Forged Alliance Forever Forged Alliance Forever Forums 2014-11-26T14:12:12+02:00 /feed.php?f=42&t=8907 2014-11-26T14:12:12+02:00 2014-11-26T14:12:12+02:00 /viewtopic.php?t=8907&p=86713#p86713 <![CDATA[Re: Make ladder globalranked again]]>
This doesn't explain the average rating difference though, because while your rating in 1v1 goes up faster if you win, it also goes down faster if you lose. It's just faster change all around, it doesn't make the average higher or lower.

I could be wrong, but I think the reason the average is different is simply that certain parameters of trueskill are different in ladder than in custom (e.g. the rating that people start with, or the deviation that people start with, or the beta parameter, or something else).

Statistics: Posted by Vee — 26 Nov 2014, 14:12


]]>
2014-11-26T12:55:12+02:00 2014-11-26T12:55:12+02:00 /viewtopic.php?t=8907&p=86709#p86709 <![CDATA[Re: Make ladder globalranked again]]>
keyser wrote:
why would it be heavier weight in rating calculations in TrueSkill?


I think it's just the formula, ever noticed that if you do a custom 1v1 your rating jumps up or down more? It might be because it's easier for TrueSkill to determine what should be changed (no teammate influences and such). When it is the expected outcome the jumps are still slight of course. I don't know the specifics, but 1v1 just seems to speed up the path to an accurate rating.

Statistics: Posted by Aurion — 26 Nov 2014, 12:55


]]>
2014-11-26T12:41:25+02:00 2014-11-26T12:41:25+02:00 /viewtopic.php?t=8907&p=86708#p86708 <![CDATA[Re: Make ladder globalranked again]]> Statistics: Posted by keyser — 26 Nov 2014, 12:41


]]>
2014-11-26T12:13:20+02:00 2014-11-26T12:13:20+02:00 /viewtopic.php?t=8907&p=86704#p86704 <![CDATA[Re: Make ladder globalranked again]]>
Vee wrote:
Displaying both ratings solves everything.

1. Ladder players get to show their e-peen
2. Teamgame players get to show their e-peen
3. New players can get ladder rating from ladder, so that people know that they aren't 100% noob and then they don't get kicked from teamgames

It would also be great if the ladder & team game ratings would be based on the same standard. AFAIK ladder ratings average ~200 points lower than teamgame ratings due to how they are implemented differently. It would make more sense to base them on exactly the same calculation, with new players starting out with the same ladder rating & team rating.


Isn't it just the case that 1v1 has heavier weight in rating calculations in TrueSkill?

Statistics: Posted by Aurion — 26 Nov 2014, 12:13


]]>
2014-11-26T11:40:23+02:00 2014-11-26T11:40:23+02:00 /viewtopic.php?t=8907&p=86703#p86703 <![CDATA[Re: Make ladder globalranked again]]>
1. Ladder players get to show their e-peen
2. Teamgame players get to show their e-peen
3. New players can get ladder rating from ladder, so that people know that they aren't 100% noob and then they don't get kicked from teamgames

It would also be great if the ladder & team game ratings would be based on the same standard. AFAIK ladder ratings average ~200 points lower than teamgame ratings due to how they are implemented differently. It would make more sense to base them on exactly the same calculation, with new players starting out with the same ladder rating & team rating.

Statistics: Posted by Vee — 26 Nov 2014, 11:40


]]>
2014-11-26T10:11:54+02:00 2014-11-26T10:11:54+02:00 /viewtopic.php?t=8907&p=86698#p86698 <![CDATA[Re: Make ladder globalranked again]]>
Zoram wrote:
Blodir wrote:But yes, we've seen many examples of players getting high rating from 1v1, but while being fairly good, not playing even close to their immense rating in teamgames. They even admit to it themselves, and it probably makes their teamgame experience much worse.

Now I don't know exactly how trueskill works its magic, so I make assumptions here. I assume that a "global" rating system including both team games and 1v1 could put a different weight on each kind of game.

Similarly, does a player who only plays seton keep his rating increasing, or does TrueSkill considers that actually he is in effect only demonstrating skills on one map and diminishes the extra points he gets after each subsequent victory on the same map ?

I thought that -maybe- with the 1v1 ladder forcing you to play a large array of random selected map, rating there might reflect a bit better the "trueskill" of a player, as victories rely less on training for one map only.


