|
On February 17 2011 12:57 Oconomist wrote:
Any progress on that program that can analyze replays and pull data from battle.net profiles?
What exactly are you trying to look at? The historical data on the battle.net site doesn't go back very far. I have a program that parses out all my replays and gives me stats on my opponent, but I'm not sure how that might be extendable. I've thought about trying to write something that also scrapes match history, but because that page only displays the day, it's possible to have some weirdness where the numbers won't match correctly with the matches (since you can't get opponent name from the web site either).
|
On February 17 2011 09:42 random user wrote:Random stat of the day: In the past 29 or so hours, of the top 500 semi-active gold players in NA server, approximately 41 of them (in other words about 8%) were promoted into plat. This sample size does not include anyone outside of the top 500. Those 41 I found were put into 23 different divisions with the following distributions: + Show Spoiler +
Khalis Yankee: 16
Tal'darim Upsilon: 2
Sarengo Quest: 2
Tarsonis Upsilon: 2
Azimar Foxtrot: 1
Dominion Iota: 1
Gantrithor Bravo: 1
Roach Phi: 1
Sargas Indigo: 1
Amerigo Lambda: 1
Artika Uncle: 1
Phoenix Epsilon: 1
Taarim Romeo: 1
Araq Omega: 1
Broodling Lima: 1
Overlord Delta: 1
Reaper Chi: 1
Zamara Echo: 1
Reaver Sigma: 1
Aldaris Sierra: 1
Aiur Psi: 1
Sargas Upsilon: 1
Valerian Rho: 1
Useful to anyone? Not sure. I decided to post this to give people some idea of how frequent promotion is within the top 500. I hope it's not too far off-topic; it seemed pretty much within the realm of division tiers to me.
That's very interesting data. Could you do it for Platinums getting promoted to Diamond?
|
On February 17 2011 16:20 LetoAtreides82 wrote:
That's very interesting data. Could you do it for Platinums getting promoted to Diamond?
I last ran the plat promotion data about a week ago, and I just re-ran it. Unfortunately, I sort of hacked (ie cut and paste) code to scape the various leagues, and I had added storing the division after I did that split. So I'll go add that code back to the plats, and have something more comprehensive later today.
Looking at the stats, I did find something I thought was interesting.
So to clarify, this population is a snapshot from about a week ago of the top 500 semi-active platinum players at that time, comparing that to where they are now.
Of them:
1 became a master (looks like after going through diamond)
(This guy:
http://us.battle.net/sc2/en/profile/653839/1/ChinPoKoMoN/
if you are curious)
130 got promoted to diamond
3 got demoted to gold
1 got demoted to bronze
(This guy: http://us.battle.net/sc2/en/profile/1499852/1/Hazonko/
if you are curious)
Which is a lot more movement that I expected.
After some time passes I can compare the snapshot from now to a later snapshot, and I'll add the code that picks up the division too.
|
|
|
United States12240 Posts
On February 17 2011 18:25 random user wrote:Show nested quote +On February 17 2011 16:20 LetoAtreides82 wrote:
That's very interesting data. Could you do it for Platinums getting promoted to Diamond? I last ran the plat promotion data about a week ago, and I just re-ran it. Unfortunately, I sort of hacked (ie cut and paste) code to scape the various leagues, and I had added storing the division after I did that split. So I'll go add that code back to the plats, and have something more comprehensive later today. Looking at the stats, I did find something I thought was interesting. So to clarify, this population is a snapshot from about a week ago of the top 500 semi-active platinum players at that time, comparing that to where they are now. Of them: 1 became a master (looks like after going through diamond) (This guy: http://us.battle.net/sc2/en/profile/653839/1/ChinPoKoMoN/if you are curious) 130 got promoted to diamond 3 got demoted to gold 1 got demoted to bronze (This guy: http://us.battle.net/sc2/en/profile/1499852/1/Hazonko/if you are curious) Which is a lot more movement that I expected. After some time passes I can compare the snapshot from now to a later snapshot, and I'll add the code that picks up the division too.
