Field of Glory Tabletop Rankings Live!

[dave_r]

Just asking because after two years of competitions there has only been a 350 point swing either way, meaning somwhat glacial movement.

Sounds familiar to another scoring system...

[hammy]

Just asking because after two years of competitions there has only been a 350 point swing either way, meaning somwhat glacial movement.

Sounds familiar to another scoring system...

I would let the system run for a while before deciding if there is an issue Dave.

If you want to have an idea of the level of volatility try checking out Robert Taylors rating graph.

[Martin0112]

Well, an ELO-system is not extremly dynamically.
It is at the beginning for sure, but after a while, you should have found your place in the ranking.
It is not a system showing up quick changes, but, for example, by winning a tournament it can bring you between 50 and 150 points, depending on your emenies. So there will be changes regularly.

But a good ranking system should not change rankings very quick, otherwise you can use the average of the last 3 tournaments, which for sure will bring a lot of changes regularly.
The real question is: What do we want to see?

I have tested about 10 different systems and rankings, starting from giving people points for achieved ranks in tournaments, creating average values, penalizing people playing less, etc.
The result was really funny sometimes, and it turns out that a system taking into account lots of results over a long period will be creating the 'best' list.
That's why we decided to use this (first only for the small german community)
While testing it for germany, I started to enter some results from the Hall of Honour... and so all begins.

I'd like to know your expectations to a good ranking system, input like this is very valuable

[Robert241167]

Hey Hammy stop picking on my ranking !!

When I'm good I'm very good but when I'm bad I'm abysmal !!

Just wait until I knock of my last places in my first 2 tournaments !!

Rob

[hammy]

But a good ranking system should not change rankings very quick, otherwise you can use the average of the last 3 tournaments, which for sure will bring a lot of changes regularly.

I agree that the system should not change too rapidly but unlike the old DBM Glicko system the ELO one does not have the reliability factor which in the DBM implementation of Glicko meant that once you had played 100 or so games it was almost impossible to make significant changes to your ranking as the formula assumed that winning three or four tournaments on the bounce was just a series of flukes regardless of who you played. The introduction of Glicko in the UK coincided with a significant improvement in my level of play at DBM. It took me four years of winning or placing in almost every tournament I played in to get my rating up from 40th to the top 4. I don't mind it taking a while to change a ranking but to me this did feel a bit wrong. Sadly I never managed to get the Glicko admins to recalculate the rankings assuming I started to play after I got to be a decent player, it would IMO have been an interresting excercise as in theory the two rankings should (with 4 years of data) have been pretty much the same.

My example of Robert Taylor is intended to show that a player who improves their level of play significantly will be rewarded with a relatively rapid change in their rating. At present I think this ranking system will be dynamic enough although some sort of 'player of the year' race would be nice to see.

[hammy]

Hey Hammy stop picking on my ranking !!

When I'm good I'm very good but when I'm bad I'm abysmal !!

Just wait until I knock of my last places in my first 2 tournaments !!

Rob

Not picking on anything Robert.

You are one of the players who's play has improved significantly after a shaky start. As a result looking at your ranking graph makes for a good example of how the system works when players skill improves. It also happens that both Dave and I lost ELO points to you at Game 09.

[Robert241167]

Just playing Hammy

Playing regularly plus loaning competitive armies has helped.

You'll be glad to know I can pretty much field the army I borrowed from you last year so hopefully they will perform as well otherwise I'll be asking for your troops back!

Rob

[rogerg]

It is really nice to have all the information, however, I would suggest a note of caution on using the rankings for seeding competitions. My personal ranking is somewhat skewed by some experimental army design in early competitions combined with a lack of army choice due to the availablity of list books. The results over the past twelve months are usually a better guide to how tough an opponent is.

Rob Taylor, who is managing to feature a lot in this thread, is a relatively new player. He might win a 'most improved over the year' prize. His current rankings in both systems will include his results for last year's BHGS Challenge. One would expect him to get a rather better result this year.

[hammy]

It is really nice to have all the information, however, I would suggest a note of caution on using the rankings for seeding competitions.

Well I am not a big fan of seending anyway so I agree with you on this one

My personal ranking is somewhat skewed by some experimental army design in early competitions combined with a lack of army choice due to the availablity of list books. The results over the past twelve months are usually a better guide to how tough an opponent is.