In team games you have to pay attansion to all enemies and all ally, cause no one will ever say that there is TML or snipe or anything, cause due to balancing you most likely "should" be best player in your team, and you can't just sit and ecowhore/defend bettet cause your team may just die then.
Basicly you can't just outplay guy that play against you in 4v4 that will surely not enough to win game. Basicly I kinda agree with blodir...for example I think even after loosing about 100 points I am still overrated :)

But I also want to say that ladder was good way to get into faf so you can gain some global rating and start join different games that are not super retarded noob only

Edit: Also team games are maps often exteremly different from ladder maps, if you look at any 2.1k rating player they often build more power after hydro instead of making more factories, and can start teaching t2 mex even before they got all their t1

So another example would be Yann ~2100 global rating, he usually just makes only mantis and captures all map to build more mantis, in team game that only work as cheese tactic and "hope he does not have t1pd" tactic

Statistics: Posted by ZLO_RD — 26 Nov 2014, 10:11


]]>
2014-11-26T08:30:27+02:00 2014-11-26T08:30:27+02:00 /viewtopic.php?t=8907&p=86693#p86693 <![CDATA[Re: Make ladder globalranked again]]> I play custom to relax and I am not going to change that, because
some people think custom games should be rated to show their "potential".
So, I will do whatever feels pleasing to me, regardless of the consequences
to the game.
I think only ladder should be rated; because there's only 1v1 ladder
Only 1v1 games should be rated. If there were a 2v2 or 3v3 ladder then
2v2 or 3v3 games could be rated too and then a mean of your performance,
A "global" rank would make sense.

The fact that a player uses ladder to downgrade his rating artificially is not
Something that can't be fixed.There was just not the will to do it.
The notion that a ladder player is only good at 1v1 and abuses this to
Artificially raise his rating is hilarious. A 1v1 better player is military
And would defeat any custom civilian at a random map. Things can only change when
The civilian knows the map too well.

Statistics: Posted by prodromos — 26 Nov 2014, 08:30


]]>
2014-11-23T16:58:20+02:00 2014-11-23T16:58:20+02:00 /viewtopic.php?t=8907&p=86446#p86446 <![CDATA[Re: Make ladder globalranked again]]>
Blodir wrote:
Zoram wrote:
Blodir wrote:The broader scale rating is supposed to reflect, the more inaccurate it will be. This is emphasized by the fact we can choose our own scale, in other words we can choose what our own rating reflects.


indeed but we could also say that about including all kind of maps in the same rating. I mean someone might be very good on field of Isis an suck hard at water maps, (for example). Is the factor or playing a 1v1 so different to all other factors that it needs its own unique rating ?

This isn't a rethorical question, I'm actually asking.

I agree that it can be said about all maps/gamemodes, which is why I always wanted a matchmaker (with a 'proper' mappool) - something close to what Starcraft 2 has.


Someone who got my point ;)
Obviously, the more "divided" the rating, the more accurate it would be, with ultimately a separate rating system for each situation (per map, per number of players). Which would be quite impractical. So the line has to be drawn somewhere indeed.

Blodir wrote:
But yes, we've seen many examples of players getting high rating from 1v1, but while being fairly good, not playing even close to their immense rating in teamgames. They even admit to it themselves, and it probably makes their teamgame experience much worse.

Now I don't know exactly how trueskill works its magic, so I make assumptions here. I assume that a "global" rating system including both team games and 1v1 could put a different weight on each kind of game.

Similarly, does a player who only plays seton keep his rating increasing, or does TrueSkill considers that actually he is in effect only demonstrating skills on one map and diminishes the extra points he gets after each subsequent victory on the same map ?

I thought that -maybe- with the 1v1 ladder forcing you to play a large array of random selected map, rating there might reflect a bit better the "trueskill" of a player, as victories rely less on training for one map only.


on a side note about the user friendliness of it all.
I was quite confused as a new player by how this thing works. I plaid a few 1v1, had a rating of 300 (or whatever) and assumed this was my rating, so I join a teamgames matching that rank requirements, only to be told that no-no, I was 700 (global rating), which was the nearly invisible number calculated from 1 or 2 previous custom games. There might be a way to make it clearer to newbies what is the difference between global rank and 1v1 ladder rank. I also have a feeling that numbers don't represent the same skill level on both scales, with global ranking = 1v1 rating + 400 points. (totally not statistically relevant figure based on my own gutfeeling ;) )

Statistics: Posted by Zoram — 23 Nov 2014, 16:58


]]>
2014-11-23T16:16:52+02:00 2014-11-23T16:16:52+02:00 /viewtopic.php?t=8907&p=86444#p86444 <![CDATA[Re: Make ladder globalranked again]]>
Apofenas wrote:
I'll try one more time. Statistic is not always what is happening in real. I speak about what i see in FAF. You speak of what you can invent from statistic.


i am talking about using trueskill effectively.
you are talking about using trueskill in a slightly suboptimal fashion.
what you are really talking about though is trueskill+, that simply works better. but that doesnt exist (yet).
certainly suboptimal trueskill is not trueskill+.


Sure, B has more chances to win. And on theta it's slightly bigger advantage than on any good ladder map, due to map being 5x5. That's still not an advantage that will win a game.


A and B are not players, but 2 trueskill ratings. A is computed taking all 50 maps into consideration, B is computed
only from theta games.


I'll answer with same example. I saw a guy who played in ladder in GPG, he got to FAF, played a some of ladder games and got to team game on map he never played.His opponent had advantage of knowing map, build order and full mass deposits from middle. Do you think he won with that? Stat wise, it should be like this. But ladder player won in reality.


and the global rating of that ladder player increased after the win, just as it should.

