Recommend
2 
 Thumb up
 Hide
31 Posts
1 , 2  Next »   | 

Caylus» Forums » General

Subject: Why is Caylus not in the top 10 ? rss

Your Tags: Add tags
Popular Tags: [View All]
Guillaume Chaput
France
Veneux les sablons
flag msg tools
Avatar
mb
The question is not really about "it's a great game it should be in the top 10 instead of El Grande or Le Havre for instance" but more about numbers I don't understand.

Caylus average rating is 8.00
El Grande average rating is 7.99

Caylus number of votes is 8736
El Grande number of votes is 7986

So! How is it possible that Caylus is number eleven and El Grande number 10! I guess it's just that I don't understand the math that leads to the geek rating so if someone has an explanation.

Thx!
Guillaume
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Joe Kundlak
Slovakia
Bratislava
flag msg tools
designer
badge
Avatar
mbmbmbmbmb
I guess it depends on the Geek Rating (7.89 vs 7.92 for El Grande).

But I might be wrong
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Łukasz
Poland
Warsaw
flag msg tools
Non vi sed virtute, not armis sed arte paritur victoria.
badge
e^{i · π} + 1 = 0
Avatar
mb
Geek rating (based on which the rank is determined) is completely different animal than average rating (which is, allegedly, just a mathematical average i.e. (sum n) / #n).

Further read:

http://boardgame.geekdo.com/wiki/page/ratings
http://boardgame.geekdo.com/wiki/page/BoardGameGeek_FAQ#toc4
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guillaume Chaput
France
Veneux les sablons
flag msg tools
Avatar
mb
Yes I know it's about geek rating, but I always assumed it was some math between average rating and number of votes.
Seems that I was wrong
But anyway! Caylus should be in the top 10!
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Vital Lacerda
Portugal
Oeiras
flag msg tools
designer
badge
2010 - Vinhos, 2012 - CO2, 2014 - kanban, 2015 - The Gallerist, 2016 - Vinhos Deluxe, 2017 - Lisboa, 2018 - Escape Plan, CO2 Second Chance and Dragon Keepers - Maybe: 2019 - ROTW Portugal and On Mars, 2020 - Kanban Deluxe Edition
Avatar
mbmbmbmbmb
The same thing is happening to Puerto Rico and Agricola.
Why agricola is still the #1 when both have the same average rating, but Puerto rico have more votes?
You can find others in the some situation, like Race for the Galaxy, wish have a bigger average and more votes then El Grande, but it still have a lower Ranking.
I can understand why a higher average game with less votes could have a lower ranking, but I can not find a reason for the opposite.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guillaume Chaput
France
Veneux les sablons
flag msg tools
Avatar
mb
In the posts about rating I found:

"In effect, usually the games with many votes will Rank higher than those games with the same Average Rating but fewer votes."

So I still don't understand
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Łukasz
Poland
Warsaw
flag msg tools
Non vi sed virtute, not armis sed arte paritur victoria.
badge
e^{i · π} + 1 = 0
Avatar
mb
chaps357 wrote:
"In effect, usually the games with many votes will Rank higher than those games with the same Average Rating but fewer votes."


Unless we know exact rule, we may only guess.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Héctor MC
Spain
Madrid
Madrid
flag msg tools
badge
Avatar
mbmbmbmbmb
Actually El Grande average is 7.995241 (63850 points / 7986 votes) and Caylus average is 8.006868 (69948 points / 8736 votes), but it's shown rounded down, so they're very very close (just a little more than 0.001).

As the "Bayesian Averaging" method explained in http://www.boardgamegeek.com/wiki/page/BoardGameGeek_FAQ#toc..., games that have more votes will be less pulled down to the 5.5 "dummy" votes, so the 800 votes difference is guilty here.

Hope this helps
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
'Bernard Wingrave'
United States
Wyoming
Ohio
flag msg tools
Avatar
mbmbmbmbmb
Here's a link to check out, from the FAQ:
http://www.boardgamegeek.com/wiki/page/BoardGameGeek_FAQ#toc...

I believe that the Geek Rating column that shows up here is the BGG Rating referred to in the FAQ.

