EPT
).Why the Blizzard ladder is great - Page 14
| Forum Index > SC2 General |
|
Elothis
111 Posts
). | ||
|
[F_]aths
Germany3947 Posts
On April 10 2011 05:25 protoss_machine wrote: The win ratio does not indicate how good you perform. Why would you like to know the win ratio?I think I have a right to know what my win/lose ratio is whenever I want to. It was fine before the patch... On April 09 2011 19:31 Aequos wrote: This is true.As I'm certain has been stated many times, the W/L ratio really doesn't mean a thing if you aren't at the top or bottom of the ladder. At the top of the ladder (masters) it cannot find opponents who can beat you reliably, so you acquire a win rate above 50%. At the bottom of the ladder (bronze) it cannot find opponents who will lose to you reliably. As such, you'll acquire a win rate below 50%. On April 09 2011 19:31 Aequos wrote: This however is not true. Your league depends on your MMR, but the MMR is independent from your league. MMR determines the skill level of your opponent, the confidence interval determines the skill range around the MMR where the match-maker looks first. (You still can get better or poorer opponents.) The confidence is also needed to know when you skill plateaus and enables the system to determine the tier of the league (the hidden division modifier) when your are moving up or down.If you go on a winning streak in a mid-league, the system re-evaluates where it believes you should be. In the short term, it will try and ensure your W/L is 50% (as opposed to overall). For example, if you win 10 games in a row in Platinum and get promoted to Diamond, chances are your W/L ratio will always be n+10/n (or at least until you have a bad lose streak). Every new season you are replaced anyway. Chances are that you stay in your league but get into another tier. Promotion and demotion does not have an influence regarding the search for the opponent. On April 09 2011 19:31 Aequos wrote: This is true only for the lowest tiers of bronze players. Most bronze players also get 50% after they play for some time.Honestly, although it is annoying, they really don't hurt anyone by removing it. It's fairly obvious to a player when they're actively playing if they've been winning a very large portion of their recent games (and hence improving). Masters players (the top of the ladder) can already see their W/L. The only players this really effects are Bronze players who actually cannot win 50% of their games. | ||
|
youngminii
Australia7514 Posts
You have completely changed my opinion of the ladder system. You raised a lot of interesting points. | ||
|
Fidd
Canada14 Posts
| ||
|
Lysenko
Iceland2128 Posts
On April 09 2011 19:22 [F_]aths wrote: Actually, it does try – over time. That's simply not the case. I took a break of a couple of months and returned to a 17-game losing streak. My MMR stabilized after about 7 net wins over losses, and at that point I was going 50/50 in new matches, but I never made up those extra 10 games lost -- my losses exceeded wins by about 10 games plus or minus a few until the end of Season 1 about eight weeks later. The matchmaking system attempts to make the chance of the *next game* 50/50. If there were excess wins or losses from streaks in the past, they simply don't have an impact on matchmaking. | ||
|
Lysenko
Iceland2128 Posts
On April 10 2011 21:42 [F_]aths wrote: will be 50% regardless (+- random noise) with enough games played. To put my earlier comments on this another way: nonrandom streaks (that is to say, streaks with a cause, like being initially rated very low or high, or a loss streak from a break) also add deviation from the 50% statistic. It is important to note that deviations from 50% can persist indefinitely because the system attempts to find 50/50 short-term matches, a tendency which encourages the persistence of win/loss streaks from the past in the statistics. Win/loss ratio, of course, would be a good way to estimate one's skill standing in the population if matching were completely random and without regard to skill. | ||
|
[F_]aths
Germany3947 Posts
On April 11 2011 18:43 Lysenko wrote: Yes.The matchmaking system attempts to make the chance of the *next game* 50/50. Blizzard did not implement a bad habit like "oh you need a win to get 50%, lets look for a noob" or "you naughty boy already have >50%, lets look for a pro to crush you." I hope that I was not misunderstood that way. Battle.net just looks for a worthy, yet beatable opponent like you said. Almost every player sometimes has a cause for a nonrandom loss. Those freewins are spread, you will likely get some of them. Even without this, it gets to the same outcome if you play enough games. In the end, your 10 "net losses" will be not discernible from random statistic noise and get meaningless. There will be no way to tell if you still have 10 net losses or if it is random coincidence. The greater the number of total games played, the greater the probable absolute difference from exact 50%, but the relative deviation should tend to get closer to 50%. On April 11 2011 18:51 Lysenko wrote: I agree when you count in intentionally thrown games. You could throw 1000 games in a row and than ladder up to anywhere while still having a horrible win ratio.It is important to note that deviations from 50% can persist indefinitely because the system attempts to find 50/50 short-term matches, a tendency which encourages the persistence of win/loss streaks from the past in the statistics. Otherwise, the win-loss-ratio is a zero-sum-game. There are reasons for a loss streak, but also for a win streak. Other players have loss streak reasons, too. You will probably get some easy wins you normally should not get, but the match making system could not know that your opponent really had a bad day, he is ill and tired and playing his first game after being 6 month in a coma. Overall, it gets you to 50% (within bounds of random noise) even when there are nonrandom causes. | ||
