May GhostRatings Published

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
CptMike
Posts: 309
Joined: Sun Dec 31, 2017 8:12 am
Location: Liège, BE
Karma: 54
Contact:

Re: May GhostRatings Published

#21 Post by CptMike » Sat May 19, 2018 6:36 am

MadMarx wrote:
Sat May 19, 2018 6:13 am
This month, CptMike, you are less than 7 points ahead of xorxes. Last month, my Elo was ~500 points, and you beat me twice in GvI while I was trying to get a better sense of how to play GvI at a higher level. Riddle me this, how much of an increase in Elo do you think you got from those two victories against someone with an Elo of 500 (I’m confident many people know the approximate answer, feel free to share as I am curious)? I don’t know the math, but I *guarantee* it was more than 7 points, so if you didn’t play those two games against me then you wouldn’t be ranked #1, thus those two games were *directly* responsible, it’s simple math.
Sure: if we remove some games, that influences the results. I agree that without both these games, I would not have got these points. But you could decide to remove my 2 victories against mhsmith0 (played during the same time as you) and I would not have got them. And if we remove the 10 games I have more than xorxes, that will also be the case… Also if I had lost one of these 4 games, that would not have given the same results either… So they were not "directly" responsible. That's a complete package.

Whatever I am not against the idea to have separate ELO's for GvI and FvA. I am just not sure that it would change anything. But I do underline that the 'weakness' of the GR system is somewhere else.

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Karma: 464
Contact:

Re: May GhostRatings Published

#22 Post by RoganJosh » Sat May 19, 2018 7:15 am

CptMike wrote:
Sat May 19, 2018 5:29 am
The reason why I am #1 and he is #2 is that during the last month I have played a tenth of games more than xorxes without losing any, whereas he lost some.
Wait, what? So, you did not lose any games, but xorxes did. And therefor you surpassed him in the rankings. And that's a problem:?: Surely, in any reasonable ranking system, if you don't lose any games then you will rise to the top.
2

Dejan0707
Posts: 118
Joined: Tue Jan 02, 2018 9:15 pm
Location: Dante's Inferno
Karma: 105
Contact:

Re: May GhostRatings Published

#23 Post by Dejan0707 » Sat May 19, 2018 1:16 pm

It doesn't neceserry mean that rank 1 is better than rank 2. Ranking is merely indicator. I remember easily defeating some player who was better ranked than me but also remember one player who was lesser ranked than me but I always considered him better player by far.

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#24 Post by MadMarx » Sat May 19, 2018 3:59 pm

It’s almost as if you guys are intentionally making arguments that have *nothing* to do with the issue I’m bringing up! ;-)

Let me try to come at this issue a different way, and then feel free to comment on what I am discussing rather than making “of course the grass should be green and the sky should be blue” sorts of comments. (Yes, I tend to exaggerate when trying to make a point, no offense intended!)

Let’s say Player A has earned a classic GR of 1000, so losing a single game costs 57 GR points, and would result in a new GR of 943. Let’s also say Player B has never played gunboat before, so his gunboat GR is 100 and a single loss costs him 3 GR points and would result in a new gunboat GR of 97. Let’s assume all that makes sense and is reasonable. Now, someone comes along and decides to combine classic GR and gunboat into a single GR, and the classic player with the GR of 1000 has never played gunboat (and his Diplomacy skill set is 100% based on chatting people up, none of which transfers to gunboat), he decides to play his first gunboat game ever, and of course he loses, which costs him 57 GR points?! To me, someone that has never played gunboat before and loses, they should lose points based on their gunboat skill, they should lose 3 GR points.

So, what I’m saying in real life, is that when I lost to CptMike in GvI, when I had little GvI experience, it would make more sense if I lost GR based on my skill, that I should have lost closer to 3 points to CptMike rather than 57 GR points (referencing the hypothetical scenario above). I am also saying that by awarding 57 GR points to CptMike rather than 3 GR points, that CptMike’s GR is over-stated, and this conceptual situation, over time, makes GR less and less meaningful since people are getting GR they haven’t earned and people are losing GR they should not have lost.

Again, I’m not trying to debate the exact nuances of how GR is calculated, I am not saying people shouldn’t gain GR when they win games, and I’m not saying GR is perfect and there is a way to get it to rank every single person correctly.

What I’m saying is that AvF is significantly different than GvI (per CptMike: “In GvI a big mistake does not mean the end. In AvF, in several configurations, the result of a 50/50 at the beginning of a game decides on the final result. AvF is much less challenging and more ‘mathematical.’”), that due to the fact that AvF was the only 1v1 offered on webDip for several months, GvI results are relatively skewed when looking at 100% of the webDip community (since that’s who GR ranks, everyone that plays), etc., etc., that GvI should not be included in GR with AvF.