Edit: You may also want to check out the final paragraph on this page, which refers to something done to counteract shill ratings:
http://www.boardgamegeek.com/wiki/page/Shilling
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Will
United States
Fresno
California
flag msg tools
Avatar
mbmbmbmbmb
IMHO, probably because its a bit dry and there are other more fun games in the top 10 for the same amount of playtime. I played it once and there's many other games I'd rather play. Its not that I didn't like it, its just not way up there on my "want to play" list.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Łukasz
Poland
Warsaw
flag msg tools
Non vi sed virtute, not armis sed arte paritur victoria.
badge
e^{i · π} + 1 = 0
Avatar
mb
Yargo wrote:
IMHO, probably because its a bit dry and there are other more fun games in the top 10 for the same amount of playtime. I played it once and there's many other games I'd rather play. Its not that I didn't like it, its just not way up there on my "want to play" list.


Second that. Caylus is very good but there are better, deeper and more thematic games. Note, however, that my top 10 would look completely different to that on BGG.

Ranking is for those who look for a point of reference - generally speaking it does not say that the game on 111th place is worse at anything compared to a game ranked on 11th place.

It also brings some fun into otherwise boring exchange of geeks comments on how wonderful games are and how much fun it is to play them - ranking plus your personal preferences plus someone with whom you do not agree => 20 page thread on why apples are better than oranges.

Cheers.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guillaume Chaput
France
Veneux les sablons
flag msg tools
Avatar
mb
Yeah I agree it's not so good to look too much at the ranking, but well, when you are looking for a new game, you usually look at the top rated games... Because chances are that they will apply to more people
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Werner Bär
Germany
Karlsruhe
Baden
flag msg tools
badge
Avatar
mbmbmbmbmb
chaps357 wrote:
So! How is it possible that Caylus is number eleven and El Grande number 10! I guess it's just that I don't understand the math that leads to the geek rating so if someone has an explanation.

There's an anti-shill algorithm in place. It isn't told how it works. But when there are user accounts that rate only one game, rate it 10 (or a 1), there are no other activities, and the user never logged in again; then it's likely that these rating will be ignored when calculating the bayesian average. Don't know how it's now, but when Caylus was new, it had a lot of these suspicious ratings.
5 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guillaume Chaput
France
Veneux les sablons
flag msg tools
Avatar
mb
Oh, so I guess this is suppose to prevent too much hype.
Make sense
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Bernard
Netherlands
flag msg tools
Avatar
mbmbmbmbmb
Werbaer wrote:
chaps357 wrote:
So! How is it possible that Caylus is number eleven and El Grande number 10! I guess it's just that I don't understand the math that leads to the geek rating so if someone has an explanation.

There's an anti-shill algorithm in place. It isn't told how it works. But when there are user accounts that rate only one game, rate it 10 (or a 1), there are no other activities, and the user never logged in again; then it's likely that these rating will be ignored when calculating the bayesian average. Don't know how it's now, but when Caylus was new, it had a lot of these suspicious ratings.


That is quiet interesting and good I think. Is it also true that a persons ratings are more powerful if their average rating is lower? Or when a person has given more ratings?

I can see why they won't tell how it works exactly, because then people will abuse the system.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Brandon Tibbetts
United States
Oak Lawn
Illinois
flag msg tools
designer
Avatar
mbmbmbmbmb
Bernaar wrote:
I can see why they won't tell how it works exactly, because then people will abuse the system.


You can easily think of all sorts of factors that should reduce the weight of a vote:

New user
Lack of votes on other games
All user's votes are high
All user's votes are low
All user's votes are high or low (no middle ground)
Vote is high or low compared to same user's other votes
Comment on game is less than a certain number of characters
Lack of user contributions

The algorithm could also "score" each of the other user's voted games in terms of similarity to the game in question using any number of BGG stats (mechanics, category, number of players, year made, designer, etc.) Based on this, an "expected" vote could be calculated for the game in question. The larger the difference between the actual vote and the expected vote, the less the vote could be weighted.

I imagine that if many of these forces are in play, it would be too time consuming to constantly adjust the BGG rating for every game in real time or even nearly in real time. More likely, a scheduled process steps through the votes of each game and adjusts them vote-by-vote over time. So if you make major changes to the criteria that would affect your votes, you probably wouldn't notice the results for days or even weeks (if at all, since it's still only 1 vote no matter what).

