Why ELO doesn't suit to BGA

[Zetes]

HI BGA players,

I want to give BGA developers some ideas about getting this elo thing to another level.
I like how the elo rating works right now.
I don't like how prestige is awarded…at today the results are that nr.1 players tend to settle there with some easy "keep me in the rank" games and, most important, it is really rare on BGA to see games between 2 players in the top 10 of a game. If BGA would post the statistics of those "top games" I think everybody would understand why they want a more dynamic system.

I will introduce an idea to make sure BGA becomes a really dynamic site….

As now prestige is mainly boosted by being in the top 3 for as many games as possible, and I agree that it is too easy to stay at the top being "lazy".
My idea is to leave the elo system as it is with math changes welcome if needed, Glitter and Doom seems to be a decent math guy

Then I would also introduce sort of seasonal contests but with tournaments. Elo ranking is to make sure you get in top 8/16/32 and gets you some prestige, then, at fixed dates, the master tournament starts with all the best players competing against each other, trophies will be permanent achievements also worth prestige. During "off season" players try to boost their elo, during the "season" they compete against the elite. This is dynamic! You like top rated players to play at least weekly? You are right…but keep it interesting so they may not become lazy…

Of course players not showing up in the season tournament will get a serious elo downer.

During " off season" I would introduce some minimum objectives you need to accomplish in order to be able to partecipate at the season tournament, example…there are 2 months "off season"…you need to have played at least 8 games against an opponent of equal level ( 1 per week)…if there are less than X players at one level ( say not enough masters), you may play those 8 games also with players of 1 level less of yours.

At the end of the off-season you also get a trophy for the ranking you have (prestige-worth)….so…all those trophies will be permanent and nobody can take that away. I would split the prestige equal, or near to equal, between nr. 1 in the offseason and nr.1 in the tournament so that you really need both trophies in order to be sure to sit permanently at the top of BGA rankings.

Choose if to play only off season, become nr.1 and then don't play the tournaments….you will forever have the icon of best player during X off-season but those achievements won't be enough to become nr. 1 overall on BGA or stay at nr.1 for that game if you never take it up against the elite players in the season tournament.

I understand very good that this system is not ok due to turn based and real time players….but…let's solve also this problem:

The top ranked players in BGA are almost all turn based players, so, how do they know they are really the best if they never will cross "only real time" players that might be better than them?

Just make 2 separate rankings, turn based and real time, than also my idea would work

It is also such a different way of play, you may say it requires totally different skills…reality is that BG's are played live, an evening, a sunday afternoon…and that requires the skill of being faster than your opponent in seeing the best move. I think that is an excellent part of the game…but I am not gonna say turn based should be removed…people don't have all that time? Ok…but..different skills.

So why not separate the rankings? You can play both styles and being in the 2 rankings at the same time if your life allows that.

Also, keep the tournament mode as it is now active, just introduce those master tournaments with heavy prestige, I really want to see the best against the best regulary, then you can say who is really the best…and mind you…I am not the brave guy here…nr. 1 will not be me, I crossed much better players than me and I don't have all that time anymore to play on BGA as I am using my experience of expert BGA player together with my creativity to become BG designer myself.

I also like the idea of balancing prestige for more or less played games….I am nr.1 for a few games, one is Tobago, not many people play that game, thx to that achievement and a similar one I am a top 20 BGA guy? No way!

Finally, you could go on with the season tournaments also for the lower ranked players to make sure many more BGA players are involved in the season tournament.

Players out of the season tournament could stay in off season mode and continue to play against each other.

I know this is a major change and I am aware that I might have overlooked some problems my idea can create. But humans always seem to solve problems, isn't it

What I know is that having top rated guys "sitting" on their throne and never see them battle it out against each other is not what I like.
For the 2 games I am nr.1 at I battled every guy (or almost) in the top ten and so I feel a proud nr.1 even if I think I should get less prestige because of the fact those 2 games have not many played games.

If you don't like this idea, do you agree a solution should be found anyway not only to cut "nr.1 lazyness" but also to have more plays between top rated guys to see who really is the best?

[demiurgsage]

For ex (extreme example) : Heartstone is doing a reset of all rankings every month. And no one can pretend that it breaks players motivation, isn't it ?

Are you sure about that?
Not so many players have enough time to get every month top10 in Legend. I'm sure most of players almost completely stop playing HS after they get high places couple of times. If you haven't HS channel with subscribers you have no need to be in Legend every month.