User avatar
ghug
Bronze Donator
Bronze Donator
Posts: 17581
Joined: Mon Mar 20, 2017 3:51 pm
Location: Seattle
Karma: 11400
Contact:

Re: May GhostRatings Published

#25 Post by ghug » Sat May 19, 2018 4:30 pm

I should note that the two variants were introduced at the same time. The reason there's more AvF play is largely that captainmeme preferred it when he set up the initial tournament.

Onto the meat of the issue, Marx is right that any time you combine different games into one rating, you're going to have edge case issues with a player who's good at one but not the other having weird impacts on the ratings. The gunboat example is indeed exactly what happens in overall GR.

Whether it's an issue that can be solved is a different question. I hardly think excluding GvI entirely is fair, given that there is a large and constantly growing group of players that take it just as seriously as FvA. I'm also selfishly not in favor of increasing the number of GR categories above where it is because there's only so much automation that can go into that process as long as ratings aren't integrated with the site. I do think there is a decent amount of logic in grouping the two together, as they do have obvious fundamental similarities, and I think both are worth capturing.

I'm happy to run the two separately as a one-off for the curious, and if the majority of our 1v1 community wants something different, I'm also happy to cede to the will of the masses.
1

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#26 Post by MadMarx » Sat May 19, 2018 5:14 pm

I stand corrected!

Regardless of when GvI was introduced, or why AvF is more prominent, the fact remains.... I’m sure people that like the world map or ancient med would like a separate GR, but there are obvious limits, so I don’t have a problem weeding GvI out of the 1v1 GR. Overall GR is currently bastardized with multiple variants, per ghug, but that seems logical to have one all-incompassing GR (does it exclude 1v1?), it would be nice if the individual categories were as accurate as possible that avoided “edge case issues”. I argue that GvI and AvF are dissimilar, even though they are both 1v1, much like a classic game and gunboat are played on the same map and thus have similarities... Clearly I only have one voice, but if we’re going to have a 1v1 GR, why not prioritize accuracy over volume, sometimes less is more...

RoganJosh
Silver Donator
Silver Donator
Posts: 556
Joined: Sun Dec 31, 2017 1:02 am
Location: Stockholm
Karma: 464
Contact:

Re: May GhostRatings Published

#27 Post by RoganJosh » Sat May 19, 2018 5:38 pm

I wasn't arguing you at all, MM, I was arguing Cpt.

Separate rankings will of course describe your FvA and GvI skills more accurate, since the combined ranking only describes your ... combined skill.

It is a bit weird though, as you point out, that the 1v1 ranking describes, not the average of your FvA and GvI skills, but your combined skill weighted by how many FvA and GvI games you choose to play...

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#28 Post by MadMarx » Sat May 19, 2018 5:52 pm

For the record, in fear of it not being obvious, none of this is meant to be any sort of personal attack against CptMike, clearly he is a top tier AvF *and* GvI player, and it’s not even meant to be an argument in my eyes, more of a discussion amongst friends (after all, like ghug, I’m much more rude and confrontational with my friends than people I do not like)! :-)

Yonni
Silver Donator
Silver Donator
Posts: 408
Joined: Thu Oct 19, 2017 6:55 pm
Karma: 368
Contact:

Re: May GhostRatings Published

#29 Post by Yonni » Sat May 19, 2018 6:00 pm

When the variants have been on the site longer and the numbers get larger, it would be interesting to test if a combined ratings is a better predictor of success than the divided rankings.

I understand the arguments MM is making but I think they're not much of an issue as your games played increases in both of them. I would hazard a guess that the combined rankings would eventually be better just because of volume. However, I play diplomacy on the diplomacy site and not nerdy checkers or whatever this nonsense is so I've got no real horse in the race. I'll publish the separated GR2 ratings shortly.

User avatar
ghug
Bronze Donator
Bronze Donator
Posts: 17581
Joined: Mon Mar 20, 2017 3:51 pm
Location: Seattle
Karma: 11400
Contact:

Re: May GhostRatings Published

#30 Post by ghug » Sat May 19, 2018 7:12 pm

I've come up with a new system that may work better (though I doubt it'll appease Marx, seeing as he's dropped :razz:).

Instead of calculating one set of Elo scores using all games, I calculate one set of Elo scores for each variant and then calculate a player's final score as the average of those two scores, weighted by number of games played. For example, if my GvI Elo were 50 with 5 games played and my FvA Elo were 200 with 20 games played, my overall rating would be (50*5 + 200*20) / (20 + 5) = 170.

This allows us to factor in both variants while more accurately reflecting what's happened in the case Marx describes where a player has very different skill levels in the two variants. It also allows us to see people's individual skill levels by variant.

A spreadsheet with both ratings is here. Interestingly, Mike's still on top. Thoughts and feedback are appreciated.
1

User avatar
ghug
Bronze Donator
Bronze Donator
Posts: 17581
Joined: Mon Mar 20, 2017 3:51 pm
Location: Seattle
Karma: 11400
Contact:

Re: May GhostRatings Published

#31 Post by ghug » Sat May 19, 2018 7:20 pm

On another note, Mercy seems to be correct about the "Classic" rating including variants. I'll work on a fix for next month, but that code's a pain to work with, so it may be a bit.