What I think would be awesome would be a snapshot of every individual player (it would only have to be 1v1 even) with remaining bonus pool tracked. Then, an activity measurement based on intervals of 90 remaining bonus pool, and the league population distribution that reflects each step of 90. If we had that, we might be able to determine what exactly the activity threshold is based upon where the distribution is closest to 2/18/20/20/20/20 (supposedly Diamond and Master would be slightly above this). If this is something you could do, that would be excellent, since unfortunately it's beyond SC2Ranks' methods of information gathering.
|
On February 18 2011 02:52 Excalibur_Z wrote:
What I think would be awesome would be a snapshot of every individual player (it would only have to be 1v1 even) with remaining bonus pool tracked. Then, an activity measurement based on intervals of 90 remaining bonus pool, and the league population distribution that reflects each step of 90. If we had that, we might be able to determine what exactly the activity threshold is based upon where the distribution is closest to 2/18/20/20/20/20 (supposedly Diamond and Master would be slightly above this). If this is something you could do, that would be excellent, since unfortunately it's beyond SC2Ranks' methods of information gathering.
That's a pretty big task. I'm not sure how many 1v1 players there are, but given there are 523 pages of diamond players in NA atm, that's over 52k players just in diamond. Looking at LA though, that's only 22 pages of diamonds, which would be 2200 players or like maybe under 40k total population? That might be doable, but is the LA population a good sample to look at?
I'm also not sure how often I can hit Blizzard's site before they start to notice and care. That's the biggest reason I haven't made a web site or published any of my code or automated them... The scripts I run are fairly intensive (the diamond promotion script hits Blizzard's site over 200 times in under a minute for example without throttling). I'm sure that's not a big deal for their web servers, but I want to be a good netizen and not do it continuously.
Even looking at something like 60k people requires hitting their web site with about 1 page view per second... I don't think I'd want to do that without getting their permission first, and I have no idea how to go about that (I think I asked a long time ago when I created a Blue Tracker for WoW and they never got back to me).
|
On February 17 2011 18:25 random user wrote:Show nested quote +On February 17 2011 16:20 LetoAtreides82 wrote:
That's very interesting data. Could you do it for Platinums getting promoted to Diamond? I last ran the plat promotion data about a week ago, and I just re-ran it. Unfortunately, I sort of hacked (ie cut and paste) code to scape the various leagues, and I had added storing the division after I did that split. So I'll go add that code back to the plats, and have something more comprehensive later today. Looking at the stats, I did find something I thought was interesting. So to clarify, this population is a snapshot from about a week ago of the top 500 semi-active platinum players at that time, comparing that to where they are now. Of them: 1 became a master (looks like after going through diamond) (This guy: http://us.battle.net/sc2/en/profile/653839/1/ChinPoKoMoN/if you are curious) 130 got promoted to diamond 3 got demoted to gold 1 got demoted to bronze (This guy: http://us.battle.net/sc2/en/profile/1499852/1/Hazonko/if you are curious) Which is a lot more movement that I expected. After some time passes I can compare the snapshot from now to a later snapshot, and I'll add the code that picks up the division too.
If only 100 players get promoted to Diamond per week that's quite bad news for my terran account, it keeps hovering between top 100 and top 250 
Do you happen to know if it's possible for a player to get promoted to Diamond when their record is around 50% ?
|
On February 18 2011 04:06 LetoAtreides82 wrote:If only 100 players get promoted to Diamond per week that's quite bad news for my terran account Do you happen to know if it's possible for a player to get promoted to Diamond when their record is around 50% ?
That's 131 (including the masters guy) out of 500.
In other words, 26% of the people in the top 500 semi-active plat players at that snapshot were in diamond a week later.
The number of actual people being promoted is much higher. And I got promoted to diamond (a week or so after master's came out) with a 51% win average, give or take, so yes, it's possible.
|
On February 18 2011 04:11 random user wrote:Show nested quote +On February 18 2011 04:06 LetoAtreides82 wrote:If only 100 players get promoted to Diamond per week that's quite bad news for my terran account Do you happen to know if it's possible for a player to get promoted to Diamond when their record is around 50% ? That's 131 (including the masters guy) out of 500. In other words, 26% of the people in the top 500 semi-active plat players at that snapshot were in diamond a week later. The number of actual people being promoted is much higher. And I got promoted to diamond (a week or so after master's came out) with a 51% win average, give or take, so yes, it's possible.