[glitter and doom]

Well, as I said, we will do a test, and if it does not solve issues we're not going to continue it. Ranking is a very complicated thing and we cannot determine a good solution only by arguing.

I support the idea to change the current system, I agree with the objectives you are trying to accomplish, and I agree that we cannot be sure how the system will work until we test it.

However, I don't understand why do you have to go with the system that seems to be the least favourable on this topic. As far as I've seen, almost no one likes the proposal. The most of the players have very strong feelings against it.

There have been many good proposals on this topic, why don't you test one of them instead?

Keep the Elo, but untie the prestige points and rankings to it. Introduce some kind of seasonal points, which will be reset (not decreased) at the beginning of each season. Each season is a new contest, everyone starts from the same position (unlike in your proposal).

[chrisa]

I am trying to follow this long thread, but I must admit that I am a bit lost and not sure anymore of what admins are proposing

Anyway, I agree with what has just been said:

Keep the Elo, but untie the prestige points and rankings to it. Introduce some kind of seasonal points, which will be reset (not decreased) at the beginning of each season. Each season is a new contest, everyone starts from the same position.

I understand the will of increasing the motivation of new players and beginners, but I would not like to see all the top-ranked (expert/master) players leaving BGA. Remember that many of them have been playing here (and have been supporting BGA) for several years, and, doing this, have contributed to make this website so funny.

For ex (extreme example) : Heartstone is doing a reset of all rankings every month.

I don't know much about Heartstone but I am not sure it is fair to compare a single-game platform with BGA, that hosts almost 100 games, with some players trying to reach a decent (Elo) level at several/many of these games.

[demiurgsage]

The proposal for beginners rating.
sourisdudesert said that most of beginners lose their rating from starting 1500. Actually they play worse than 1500 so they lose their rating. But this thing can be changed. Lets beginners start from 1350 or 1400 and get the lack of rating during first 15-25 games. E.g. 10 additional ELO points per first 5 games and 5 additional ELO points per other 10 games.

Also the K-factor on BGA is too big. I lost about 30-35 points in most of my first 2p games. Of course it's disappointing. (In game with 6 players you can easy lose 100 points) If K-factor would be lower I would lose less points and it would be less disappointed.

And you should change ELO formula for multiplayer games.
The best suggestion I think is

• 1st player get a victory against every other player.
• 2nd to last players get a loss against average opponent (no need to last player help to win a player with most ELO)
• after that 2nd place get some points on the same basis but divided by some factor (e.g. divided by 5). 1st player is excluded from these calculations.

Maybe this system isn't the best for all games but it's better than current when you get N/(2 * (N - 1)) of ELO you would get against every player in 2p.

[glitter and doom]

My idea is to leave the elo system as it is ...

Then I would also introduce sort of seasonal contests but with tournaments...

Thanks for your post Zetes, there are some very interesting ideas. I like the idea of official seasonal tournaments. Tournaments are already very popular, it would be great to introduce some kind of official championships for each game.

[latexrhombus]

Just make 2 separate rankings, turn based and real time, than also my idea would work

It is also such a different way of play, you may say it requires totally different skills…reality is that BG's are played live, an evening, a sunday afternoon…and that requires the skill of being faster than your opponent in seeing the best move. I think that is an excellent part of the game…but I am not gonna say turn based should be removed…people don't have all that time? Ok…but..different skills.

I agree with a lot of the things Zetes said in the other parts of his/her post like those about prestige based on tournaments rather than linked to ELO. The only change I would suggest there is the idea proposed by Cosmosis earlier that the prestige earned from the "season" goes away over time as to not create an insurmountable prestige wall for future players (but they can of course keep the trophy visible in their profile).

I quoted the section above because that is a pretty new comment in this thread and I think it makes a lot of sense. Playing turn-based and playing real-time really are quite different. I think they should probably be treated differently for rating-purposes. I realize this is way late to mention this, but the current Legendary Inventors contest has some definite potential for turn-based players to sort of swoop in at the last moment and take massive ELO gains that would be difficult to confirm if they were legitimate or not. I bring this up because I believe the same can be said for the existing ELO in some cases (certainly not all).