|
caruso
Germany733 Posts
On April 07 2011 00:41 oGsTheStD wrote: they should make it optional (only player can see it not others), but have the default show no losses. That way if your troubled enough to want the losses you will find it, but you can also remain blissfully ignorant QFFN (Quoted for fantastic name) | ||
|
vek
Australia936 Posts
While hiding stats might seem like a good idea to get people involved I don't think we are seeing much difference in percentage of players who ladder vs those who just play custom games compared to WC3. The percentage of people who actually ladder is still very low. I see ladder as the competitive component of the game. Some people are just not interested in competition - they only want to have fun. The lack of stats just gave me the feeling that the whole thing was fake so I found it extremely hard to care and take pride in my "rank". I never had any concrete way of telling if I improving and lost all motivation to play ladder games. If you could see peoples MMR or even filter by MMR it would be fantastic for setting up custom games to practice against people of even skill. The matchmaking system is absolutely great at matching you against someone of relatively even skill. Blizzard really did a great job there. It's just a real shame the ladder has so little information and is still impossible to browse. | ||
|
WindCalibur
Canada938 Posts
I bought this game right after it was released. I was a C zerg on iccup before and I was really excited about sc2. Being sort of experienced with RTS games and have pretty good mouse accuracy (played a game called osu! where I am probably the top Canadian player), I was able to get into platinum, promote to diamond, and have a 60-70% win ratio. However, even then, I felt uncomfortable about my skill. Unlike BW, I had no build order, weaker mechanics, and was lost in a lot of situations. Then I stopped playing for a long long time until only roughly a month ago when I started to ladder again. I came back with 2300 bonus pool still in my league and felt like improving and having a good time. I got stomped so hard at first for some reason that it was disgusting. I remember going from my previous win ratio to around 40 percent win ratio having streaks of 10 loses at occasions. But that didn't really bother me at all because I only cared about improving rather than my win ratio. I kept playing and eventually started winning again, ending up with a near 50% win ratio near the end. Being a BW player, I knew that mechanics meant everything, and that playing standard, improving mechanics, was the way to improve. Yes there were times where I would get cheesed every game and lose often but I kept trying to improve so I can become cheese-proof. Then ladder reset and after I started winning, I wanted to see how I can promote into masters. I read threads about MMR and learned that in order to be in masters, you probably have to beat masters. This worried me at first because I was only versing diamond players, but I know that if I just keep improving via mechanics then I can make it. Watching replays after loses, following strict build orders, scouting properly, macroing and watching pros etc, allowed me to continuously vs stronger players and improve. At first, I was versing mid-diamond, then it was high-diamond, eventually masters. Sitting at around 100 points above rank 2 in my division and consistently beating masters, I finally saw the promotion sign and got into masters. After all that, I just have to say that it is your own responsibility to improve. If you are truly competitive and want to be good, screw win ratios and instead focus on improving. There are tons of threads on Teamliquid on how to gauge your improvement and how to get better. Doesn't matter if you are a bronze or diamond, if you want to improve, why do you even care about win ratios? You should only feel good when you know you are improving and are good rather than having a beasty win ratio against weaker opponents. | ||
|
Lysenko
Iceland2128 Posts
