Improving ladder through map selection

[RealityCheck]

First, I am fully aware how scarce is the developer's time. Therefore I am not insisting that my idea should be ever implemented. Let's just say that I want to have a mostly theoretical discussion about how to improve FAF ladder.

I was thinking for months about what the ladder system should be like, but I can't seem to work out all the details needed to make it bulletproof.

So let me just share my basic idea:

Edit: The problem I'm trying to solve is best explained by KeyBlue:

A player who is the best on theta passage (and other 5x5 like maps), could be merely average on Roanoke Abyss (or other 20x20 maps). So it would be more enjoyable for this ThetaPro to play against a pro theta player on Theta and against an average Roanoke player on Roanoke.

So if someone has some weaknesses and some strenghts he would get games that match his strenghts and weaknesses.
No easy theta wins and no horrible Roanoke beatdowns.

The matchmaker will be allowed to match players with larger rating difference than now - for example, matching player with rating 700 vs player with rating 1500. However, the matchmaker should try to balance the match by selecting a map that favors the weaker player. (Edit: Or, put otherwise, the matchmaker tries to find a map on which both players have equal or similar strength.) The matchmaker will maintain statistics about players performance, so it will know everybody's chance to win on a certain map.

Hopefully this rule will cause the following effects:

- it will be easier for everybody to find a ladder opponent.
- weaker and casual player will tend to play their favorite maps more often.
- stronger players will get map variety and more challenge.

FAF beginners will be able to select their favorite maps. Thus their initial period will be easier and more enjoyable, since they will be able to learn the game on maps they know. When they become better and start playing against weaker opponents, they will start getting different maps.

With this system the map pool can be larger and consist of 50 or 100 maps. Nobody will be forced to learn them. You can get a map you have never played before, but your opponent will be weaker than you, so you will have a fair chance to win regardless.

I think such a system can work well with Supreme Commander, because with this game the gameplay can be very different depending on the map. Most players have weaknesses in different areas and the matchmaker can exploit this to create balanced matches.

[NapSpan]

That system existed before with no greater results, the ladder we have now is intended to make players play different maps, not always the same ones. Ladder is not so bad right now.

[biass]

It didn't exist like this, only in you had a veto system so you could opt out of shit maps instead of getting them 100% of the time
(Haven reef...)

[RealityCheck]

That system existed before with no greater results, the ladder we have now is intended to make players play different maps, not always the same ones. Ladder is not so bad right now.

This is interesting. Do you know why was this system scrapped? Because players tended to play the same maps?

Also, was this system indeed the same like what I am proposing? Because in my mind, my proposal can actually lead to a larger map variety. You will only play your favorite maps when you play against stronger opponents. When you play against weaker opponents, you will play their favorite maps. And since the map pool will be larger, then those maps will vary more.

Also, your "favorite" maps can change over time, because when I am saying "favorite" maps, I mean maps where you have best results. The matchmaker can also select map not because the weaker player is strong on it, but because the stronger player is weak on it.

[Astrofoo]

I think this is a very interesting concept. Basically you are balancing games not through rating, but through map diversity. Very unique idea imo. I don't know how difficult it would be to make this ladder system, it does seem a bit ambitious but with a game like FAF it could work nicely. Mainly do to the fact we have a smaller playerbase so that means there are less players at every skill level. +1 to your theory

[Hawkei]

It is an interesting idea. The non-random element in the map selection could create bias and artificially inflate a players ranking. It does remove much of the inherent randomness, and therefore, would have an effect on rating and what this means. As the only pure measure of skill can be done on a perfectly random system.

I'd recommend caution when widening rank difference threshold. If this difference becomes too great it will just result in players quitting.

[ZeRen]

It is an interesting idea. The non-random element in the map selection could create bias and artificially inflate a players ranking. It does remove much of the inherent randomness, and therefore, would have an effect on rating and what this means. As the only pure measure of skill can be done on a perfectly random system.

I'd recommend caution when widening rank difference threshold. If this difference becomes too great it will just result in players quitting.

maybe hide rating in game?

[RealityCheck]

Effectively, players will have a rating for each map. Average ladder rating can also be calculated, but the matchmaker will only take into account the map ratings.

Suppose that player A, who is good on Theta Passage is put against player B, who is good on Roanoke Abyss and weak on Theta. Player A has average rating of 700 and player B has average rating of 1500. However their map ratings for Theta Passage can be respectively 1000 and 1200. Now this is a reasonable difference - 200 rating points. This rating expresses their true strengths on this exact map, not some average strength for all maps. In some cases the matchmaker will be even able to find maps on which players have equal strength, even though they have for example 400 points difference in their average ratings.
The match can be launched with players map ratings shown ingame and then this will be perceived as fair.

My example may be crude btw. I don't know if there are many players with rating of 1500 who have such glaring weaknesses. But the point is that the matchmaker tries to make a best match based on ratings that are more real and concrete - instead as of now, where the matchmaker has only average ratings to work with.

[KeyBlue]

I think the big issue here is that such a system can't work with 50-100-... maps. (issues I see are in spoiler)

Spoiler: show

Lets use 50 maps for the examples.

Lets say it takes 5 games to learn a persons rating for a map. That means you need 250 games to be rated on every map.
And only if you were able to play each map exactly 5 times.
During this period you need to use the average rating, because you don't have enough information yet for all the maps.
So the rating for each map wasn't usefull yet.

Now consider that a player becomes better during his first 250 games.
This is especially so for new players, one would think.
Can you really use the 25 first games he played to get a proper rating? I doubt it.
So at what point will it become beneficial to use the map rating and not the average rating?

And what if you haven't played 1 map in 1000 games, because other maps were more beneficial for matchmaking?
When will you let average rating take over from map rating?

I don't think map rating can really show someones real and concrete skill on a specific map.
Games take to long and the frequency of each map is too low for each map rating to be up to date.

Maybe it is more feasible if you had a few categories like: 5x5,10x10 land, 10x10navy, 20x20 land, ...
This way you can keep each categorie up to date , since there is only a few.
Ofcourse it is difficult to have each map in one these categories, but the exceptions could just use average rating, like now, or an average of a subset of categories.
example: uncategorised 10x10 = average of 10x10 navy and 10x10 land

[Lieutenant Lich]

The idea is great but matching 700 vs. 1500 is not good and will definitely result in Hawkei's scenario - players will quit. I'd say the rating gap should be 400 from the players rating (from x-400 to x+400). Then if you lose vs. a stronger player, you should be able to learn off your mistake because the rating gap is not too great.