That is one of the problems with any all games count system. Using oddball armies can hurt your ranking

It does look from the graph though that your ranking has mostly recovered from your poor start and as long as you keep doing as well or better than they system expects then it will increase.

[dave_r]

I have tested about 10 different systems and rankings, starting from giving people points for achieved ranks in tournaments, creating average values, penalizing people playing less, etc.
The result was really funny sometimes, and it turns out that a system taking into account lots of results over a long period will be creating the 'best' list.
That's why we decided to use this (first only for the small german community)
While testing it for germany, I started to enter some results from the Hall of Honour... and so all begins.

I'd like to know your expectations to a good ranking system, input like this is very valuable

Well.... I would like them to be an accurate representation of a player's skill as it stands - not as it was three years ago, which was the problem with Glicko. I would also like to see significant movement three years into the rankings so that people get better or worse. Again the example with Glicko is seeing somebody at the top of the rankings when they haven't played a competition game in over three years.

The big problem is of course tallying in the internation systems. It is still largely the case that people in the US only play people in the US, hence, you may get a player with an ELO rating of 2000 in one country, which equates to somebody in another country with an ELO rating of 1700. Glicko never resolved this and any rankings are really going to struggle as their simply isn't the tournament cross-over to get reliable rankings.

Glicko compounded this even further by the South West Syndrome, where players in the SW leagues had a high rating, but weren't as good as those on the main BHGS circuit with a similar rating.

If you can solve this problem then well done.... Of course, you have largely cheated by not allowing doubles (Which I think is a very good start)

[nikgaukroger]

It is really nice to have all the information, however, I would suggest a note of caution on using the rankings for seeding competitions.

Well I am not a big fan of seending anyway so I agree with you on this one

Quite.

The possible issue is easily avoided by not bothering with seeding.

[Martin0112]

I have tested about 10 different systems and rankings, starting from giving people points for achieved ranks in tournaments, creating average values, penalizing people playing less, etc.
The result was really funny sometimes, and it turns out that a system taking into account lots of results over a long period will be creating the 'best' list.
That's why we decided to use this (first only for the small german community)
While testing it for germany, I started to enter some results from the Hall of Honour... and so all begins.

I'd like to know your expectations to a good ranking system, input like this is very valuable

Well.... I would like them to be an accurate representation of a player's skill as it stands - not as it was three years ago, which was the problem with Glicko. I would also like to see significant movement three years into the rankings so that people get better or worse. Again the example with Glicko is seeing somebody at the top of the rankings when they haven't played a competition game in over three years.

The big problem is of course tallying in the internation systems. It is still largely the case that people in the US only play people in the US, hence, you may get a player with an ELO rating of 2000 in one country, which equates to somebody in another country with an ELO rating of 1700. Glicko never resolved this and any rankings are really going to struggle as their simply isn't the tournament cross-over to get reliable rankings.

Glicko compounded this even further by the South West Syndrome, where players in the SW leagues had a high rating, but weren't as good as those on the main BHGS circuit with a similar rating.

If you can solve this problem then well done.... Of course, you have largely cheated by not allowing doubles (Which I think is a very good start)

Well, I think this is a very good point to discuss with a beer or 2.
Will you come to brussels or Athens?
We can meet there and have some nice discussions.
By the way, we have at least an idea on how to avoid having people showing a big ELO but not playing anymore.
We think that we will have a filter in for all people not playing for let's say 3 or 4 years.
They will be in the ranking, but not shown, only popping up again if playing again.
As we are now only 2 years with the game, this is for the future, but an idea.

[dave_r]

I would certainly like to go to Athens. If it is cheap enough I will go. I have two consecutive second places to defend....

Surely worth a beer or two to discuss

[Ghaznavid]

Well.... I would like them to be an accurate representation of a player's skill as it stands - not as it was three years ago, which was the problem with Glicko. I would also like to see significant movement three years into the rankings so that people get better or worse. Again the example with Glicko is seeing somebody at the top of the rankings when they haven't played a competition game in over three years.

