|
For a long while I have had an interest in the hidden mechanics of the matchmaking and ladder system in SC2. I've been in Gold league and below since release except for a brief stay in Platinum. At this point, on my main account, I've played about 3000 1v1 games since release.
Anyway, I've found myself wondering if my having such a long game history has an impact on how easy or difficult it is to move my MMR. Now, I should point out that the people I'm playing on my main account are a mixture of Gold and Platinum players, and I'm winning about 50/50. I'm not getting matched with Diamond players very much. So, I am striking out with an experiment.
I dug out my WoW Cataclysm collector's edition and found the Starcraft guest pass in there, made a new account, and played 5 placements on a brand new account.
I won all 5 and placed in Diamond.
Here's the experiment: I'm going to try to focus my highest quality play on this new account. I'll play one game on the old one to warm up, and play a maximum of three or four in a row on the new one, then if I want to keep playing I'll switch back to the old account.
If the ladder system is working as advertised, over time I should fall to middle-platinum to high gold, where my old account is. On the other hand, if MMR is not in fact a great predictor of future game performance, anything could happen!
My existing account:
![[image loading]](http://i.imgur.com/OGRTE.jpg)
http://us.battle.net/sc2/en/profile/472254/1/Lysenko/
My new account:
![[image loading]](http://i.imgur.com/c6MaP.jpg)
http://us.battle.net/sc2/en/profile/3243569/1/Lysenko/
Why am I doing this?
It's entirely an investigation of how the ranking system deals with a new account vs. one with a long history. I'm not one of those people who believes that "It placed me in league A but I should really be in league B." I'm acutely aware of the flaws in my play and don't have an ego about it.
What am I hoping to get out of it?
One likely outcome is that as I play a lot of games I end up right in the same spot in the ladder with my new account as with the old one. However, getting there can be a benefit to me. By placing in Diamond to start with, I am getting matched against better players and can possibly learn more from my losses.
On the other hand, if the ladder system is really bad at distinguishing between Gold and Diamond players, it's possible I'll settle at a stable place in Diamond, or maybe high-ish Platinum. If this is the case, then my long history working my way up from Bronze to Gold over about 3000 games on the old account might actually be holding back my ladder status. In this case, I'll wind up with a higher-rated account to play, and hopefully have better, more challenging games that push me farther as a player.
A third possibility is that the ladder system works just as advertised, but as I lose a lot of games on the Diamond account, I learn enough to bring my performance level up to, say, Platinum, and stay there. If this is the case, going back to play on the Gold account after the new one has stabilized should get me promoted!
So far, on the new account, I've played four games and won 1 of them. Two were against diamond players and two were against top-end Platinum players. All of them were pretty close in terms of economy, so I'll have to play a lot more to know where this is going. Counting my placements, this means I'm currently 3-3 against Diamond and high Platinum opponents.
Edit: I should also say that my impression based on the first four games is that there's not a gigantic difference between high Gold and low Diamond. It feels like a matter of small advantages and slightly better exploitation of them. Getting my own play to that level certainly feels doable.
Edit 2: I am saving all of my replays and if anything even remotely interesting happens with this experiement, I'll post them.
Edit 3: Every game I play, on each account, is a straight-up game. I do sometimes get accused of cheesing when I aggress early, but I'm certainly not doing anything like 6-pooling every game or anything like that.
  
|
hmm interesting though i doubt you will get anything from it
|
I've. Noticed the same thing that is apparent in this system I had a second account given to me. My original account was top 8 plat this new account put me in diamond playing low masters. My original acc has 800 wins on it the new one has about 25. So yes the variation is more likely to swing to higher extremes given the smaller amount of data it has to gather info from.
|
On November 20 2011 14:59 BobTheLob wrote:
hmm interesting though i doubt you will get anything from it
I'll certainly get some games vs stronger players than I've been facing! Even if I lose a huge majority of them, the replays will give me something to study in figuring out what I should be working on.
|
Baltimore, USA22256 Posts
Good luck. Like you said, even if you don't really learn anything, it's still fun to poke & prod. 
|
This is really interesting. I kinda feel like you'll level up or down as soon as you started to get a good streak going and then it'll go fast. Just look at the win traders with thousands of games in lowest possible bronze getting up to master in a hundred games or so.
|
On November 20 2011 15:30 Eatme wrote:
This is really interesting. I kinda feel like you'll level up or down as soon as you started to get a good streak going and then it'll go fast. Just look at the win traders with thousands of games in lowest possible bronze getting up to master in a hundred games or so.
If you can continue to win a large majority of your games for 30-100 games in a row, you'll certainly get promoted. However, that doesn't necessarily mean that there aren't MMR ranges where differences are more arbitrary. For example, suppose you had an MMR system overlaid on a game that involved flipping a coin to see who wins. You'll get a Poisson** distribution of MMRs, but nobody has an actual advantage over anyone else, and it's mainly one's history that determines where one lands in that distribution.