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#32 Post by MadMarx » Sat May 19, 2018 9:04 pm

ghug wrote:
Sat May 19, 2018 7:12 pm
A spreadsheet with both ratings is here. Interestingly, Mike's still on top. Thoughts and feedback are appreciated.
More data = More fun!

Thanks ghug! It’s not like I’m trying to push for a system that ranks me as high as possible, simply a system that makes sense and is as reasonably accurate as we can expect given certain limitations... I like it!

User avatar
ghug
Bronze Donator
Bronze Donator
Posts: 17581
Joined: Mon Mar 20, 2017 3:51 pm
Location: Seattle
Karma: 11400
Contact:

Re: May GhostRatings Published

#33 Post by ghug » Sat May 19, 2018 10:21 pm

I know, I'm just teasing.

Is this something people would prefer to see to the normal version? Are there tweaks people would like?

jmo1121109
Lifetime Site Contributor
Posts: 1099
Joined: Fri Sep 29, 2017 4:20 pm
Karma: 2944
Contact:

Re: May GhostRatings Published

#34 Post by jmo1121109 » Sun May 20, 2018 1:16 am

Yeah if you could learn php and just integrate it to the site and save me some work that'd be awesome!

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#35 Post by MadMarx » Sun May 20, 2018 2:44 am

ghug wrote:
Sat May 19, 2018 10:21 pm
Is this something people would prefer to see to the normal version? Are there tweaks people would like?
Personally, I still think AvF and GvI are different enough that combining them is an extremely low priority and relatively meaningless... The AvF GR is most important, and throwing in a separate GvI GR would be pretty cool, but I probably wouldn’t clutter things up beyond that...

User avatar
ghug
Bronze Donator
Bronze Donator
Posts: 17581
Joined: Mon Mar 20, 2017 3:51 pm
Location: Seattle
Karma: 11400
Contact:

Re: May GhostRatings Published

#36 Post by ghug » Sun May 20, 2018 3:00 am

I still don't see why FvA is more important. It seems like people play them both very seriously at this point.

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#37 Post by MadMarx » Sun May 20, 2018 3:43 am

ghug wrote:
Sun May 20, 2018 3:00 am
I still don't see why FvA is more important. It seems like people play them both very seriously at this point.
Uhhh, there’s a Champions League for FvA, the third times it’s been done, but no such thing for GvI. GR2 shows ~2.5 times as many active FvA as GvI. And, most importantly, GvI is stupid. :-)

I think it would be great if we had GR for both AvF and GvI, but if we can have only one 1v1 GR it should be AvF rather than some bastardized version that combines the two (and, obviously, nobody is advocating for a single 1v1 GR to only be GvI).

jmo1121109
Lifetime Site Contributor
Posts: 1099
Joined: Fri Sep 29, 2017 4:20 pm
Karma: 2944
Contact:

Re: May GhostRatings Published

#38 Post by jmo1121109 » Sun May 20, 2018 5:19 am

Classic - France vs Austria (2 Players) - 6870 game(s) played on this server
Classic - Germany vs Italy (2 Players) - 3583 game(s) played on this server

While both are popular, FvA is clearly the more played.

User avatar
ghug
Bronze Donator
Bronze Donator
Posts: 17581
Joined: Mon Mar 20, 2017 3:51 pm
Location: Seattle
Karma: 11400
Contact:

Re: May GhostRatings Published

#39 Post by ghug » Sun May 20, 2018 6:11 am

That's a third of all games, which strikes me as significant.

Does what I posted today not fit your needs? It has individual ratings listed for each variant if that's what you want, and it addresses the issue you had with the old system. I fail to see how excluding a third of the games is any less of a bastardization.

MadMarx
Gold Donator
Gold Donator
Posts: 293
Joined: Sun Dec 31, 2017 12:01 am
Karma: 148
Contact:

Re: May GhostRatings Published

#40 Post by MadMarx » Sun May 20, 2018 3:04 pm

ghug wrote:
Sun May 20, 2018 6:11 am
Does what I posted today not fit your needs? It has individual ratings listed for each variant if that's what you want, and it addresses the issue you had with the old system. I fail to see how excluding a third of the games is any less of a bastardization.
I think what you posted was excellent, but initially you said you didn’t want to add any more categories to GR, so assuming each additional category you add is increasingly a pain, I’d definitely have AvF since you already have one 1v1 category, hope adding one additional category isn’t too taxing to include GvI, and I’d stop short of bothering you further with a combined category, but if you’ve reconsidered and can add everything then that’s great!

Post Reply

Who is online

Users browsing this forum: No registered users and 58 guests