On April 11 2011 19:17 [F_]aths wrote: But this eventually gets to the same outcome if you play enough games. In the end, your 10 "net losses" will be not discernible from random statistic noise and get meaningless. There will be no way to tell if you still have 10 net losses or if it is random coincidence. The greater the number of total games played, the greater the probable absolute difference from exact 50%, but the relative deviation should tend to get closer to 50%. Admittedly, if a 10-game losing streak were the entire deviation from an otherwise 50/50 record, after 100,000 games it would be lost in the noise. The problem is that each individual game is not statistically independent of the one preceding it. I believe that if you analyze a long period of any player's games, you'll find that a loss is more likely to be followed by a loss than 50%, and a win is more likely to be followed by a win. There are several practical reasons for this: * A losing player's more likely to get upset in a way that screws up their play in subsequent games. * A fundamental change in strategy or insight can yield a winning streak that's sustained until the player settles in at a new MMR level. * Players at a given MMR are more or less challenging at certain times of day and days of the week, and multiple games at a given time are likely to lean toward winning or lean toward losing for this reason. * Playing multiple games under transient conditions that affect results, like during network problems, on a laptop while traveling, or when playing conditions like heat/cold are more or less optimal will tend to introduce correlations between games. Because there are game-to-game correlations in results, convergence on the mean happens a LOT more slowly than for a statistically-independent random process, and under some conditions (say, a player who improves very rapidly over a long period of time) there will not be a convergence on the mean at all. I should point out that I agree with the basic premise that win/loss ratios are not meaningful in evaluating skill, but my point is that they lack meaning not because they always converge on 50%, but because they deviate from 50% for reasons that may have nothing to do with a player's actual skill at the game. | ||
|
naggerNZ
New Zealand708 Posts
While it would be nice to have, for those who understand how the ladder works, the reality of it is, probably 90% of players don't fully understand how the ladder works, and adding irrelevant information just fuels complaints and confusion. | ||
|
Fryght
Netherlands254 Posts
It just doesn't make much sense o_O Also, I got plenty matches against Diamond players, while we were in bronze league, with my 2v2 AT team. Really makes me wonder...is the MMR system like "These guys are on a win streak, time to crush them. Let's hit them with players that are 4 leagues higher than them!". Hell, I've even played against Master league players, when being nowhere near their league. | ||
|
naggerNZ
New Zealand708 Posts
This is how I understand it to work, however I may be wrong. | ||
|
Swwww
Switzerland812 Posts
| ||
|
DisneylandSC
Netherlands435 Posts
Also I find myself trying out the other races and trying new / different strategies. Also because of MMR your win / loss will be about 50% anyhow so who cares about the actual number. Also the matchmaking is awesome. 1v1 in BW was terrible imo, it was almost always a one-sides slaughter fest. Now I get so much interesting and fun games. But yeah, some people are just afraid of change. | ||
|
[F_]aths
Germany3947 Posts
On April 11 2011 19:41 Lysenko wrote: Much fewer games are sufficient to make a deviation of 10 indiscernible to noise. I will try to figure out how to calculate the amount of games where a deviation of 10 cannot be told from noise. (From intuition, it is 100 or 200 games.)Admittedly, if a 10-game losing streak were the entire deviation from an otherwise 50/50 record, after 100,000 games it would be lost in the noise. edit: Now I believe its more like 2000 games you need to have to explain a deviation of 10 with pure noise, I still try to get the formula. I am confident though the numer of games needed is lower than 10.000. Probably much lower. On April 11 2011 19:41 Lysenko wrote: I made the same observations. For example if I play Sunday evening, I feel I get much worse opponents than other days. Recently I got some insight to zerg play which got me to a winning streak, but then I got opponents who just outplayed me. The problem is that each individual game is not statistically independent of the one preceding it. I believe that if you analyze a long period of any player's games, you'll find that a loss is more likely to be followed by a loss than 50%, and a win is more likely to be followed by a win. There are several practical reasons for this: + Show Spoiler + * A losing player's more likely to get upset in a way that screws up their play in subsequent games. * A fundamental change in strategy or insight can yield a winning streak that's sustained until the player settles in at a new MMR level. * Players at a given MMR are more or less challenging at certain times of day and days of the week, and multiple games at a given time are likely to lean toward winning or lean toward losing for this reason. * Playing multiple games under transient conditions that affect results, like during network problems, on a laptop while traveling, or when playing conditions like heat/cold are more or less optimal will tend to introduce correlations between games. On April 11 2011 19:41 Lysenko wrote: Nonrandom reasons work both ways. When you came back from a break, you essentially gave free wins to others. But it is likely that you sometimes get easy wins from other players, too. Most of the times, you