I'm not saying SC2 is like that. It's a very skill-intensive game, and I'm certainly systematically worse than most Diamond players. However, it's possible that it's more of a coin-flip than it seems, in certain skill ranges (particularly those where people know to scout and respond to certain approaches and not others.)
** I originally said the coin-flipping game would have a normal distribution, but it would have a Poisson distribution, since your score at any given time depends on the flips that came before.
|
No matter what happens, after you're guest pass expires, you'll have games against higher-level play under your belt, and you'll be able to take that into your play on your main. And is there really not that big of a difference? I would imagine that to be in a place like High Diamond-Low Masters, but in a place like High gold to low diamond? That's interesting
|
On November 20 2011 16:15 Latrommi wrote:
No matter what happens, after you're guest pass expires, you'll have games against higher-level play under your belt, and you'll be able to take that into your play on your main. And is there really not that big of a difference? I would imagine that to be in a place like High Diamond-Low Masters, but in a place like High gold to low diamond? That's interesting
Well if my rating continues to linger higher than my main's, I probably will buy another copy of the game to keep the account going for a while.
As for how far off high gold to low diamond is? It's too early to say what the trend is, but the games don't feel that uneven, and the score screen looks like it agrees.
My plan is to collect a bunch of data and use some of the metrics people have already worked out to evaluate.
|
I'm sorry to do this to you, but at our pharmacy, we service a patient with the last name Lysenko.
|
This sounds quite interesting, and I would love to see your MMR from previous games holds you back
On a side note, I used to be in bronze and that does not seem to have any effect now, since I just got promoted into diamond (Although I have only played about 1/10 the number of games that you have). It might just be that your MMR can find really close matches to you on your original account and thus your uncertainty factor will not increase as much.
I also didn't know that I could actually use that stats class I took to rationalize the match making system ^ ^
P.S. May want to edit those pictures so that your real name does not appear. Would hate to see you get trolled.
|
On November 20 2011 18:13 inbox24 wrote:
I'm sorry to do this to you, but at our pharmacy, we service a patient with the last name Lysenko.
That's interesting.
I chose my name because Trofim Lysenko was an early Soviet agronomist who was an advocate of Lamarckism, a discredited theory that creatures could develop inheritable traits in response to their environments. He became a prominent member of the Soviet Academy of Sciences and presided over purges of biologists who believed in Mendelian genetics which resulted in them being imprisoned or killed.
The threatening, Stalinist overtones to the name, as well as the Zergy idea of "evolving" traits in response to what's going on in the game, made me choose Lysenko as my name for SC2. There's a link to the Wikipedia entry on Lysenkoism in my signature.
You should ask the patient whether he or she is related to Trofim Lysenko.
|
On November 20 2011 18:54 Peanut Butter wrote:
P.S. May want to edit those pictures so that your real name does not appear. Would hate to see you get trolled.
My real name has been out there from previous blogs on TL as well as other activities in the community and my professional life, so I doubt this will be much of a problem. As for the rest of your comments, thanks for your thoughts!
|
"I've played about 3000 1v1 games since release."
Wait how? i've played 80 1v1's and im diamond^^
|
On November 21 2011 17:00 Flippin1337 wrote:
"I've played about 3000 1v1 games since release."
Wait how? i've played 80 1v1's and im diamond^^
You've probably played some other RTS before SC2? My pre-SC2 RTS experience was about four hours of WC3 and SC1 single player, total. Also, playing Zerg, having no low-tier anti-air made things a lot harder when I was just learning the game.
I get the sense, based in part from the few first games of this experiment, that the skill gap between high Gold and mid-Diamond has been compressing. Six months ago, I'd play a Diamond league friend and just get completely destroyed by their superior macro. Today, my games vs. Diamond players are much closer. I won't say an even match, but that's partly what this experiment is to test.
My feeling is that this makes sense to some extent -- different people learn at different speeds, but at a certain point, all you have left on the ladder are the people who genuinely have a continuing interest in Starcraft, and who know the basics. If you think about the really basic skills of building an army and making semi-reasonable tech choices, at release maybe 20% of people had those skills tops (from beta or lots of SC1 experience) and today maybe 70% have them, with speed and quality of decision making distinguishing the leagues.
To put it another way, as the overall skill level rises and Gold league tech timings approach optimal, it becomes a lot more likely for a slightly less-skilled player to take a game off a more-skilled player with a solid timing attack or something sneaky.
|
Update on this experiment. After playing several more games on the diamond league account, I'm now 1-7! This is certainly consistent with being highly over-placed and MMR being reasonably predictive of even games, although it will require more games to be sure, since I find my gold account tends to be streaky as well.
I still feel like the games themselves are not THAT uneven. It's more just that I steadily fall behind by enough to matter.
I've spent a couple nights playing on the gold account just to get some games in and stay thinking about the game while I consider what I can do to improve.
|
Final update: I didn't play on the new account very much at all. However, at the start of S5, both accounts placed into Platinum.
|
|
|
|
|
|
EPT