Also, it seems the "reset" is going to happen and like others I just wanted to take one more shot at saying there are some really nice alternatives here. I am mostly done talking about ELO decay, but even that would result in WAY less total ELO loss per player than a reset and still encourage participation. I also want to address the Hearthstone reference because their ranking system is really quite different. Up to level 20? (I believe) you cannot even lose ranking points by losing a match. Obviously on BGA you lose ELO every time you lose a match. And even once a player reaches level 20 and can lose ranking, they can never drop below level 20. So a reset makes a lot more sense there since it is like a race that encourages participation (you do not LOSE anything by participating, which is why a lot of top level players do NOT seem to play on here).

[sourisdudesert]

You may find the following Excel sheet interesting :

https://x.boardgamearena.net/data/newsi ... lysis.xlsx

For each games, for the 10 top players, I display :

• number of day since the last ranked game.
• number of ranked games in the last 60 days.

Yellow indicates players that has played 5 games or less in the last 60 days, which is very few to be considered as a "top 10" players.

As you can see :

• There is way too much yellow => If we keep ELO as it is no (without reset or decay), we cannot use it anymore for rankings.
• The only games with a decent rankings are the new ones : for most of the recent games, top players have all played a significant number of games and no one is "defending a position" => This is what incite me to recreate these conditions for all games at each seasons.

[sprockitz]

My two cents...

Elo Decay or resetting Elo are terrible options! Either one destroys what we've worked very hard to gain. I primarily only play from Oct-Mar because, while I love board games, I'd rather be outside when the weather is nice...you shouldn't punish me by decaying my Elo when I'm not active. I won't suddenly become average at a game by not playing it for 6 months. Resetting Elo is just a harsher form of decay (or maybe equivalent depending on how decay is implemented).

I like the idea of creating a new seasonal system where everyone starts at the same point each time through, but keep the lifetime Elo. The current Elo is a better representation of skill than any decaying or resetting Elo would be. A seasonable competition is more a measure of skill combined with activity, this is fine and could encourage people to play more since they have a better chance at a top spot, but would also force people with less time to only focus on a few games each season instead of exploring the almost 100 games available here.

How to deal with Prestige:

I think keeping lifetime best player (and top 20) is important.

I think giving more prestige for more popular games is good.

I think limiting prestige someone can gain in a given game is good (perhaps only prestige for overall and prestige for best seasonal finish)...though I don't know how many players come to this site trying to compete to have the most prestige...I know it never even crossed my mind when I started...and only occurred to me after playing for months and realizing I'm winning at almost every game I'm playing that it could actually happen.

I think changing the point distribution for Elo awards is good. Right now there is too much focus on the very top spots. Going from 50k for 1st down to 12k for 3rd and just 3k for 6th is a HUGE dropoff. Keeping 50k for 1st, why not try 35k for 2nd, 25k for 3rd, 20k for 4-5, 15k for 6-10, 10k for 11-20, and 5k for Elite (ranked in top 5% of active players). The added Elite rank helps for the more popular games, but you could scale the whole system for less popular games, starting with say 30k for 1st. Seasonal ranks would give something similar to these numbers, but again you'd be limited to the prestige gained from your best season + your overall (or just do best season or your overall if you really want to take the importance on overall Elo out as one seasonal 1st has earned you max points).

[chrisa]

@sourisdudesert: Thanks for these numbers.
Unfortunately, the only game for which I am in the top-10 (Expedition NorthWest Passage) does not seem to be listed in your file. I would have been interested to see the numbers for this game. And I would probably have commented that NWP games are quite long, especially when played turn-based (which is, by the way, recommended by BGA for this game - a warning message is displayed whenever you open a real-time table). It is not unusual that NWP games last more than one month, even with only two players, and it can be even longer with 3-4 players. Then I would not say that

5 games or less in the last 60 days, which is very few to be considered as a "top 10" players.

Very few players can play more often than that, given the length of this game...

Of course, this is not true for all games, some are very short compared to this one. But some players are very good at many games, I understand that they cannot play them all within a 60-day window.

I am not at all advocating for keeping Elo as a basis for ranking, I am fully in favour of seasonal trophies gained from seasonal points.
But I think Elo should be kept as is since it gives an idea of players' skills. I know that some players with high Elo do not play anymore ranked games (or against beginners), so as not to risk to lose their titles. Although I try to keep playing with highly-ranked people, I somehow understand them because you really lose too many points when you lose a game, even with someone of the same level as you. Additionally, it is always possible to check the game history of a player to see whether he plays regularly with good players or not, and then to have a better idea of what his Elo says.