don't know that. You cannot tell nonrandom reasons of other players from noise.Because there are game-to-game correlations in results, convergence on the mean happens a LOT more slowly than for a statistically-independent random process, and under some conditions (say, a player who improves very rapidly over a long period of time) there will not be a convergence on the mean at all. I should point out that I agree with the basic premise that win/loss ratios are not meaningful in evaluating skill, but my point is that they lack meaning not because they always converge on 50%, but because they deviate from 50% for reasons that may have nothing to do with a player's actual skill at the game. I normally fuck up the first game I play in a session. I forget drones, get supplystuck, don't click at the right positions and so on. The probability I lose that game is >50% regardless of my current MMR or win ratio. But that loss lowers my MMR, I probably get a somewhat easier opponent the next game so that I can probably win. Such short-term nonrandom causes are evening out very fast. Every player is affected by the same MMR formula. Let's say you rage after a loss and will lose the next game with 55% probability versus someone with same MMR. But you may be now get an opponent who also just lost, so it evens out. Longer-term causes (like taking a break from Starcraft for months) of course have a greater impact. It still evens out since other players take breaks from the game, too. You're right of course that the AMM tries his best to offer 50% win chance for the next game, it does not consider the current win ratio to skew the player search. | ||
|
Sanchez_
Australia40 Posts
I really think they should show losses for diamond at least though maybe even platinum. If i was high diamond it would feel to me like they were disrespecting my achievement still not showing losses the same as people in bronze. | ||
|
BeMannerDuPenner
Germany5638 Posts
1. ladder outside of masters isnt competive . a LADDERs whole point is competive play. this is always the case at all levels. if its not for competive play then get rid of points/rank/leagues as well and just make it a huge matchmaking pool. but that isnt a ladder then cause its not competive right? also i know gold players that fight for evry point/rank and constantly compare themselves to others. and your "bob" does too since apparently he cares about getting to the top of his division. 2. winrate is the end all what people care about (as in the "guy scared to play cause of his 51% winrate" example). if people are positively affected by that change they already had the wrong mindset. if someone is scared of ladder cause of demotion, losing points/rank etc it doesnt change ANYTHING. i dont get how you can talk about "bob" caring about getting top8 in his division but at the same time completly ignore all the ladderfear connected to losing ranks/points. it makes no sense. what do you think will piss "bob" more off? going from 51% to 50.89% winratio or losing 17 points that make him lose 2 ranks and puts him further away from his top8 goal? 3. ladder is/should be a playground it shouldnt and only is as a result of the poor custom game system. its almost impossible to find a custom game where the players are evenly matched for obvious reason. if we could name our games ("1v1 master 3k+ join!" ) / have a nonladder matchmaking/sort them by skill of the player in the lobby etc there would be no need to wash down the competive aspects of the ladder since evrybody could play all day fine without even touching the ladder. overall i plain disagree with taking away the most basic statistical info from 99% of the playerbase (yeah im masters but only diamond in most team brackets) just because some people focus on the wrong things. i still see it as a huge slap in the face for evryone with a somewhat competive mind below masters and a immensly stupid thing to do. bronze only? maybe. optional and made standart setting so the super noobs wont even know they can change it?sure why not. but forcing this on almost evry player out there is just a big no. | ||
|
[F_]aths
Germany3947 Posts
On April 11 2011 20:36 Sanchez_ wrote: In Diamond (and even in the lower and middle regions of master), the win ratio will be around 50% when you play enough games. It is pointless to show the losses in Diamond. It only leads people to consider the win ratio – while that ratio says nothing about their skill. Cutting the losses out, allows Diamond league players to focus on the game, not on a statistic value which does not tell you anything about your skill. If you check the league of the opponents you beat, you get an impression where you really are.I agree that the ladder generally works well. I really think they should show losses for diamond at least though maybe even platinum. If i was high diamond it would feel to me like they were disrespecting my achievement still not showing losses the same as people in bronze. Blizzard displays the win ratio for all masters because some of the top master players, who will not be able to get ranked to Grand Master, are still so good that the AMM cannot properly work in time and therefore matches them against a weaker opponent to finally get a game. | ||
| ||