Statistics: Posted by rootbeer23 — 23 Nov 2014, 16:16


]]>
2014-11-23T14:27:50+02:00 2014-11-23T14:27:50+02:00 /viewtopic.php?t=8907&p=86438#p86438 <![CDATA[Re: Make ladder globalranked again]]>
Blodir wrote:
Sir Prize wrote:Of course he would, but it would be funny when he drops 800 ratings points while fantasising about becoming a 3000 :P

And if he did spam ladder the system would figure out what he should be pretty quickly.

Gala didn't mean he would start playing ladder. He just meant that if ladder was rated again he would have higher rating opponents to farm rating from.



Yes, I was replying to Gorton's comment, my point was gala will lose a lot of epeen rating if we create a combined ladder due to being rated 1060 at 1v1. I'm not dumb enough to expect gala to try a new map or game type, was just explaining a hypothetical :)

Statistics: Posted by Sir Prize — 23 Nov 2014, 14:27


]]>
2014-11-23T12:10:33+02:00 2014-11-23T12:10:33+02:00 /viewtopic.php?t=8907&p=86434#p86434 <![CDATA[Re: Make ladder globalranked again]]>
Sir Prize wrote:
Of course he would, but it would be funny when he drops 800 ratings points while fantasising about becoming a 3000 :P

And if he did spam ladder the system would figure out what he should be pretty quickly.

Gala didn't mean he would start playing ladder. He just meant that if ladder was rated again he would have higher rating opponents to farm rating from.

Zoram wrote:
Blodir wrote:The broader scale rating is supposed to reflect, the more inaccurate it will be. This is emphasized by the fact we can choose our own scale, in other words we can choose what our own rating reflects.


indeed but we could also say that about including all kind of maps in the same rating. I mean someone might be very good on field of Isis an suck hard at water maps, (for example). Is the factor or playing a 1v1 so different to all other factors that it needs its own unique rating ?

This isn't a rethorical question, I'm actually asking.

I agree that it can be said about all maps/gamemodes, which is why I always wanted a matchmaker (with a 'proper' mappool) - something close to what Starcraft 2 has.

But yes, we've seen many examples of players getting high rating from 1v1, but while being fairly good, not playing even close to their immense rating in teamgames. They even admit to it themselves, and it probably makes their teamgame experience much worse.

Statistics: Posted by Blodir — 23 Nov 2014, 12:10


]]>
2014-11-23T10:08:01+02:00 2014-11-23T10:08:01+02:00 /viewtopic.php?t=8907&p=86431#p86431 <![CDATA[Re: Make ladder globalranked again]]>

And if he did spam ladder the system would figure out what he should be pretty quickly.

Statistics: Posted by Sir Prize — 23 Nov 2014, 10:08


]]>
2014-11-23T09:13:57+02:00 2014-11-23T09:13:57+02:00 /viewtopic.php?t=8907&p=86429#p86429 <![CDATA[Re: Make ladder globalranked again]]> Statistics: Posted by Gorton — 23 Nov 2014, 09:13


]]>
2014-11-23T08:56:09+02:00 2014-11-23T08:56:09+02:00 /viewtopic.php?t=8907&p=86428#p86428 <![CDATA[Re: Make ladder globalranked again]]>
galacticfear wrote:
I am in favour of this idea because then all the 1v 1 players will get some more rating for me to absorb allowing me to easily reach 3000 :D
The hilarious side-effect this idea will have on the ratings epeens of Setons specialists is the strongest argument to introduce it. Hint: http://content.faforever.com/faf/leader ... lacticfear


I'll try one more time. Statistic is not always what is happening in real. I speak about what i see in FAF. You speak of what you can invent from statistic.
That's exactly what you're doing by saying a combined ladder would be more valid than individual ones.

Basically your examples (and you can't generalise from case studies and hope to win an argument about population-wide effects) are due to people being briefly under rated when they haven't played yet and there's not enough data on them. The effect is gone once they start playing team games regularly. The idea of displaying both ratings next to eachother in a lobby solves this without the distortion you'd introduce by combining them.

Statistics: Posted by Sir Prize — 23 Nov 2014, 08:56


]]>
2014-11-23T07:44:03+02:00 2014-11-23T07:44:03+02:00 /viewtopic.php?t=8907&p=86426#p86426 <![CDATA[Re: Make ladder globalranked again]]>
Sure, B has more chances to win. And on theta it's slightly bigger advantage than on any good ladder map, due to map being 5x5. That's still not an advantage that will win a game.

I'll answer with same example. I saw a guy who played in ladder in GPG, he got to FAF, played a some of ladder games and got to team game on map he never played.His opponent had advantage of knowing map, build order and full mass deposits from middle. Do you think he won with that? Stat wise, it should be like this. But ladder player won in reality.

Statistics: Posted by Apofenas — 23 Nov 2014, 07:44


]]>