Hey, I'm just happy that Caylus crawled ahead of Race for the Galaxy! This thread brought that to my attention. Now we just need to work on getting Dominion out of there.
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Guillaume Chaput
France
Veneux les sablons
flag msg tools
Avatar
mb
Noooooo don't touch Dominion!!!
Didn't you hear? The new game to hate is Space Hulk! devil
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Andre Metelo
United States
Kennesaw
Georgia
flag msg tools
mbmbmbmbmb
That actually raises a question regarding one thing that I've noticed on some games.

There are games that have more votes and higher average user rating than games placed above them on the geek rate.

Examples: (as of 10/8)

_____________ Geek rate______User rate__Votes
Hannibal_______7.72___________7.99_______2077
Goa____________7.70__________ 7.82_______4566
Shogun_________7.69__________ 7.84_______3910
It this was just an amount of fixed dummy votes at 5.5, Hannibal should clearly have a lower geek rate.

_____________ Geek rate______User rate__Votes
Imperial______ 7.65__________ 7.82_______3275
TI3___________ 7.64__________ 7.84_______4415

_____________ Geek rate______User rate__Votes
GO____________ 7.59__________ 7.83_______4163
AoE III_______ 7.63__________ 7.80_______3566

_____________ Geek rate______User rate__Votes
Galaxy Tr______7.52__________ 7.64_______4366
Settlers_______7.51__________ 7.64______20031
Settlers is a little tricky because due to rounding we may end up with the numbers difference on the user rate (up to 0.99 points in the worst case). But if the assumption that 300 votes is enough to make a significant difference, 16K should affect it as well.

My guess, the rating uses some other value that it is not 5.5. Maybe it uses the median of all votes? But the dummy 5.5 vote will not generate these differences.

My 2 cents


(edit for reading and formatting. How do a force a tab into this?)
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Phil Alberg
United States
Cumberland
Rhode Island
flag msg tools
Whoopee Doopee, we have fun!
badge
You lost the game!
Avatar
mbmbmbmbmb
Why does it matter?
2 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Flying Arrow
United States
Pennsylvania
flag msg tools
badge
Avatar
mbmbmbmbmb
metelo wrote:
That actually raises a question regarding one thing that I've noticed on some games.


What you bring up is what pretty much the whole thread was about. It was interesting to see more examples, though.

One place where it is easy to see that the shillbusting is working to look at RPGQuest and Capes & Cowls: The Superhero Board Game. In this case, it's easy to see why RPGQuest has a worse ranking even though it has more votes and a higher rating. Look at the voters - half of them are clear shills (single rating and it's RPGQuest for a 10).
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Andre Metelo
United States
Kennesaw
Georgia
flag msg tools
mbmbmbmbmb
FlyingArrow wrote:
metelo wrote:
That actually raises a question regarding one thing that I've noticed on some games.


What you bring up is what pretty much the whole thread was about. It was interesting to see more examples, though.

One place where it is easy to see that the shillbusting is working to look at RPGQuest and Capes & Cowls: The Superhero Board Game. In this case, it's easy to see why RPGQuest has a worse ranking even though it has more votes and a higher rating. Look at the voters - half of them are clear shills (single rating and it's RPGQuest for a 10).


My bad,, did not include the question itself.. HOW THE GEEK RATE IS CALCULATED? As it is, seems like a black box.. never a good thing. There are a lot of guesses, but I never found a formal and final description of it. Not that it really matters, but the engineer in me likes formulas.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Jay Sachs
United States
Woodinville
Washington
flag msg tools
badge
Despite your hope, there is not even any inherent symbolism; gravity is simply a coincidence.
Avatar
mbmbmbmbmb
Werbaer wrote:
There's an anti-shill algorithm in place. It isn't told how it works.