The problem is the combination of 'accurate representation of a players skill as it stands' and 'see significant movement'. Its a bit like Heisenbergs uncertainty principle. Accuracy needs lots of results for precision and to overcome flukes, while 'significant movement' argues for a limited set of results and 'skill as it stands' argues for only recent results, excluding flukes, which is also limiting the number of results and asks for the elimination of flukes somehow (which is what Glicko attempts with the reliability rating).
Getting all of those together is always going to be a compromise. The closer you get to one of those goals the further you move away from the others, there is simply no system that can accommodate everything perfectly. So we will always have to make compromises.
That's also why we offer to incorporate the various national ranking systems on request (basically showing the ELO-Rankings and the Rankings created using that nations own system side by side).

The big problem is of course tallying in the internation systems. It is still largely the case that people in the US only play people in the US, hence, you may get a player with an ELO rating of 2000 in one country, which equates to somebody in another country with an ELO rating of 1700. Glicko never resolved this and any rankings are really going to struggle as their simply isn't the tournament cross-over to get reliable rankings.

Correct and basically there is no solution to this. We have to accept the fact that the world wide (or even Europe wide) Rankings are less reliable then the nationwide rankings.

I would certainly like to go to Athens. If it is cheap enough I will go. I have two consecutive second places to defend....

As long as I don't have to play you in the 1st round.
If you want it to be affordable I suggest you decide soon though. Not sure about the UK but in Germany, the cheaper flights to Athen (not the cheapest airport in Greece to begin with) are rapidly selling out.

[rbodleyscott]

Is buried in there somewhere a statistic that indicates how drawish/decisive a particular army has been, other than manually tallying the distribution of results for a particular army from the army details?

I've added an table showing the relative "army decisiveness" right above the bar-graph on the army-details pages.

May I suggest that this would be better expressed as a % of the total games included for that army.

Not sure if this idea got lost in the depth of the thread.

[madaxeman]

Not picking on anything Robert. You are one of the players who's play has improved significantly after a shaky start.

I suspect I fall in the category of someone who is consistently shaky with occasional random improvement.

[azrael86]

Well.... I would like them to be an accurate representation of a player's skill as it stands

It's possible to take this a bit too far, and the use of the Chess system is rather misleading. Chess has no chance element, and a known level of balance in every game (about 55/45 IIRC): obviously wargames involve several significant variables, both in terms of matchups and of course random factors. I'm not complaining about this, just saying that it isn't quite the same as a FIDE ranking where there is no chance component.

The big problem is of course tallying in the internation systems. It is still largely the case that people in the US only play people in the US, hence, you may get a player with an ELO rating of 2000 in one country, which equates to somebody in another country with an ELO rating of 1700. Glicko never resolved this and any rankings are really going to struggle as their simply isn't the tournament cross-over to get reliable rankings.

Well, the way to fix the international problem is to have a fully compatible version that you can play online?

Can Slitherine advise if this is ever likely to happen ? Or is it too niche a market ?

[dave_r]

You mean as well as the FoG PC game that they have just released with lot's of trumpeting sounds and so on?

[Martin0112]

We have taken into consideration that chess for example is onlöy knowing 3 results, win/draw/loss
This is different in our ELO-system, as it calculates the value based on the detailed result, so a game which ends 15:5 is not having that much influence on your ELO if you win 25:0
This makes the ELO for FOG more granular.

To compare different countries is always an issue as long as the countries are playing like in a nutshell.
The best way to compare results is having as much international comparisions in the ranking as possible.
Each game is helping here.
And you can see on the webpage, on the details for each player, there is an overview-page for all countries, showing the average ELO for this country.
Even with these few results between people from different countries, there are some changes in the average ELO for each country.
This can give you a feeling on how much your ELO is different in an international comparision.

The more tournaments we have with some international appearance, the better the value will be.
So tournaments like Athens, Brussels, Britcon or ITC will for sure help to get the results even more comparable.

I don't think a better system is to be found, even in soccer it was for years that all the national champions were playing in the european cups, before he Champions League was founded. And even there lots of countries are moaning about the modus.

[azrael86]

You mean as well as the FoG PC game that they have just released with lot's of trumpeting sounds and so on?

Yes. My understanding, and I haven't played it personally, is that the recent PC game is similar but isn't a complete rendering of the rules. So my question is simply - will a full version ever be made?