 in the count column, and then keep your mouse hovering over it, then you'll get the 'Games: 5' pop-up with a small menu attached to it. That menu has three options: Keep Only, Exclude, and View Data. If you click View Data, then you get a pop up window where it pulls those five entries from the dataset, and you can read of all the information that's stored about them from there, including the GameID.

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Re: Gunboat Stats

#16 Post by RoganJosh » Fri Jan 01, 2021 8:03 pm

tr1285 wrote:
Thu Dec 31, 2020 3:09 pm
One thing I couldn't find is a breakdown of how often the different powers make it into draws. I think that would be interesting too.
I added a 'Results by Power' tab, where you now can see such things.

One more interesting thing:
The decrease in number of games is almost entirely a decrease in live games.

AryaStark
Posts: 3
Joined: Fri Sep 18, 2020 5:23 am
Contact:

Re: Gunboat Stats

#17 Post by AryaStark » Sat Jan 02, 2021 12:50 am

Great stuff here! I was wondering if you could add median pot statistics rather than just average stats as well. Great work though this is awesome!

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Contact:

Re: Gunboat Stats

#18 Post by RoganJosh » Sat Jan 02, 2021 1:30 am

Well, the median pot is zero for all common openings except for a few French ones. You can see that from the pie chart on the Standard Openings tab. These distributions are not standard, so it's not obvious what statistics to show.

AryaStark
Posts: 3
Joined: Fri Sep 18, 2020 5:23 am
Contact:

Re: Gunboat Stats

#19 Post by AryaStark » Sat Jan 02, 2021 1:33 am

RoganJosh wrote:
Sat Jan 02, 2021 1:30 am
Well, the median pot is zero for all common openings except for a few French ones. You can see that from the pie chart on the Standard Openings tab. These distributions are not standard, so it's not obvious what statistics to show.
Ahh ok, thanks for the reply

Wusti
Posts: 409
Joined: Thu Oct 19, 2017 10:12 pm
Contact:

Re: Gunboat Stats

#20 Post by Wusti » Sat Jan 02, 2021 4:40 am

I can't believe the difference that Phase length makes. Essentially the longer the phase length - the more likely you are to Draw.

Edit: Cancel that - it seems to be unique to 2 Day phase length.

I'd actually like to see this run up in SAP's Analytics Cloud product which has some Machine learning and Smart Insights built in might come up with something that's not immediately apparent.

Is it a straight csv file source, and how big?

Post Reply

Who is online

Users browsing this forum: No registered users