Interesting. In encryption, security-through-obscurity is not highly regarded. In other words, just because the encryption algorithm is not published doesn't make it secure. I wonder if anti-shilling algorithms ought to be classified as such? In other words, is there an anti-shilling algorithm that even if it were public, it would still be effectively impossible to circumvent it?
2 
 Thumb up
0.02
 tip
 Hide
  • [+] Dice rolls
Flying Arrow
United States
Pennsylvania
flag msg tools
badge
Avatar
mbmbmbmbmb
metelo wrote:

My bad,, did not include the question itself.. HOW THE GEEK RATE IS CALCULATED? As it is, seems like a black box.. never a good thing. There are a lot of guesses, but I never found a formal and final description of it. Not that it really matters, but the engineer in me likes formulas.


Actually, I don't think the black-box is that bad of a thing. We know there is some sort of shill-busting, but we don't know exactly how. If we knew exactly how, it would be pretty easy for people to work around it.
 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Flying Arrow
United States
Pennsylvania
flag msg tools
badge
Avatar
mbmbmbmbmb
jaysachs wrote:
Werbaer wrote:
There's an anti-shill algorithm in place. It isn't told how it works.


Interesting. In encryption, security-through-obscurity is not highly regarded. In other words, just because the encryption algorithm is not published doesn't make it secure. I wonder if anti-shilling algorithms ought to be classified as such? In other words, is there an anti-shilling algorithm that even if it were public, it would still be effectively impossible to circumvent it?


I doubt it. There is no real way to tell the difference between a shill and a real account. It's just a lot of work to make a fake account look real, but it would be easy to do if you knew exactly which things you had to do (rate 50 games?) and which things you could ignore (post 50 comments on games and 1 review?).
1 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
Russ Williams
Poland
Wrocław
Dolny Śląsk
flag msg tools
designer
badge
Avatar
mbmbmb
FlyingArrow wrote:
jaysachs wrote:
Werbaer wrote:
There's an anti-shill algorithm in place. It isn't told how it works.


Interesting. In encryption, security-through-obscurity is not highly regarded. In other words, just because the encryption algorithm is not published doesn't make it secure. I wonder if anti-shilling algorithms ought to be classified as such? In other words, is there an anti-shilling algorithm that even if it were public, it would still be effectively impossible to circumvent it?


I doubt it. There is no real way to tell the difference between a shill and a real account. It's just a lot of work to make a fake account look real, but it would be easy to do if you knew exactly which things you had to do (rate 50 games?) and which things you could ignore (post 50 comments on games and 1 review?).

My first reaction was to doubt as well, but I could imagine some ideas. E.g. at least some shill accounts could be detected by looking at what IP address they came from, if they were created with many other new accounts that all rate the same game, etc. That wouldn't catch all shill accounts, of course, but would actually catch many of the obvious ones.

I can also imagine there needing to be some text posted by the shill account before its ratings are used, and once you have text from someone, analysis can be done on it. Most text from spammers is easily detectable, for instance.

Requiring multiple games to be rated seems a good requirement too. I can imagine various statistical tests to try confirming that a "reasonable" or "likely" set of games has been rated, e.g. not 1 hardcore wargame, 1 eurogame, 1 abstract, 1 dexterity, 1 party, 1 word game.

There are thousands of users; quite possible their data can be used to establish models of "typical" users, and accounts that deviate from the "norms" could be provisionally ignored as shills. There might be many false positives (real people falsely detected as shills and ignored in the ratings) but the cost of that is not necessarily too high.

If all the shill-detecting methods were publicized, it would indeed make it easier for a determined shill to circumvent them. But the key is that most shills probably don't care enough to go through that trouble. Also, analogous to how encryption publishes the algorithms but not the keys, the shill-busting algorithm could publish the algorithms but not the numbers. E.g. simply say "A minimum number of forum posts must have been made" without saying what that number is. Algorithms that analyze text for being sufficiently non-spammy/non-shilly can be publicized without specifying what the desired level achieved of some output parameter is. Etc.

I agree it seems different from the case of encryption/security, but it seems interesting and possibly fruitful to think about whether the shill algorithm could be public and still work. I bet that it could to some degree, more than we might think, even if not as well as encryption/security.
5 
 Thumb up
 tip
 Hide
  • [+] Dice rolls
1 , 2  Next »   | 
Front Page | Welcome | Contact | Privacy Policy | Terms of Service | Advertise | Support BGG | Feeds RSS
Geekdo, BoardGameGeek, the Geekdo logo, and the BoardGameGeek logo are trademarks of BoardGameGeek, LLC.