Skill levels system

(INF3RN0) #21

Actually w/l is. If you’ve followed any game with a skill ladder all top players swear by the fact that if someone is indeed skilled enough they will carry their team. This holds up in xT from my experience, and although it might not always put players exactly where they belong, it does the job well enough. OBJ play and shooting are both important, but let’s be honest here… doing objs is not a skill, it’s a choice. Any player who can shoot can do the objs just as well as anyone else, but if they choose not to then they lose. In the end they are more skilled as a player, but less likely to win games. If it required a special skill to do objs it might make sense that it was equal to shooting, but it doesn’t so at a certain point you need good aim to really make a difference. That’s why a combination of wins representative of intelligent play and killing ability mark the top of the chart.

The exception is that teamwork plays a bigger factor than shooting sometimes or at least can close the gap, but not even the smartest players can win without aim.

(Phandy) #22

Inferno’s right. All major rating systems are normally based on ELO - http://en.wikipedia.org/wiki/Elo_rating_system. And the idea is that they will normally only used w/l as its determinating factor. It’s the idea that over a certain amount of games, when winning/losing is the most important factor of a match, a better player will average more wins against a worse player. The same works for teams, based on some kind of average rating.

(Mustang) #23

But it doesn’t have to be the only consideration.

For example a player could have a low individual skill, whilst at the same time having a high skill when playing with a particular team that always wins, this would mean his skill would be overrated when in public matches and if teams are balanced by this his team would get overrun.

You could guard against this by considering multiple factors, e.g. 50% team skill (W/L), 30% individual skill (K/D), 20% other (objective play, time alive, etc.)

I think everyone agrees that ELO weighting is beneficial though.

(Anti) #24

Ultimately Elo style ratings are not ideal, but they are the best we can do. Any system that starts to try and factor in performance as well as w/l brings in far too many unstable factors that can skew the resulting score and make it less accurate over time.

I think the best approach is to use Elo like systems to ranked players but to give each player a personal grade after each match, to tell them how they did in their role with their character compared to the population of other players, in the same way games like Street Fighter 4 grades performances (albeit that their system is score based rather than population based).

That way players can tell when they’re doing well or not, but ratings for matchmaking stay as accurate as possible.

(Anti) #25

[QUOTE=Mustang;490907]But it doesn’t have to be the only consideration.

For example a player could have a low individual skill, whilst at the same time having a high skill when playing with a particular team that always wins, this would mean his skill would be overrated when in public matches and if teams are balanced by this his team would get overrun.

You could guard against this by considering multiple factors, e.g. 50% team skill (W/L), 30% individual skill (K/D), 20% other (objective play, time alive, etc.)

I think everyone agrees that ELO weighting is beneficial though.[/QUOTE]

This is why many games keep multiple Elo ratings per player, with each rating counting for a specific type of play (i.e. solo play, group play, clan play, mode etc).

The systems that tend to fall down are those that use one rating for all, for example ranked play in Dota 2 currently treats All Pick matches with the same rating as the two draft modes. This means highly rated players from AP, who only know 4-5 characters, are vastly overrated when they come into the draft modes where they’re likely to get a random character.

(Phandy) #26

[QUOTE=Mustang;490907]For example a player could have a low individual skill, whilst at the same time having a high skill when playing with a particular team that always wins, this would mean his skill would be overrated when in public matches and if teams are balanced by this his team would get overrun.

You could guard against this by considering multiple factors, e.g. 50% team skill (W/L), 30% individual skill (K/D), 20% other (objective play, time alive, etc.)

I think everyone agrees that ELO weighting is beneficial though.[/QUOTE]

The one advantage of using only W/L though is that you truly enforce a play to win mentality, which is good. Because if the game rewards teamwork with winning, people will want to do whatever they need to win. If you include K/D etc people might focus on those other things with disregard for winning or teamwork etc.

For the example you gave, DOTA does something interesting which is, within the same MM system, tracks your individual MMR and your partyMMR. Then depending on whether you are solo/partied, uses the appropriate rating. They also do something where parties gain a bonus to their MMR because its assumed that parties will play better than if solo (they have extra communication and teamwork etc).

Like Anti said, its surely not perfect but generally from the games I’ve played using an ELO type system, it works pretty well.

(DarkangelUK) #27

CoD is a perfect example of this, all everyone cares about is their KDR. Doesn’t matter if you lost the match, if your KDR was positive then they classed that as a win in itself.

(RasteRayzeR) #28

A play to win mentality is good as long as it enforces teamplay, but in games like LoL it lead to people laregly leaving match making or fighting about the roles they will play in the game a lot because they don’t want to lose. This has disrupted for a long time the match making. I can remember having to go through 10 or 12 match makings in a row before I could join a game. As an answer Riot had to implement a reputation system and bans.

It can work, but you need a community that is disciplined.

(Mustang) #29

The primary motivations in my eyes are:

[ul]
[li]At the end of the match a player can see if they’ve played a good match or a bad match by their own standard (i.e. am I improving)[/li][li]Public teams are always as balanced as possible and not shuffled randomly or weighted by play time / character level alone[/li][li]Teamwork is promoted above individual success[/li][/ul]

These are the goals, so any system should be designed with these in mind.

(rookie1) #30

Another thing to look on : Is the player doesnt do well because of laging or stuff on servers .One reason a premilary evaluation test done localy on his machine can gather datas that can be usefull somehow

(INF3RN0) #31

[QUOTE=Mustang;490907]But it doesn’t have to be the only consideration.

For example a player could have a low individual skill, whilst at the same time having a high skill when playing with a particular team that always wins, this would mean his skill would be overrated when in public matches and if teams are balanced by this his team would get overrun.
[/QUOTE]

It can be true that a lower skilled player will perform better if they are paired with high skilled players, however the credit does not necessarily go their direction. At a certain point they would start to bring their team down and eventually drop out- this is normal behavior though sometimes people can get on a lucky streak for a bit.

I feel like CSGO has an extra variable in terms of KDR/Score because I’ve ranked up after a loss which I could only attribute to having the highest kills/score. I think KPM/DPM might be a good additional factor for micro-boosts. At a certain point anyone not carrying their weight or playing for the team will get punished, but as it has been mentioned winning games is all that matters and it tends to sort itself out over time.

(prophett) #32

[QUOTE=INF3RN0;490926]
I feel like CSGO has an extra variable in terms of KDR/Score because I’ve ranked up after a loss which I could only attribute to having the highest kills/score.[/QUOTE]

It does, but only if you are killing equal (or especially) higher skilled players. This will enable you to rank up or maintain your rank, even if you lose the match.

Going 10-5 vS higher ranked players holds more weight than going 20-5 vS lower tier players. You aren’t going to get any credit for mopping the floor with gold novas (but they get a lot of credit if they kill you).

Best explanation I could find, although I don’t like that he changed the name of the ranks;

DMG represent!