Okay that makes me feel better, thanks. Getting to diamond is a lot harder now than it was back around the time the game came out lol.
|
United States12240 Posts
On February 18 2011 03:47 random user wrote:Show nested quote +On February 18 2011 02:52 Excalibur_Z wrote:
What I think would be awesome would be a snapshot of every individual player (it would only have to be 1v1 even) with remaining bonus pool tracked. Then, an activity measurement based on intervals of 90 remaining bonus pool, and the league population distribution that reflects each step of 90. If we had that, we might be able to determine what exactly the activity threshold is based upon where the distribution is closest to 2/18/20/20/20/20 (supposedly Diamond and Master would be slightly above this). If this is something you could do, that would be excellent, since unfortunately it's beyond SC2Ranks' methods of information gathering. That's a pretty big task. I'm not sure how many 1v1 players there are, but given there are 523 pages of diamond players in NA atm, that's over 52k players just in diamond. Looking at LA though, that's only 22 pages of diamonds, which would be 2200 players or like maybe under 40k total population? That might be doable, but is the LA population a good sample to look at? I'm also not sure how often I can hit Blizzard's site before they start to notice and care. That's the biggest reason I haven't made a web site or published any of my code or automated them... The scripts I run are fairly intensive (the diamond promotion script hits Blizzard's site over 200 times in under a minute for example without throttling). I'm sure that's not a big deal for their web servers, but I want to be a good netizen and not do it continuously. Even looking at something like 60k people requires hitting their web site with about 1 page view per second... I don't think I'd want to do that without getting their permission first, and I have no idea how to go about that (I think I asked a long time ago when I created a Blue Tracker for WoW and they never got back to me).
It would be literally 100x more work than SC2Ranks does on a single server and bracket =) I don't know if LA would be a quality sample because of the anomalies we've witnessed in the past, but maybe as a preliminary run we can notice some trends? If it's still inconclusive we would need NA data though.
I can't really estimate their web server traffic, but their bandwidth alone can certainly handle it (they were saying that for their WoW world server, their bandwidth load never exceeded 5% of its capacity even when they uncapped the connections on the client end).
|
LA doesn't have anomalies, it just can show things that it's hard to see on bigger servers ^^ Taiwan has anomalies!! :D
You don't have to do a 100% pass either. You can skip silver if you wish, as it should have as much active players as gold and platinum anyway, just for having gold and platinum it should be fine.
Then you can try SEA server next and confirm/deny/refine what we can grasp from LA. ^^
|
United States12240 Posts
On February 18 2011 05:13 SDream wrote:
LA doesn't have anomalies, it just can show things that it's hard to see on bigger servers ^^ Taiwan has anomalies!! :D
You don't have to do a 100% pass either. You can skip silver if you wish, as it should have as much active players as gold and platinum anyway, just for having gold and platinum it should be fine.
Then you can try SEA server next and confirm/deny/refine what we can grasp from LA. ^^
I would want Silver info, but especially Bronze. The most important part is finding the threshold where Bronze is around 20%. That also impacts the other leagues in turn, so Silver would be just as important.
|
2 gold divisions on LA server confirmed.
I doubt that there isn't 2 tiers on NA and any other server.
A player has joined this division: http://us.battle.net/sc2/en/profile/117477/2/HenX/ladder/21231#current-rank
at the date of 2/182011, while the newest division still have only 16 players.
So, yeah, confirmed.
I'll ignore silver and look at bronze soon. Silver has probably as many tiers as platinum anyway...
Excalibur, do you know why are there 2 divisions listed in sc2ranks (on LA bronze) with 0 players? One of them is listed on sc2ranks as having 20 players though, so was there players there once? o.o
|
United States12240 Posts
I've seen that happen before, SC2Ranks lists them as "dead divisions" but I don't exactly know what that means. There doesn't seem to be a way to translate a division ID to a Battle.net URL so I doni't think this is something we're going to know. I've just ignored those "dead divisions" as anomalies but I'll ask Shadowed about what creates them.
|
Sorry I've been busy at work:
Here are the diamonds. Snapshot is from whenever I posted last about plats getting into diamond till now:
27 promotions to diamond. 1 demotion to gold
Again the sample size is the top 500 by points of semi-actives in plats from a snapshot a day or two ago.
Division distribution:
+ Show Spoiler +
Braxis Omega: 12
Azimar Charlie: 2
Auriga Zeta: 2
Infestor Whiskey: 1
Tassadar Juliet: 1
Phash Gravity: 1
DuGalle Indigo: 1
Templar Eta: 1
Queen Iota: 1
Azimar Romeo: 1
Kelerchian Hawk: 1
Kalathi Indigo: 1
Hyperion Dixie: 1
Warp Prism Mars: 1
|
In regards to spidering, I still think it's a good thing to do to get permission before spidering over 50k pages per day, ever day, for an extended period. If you have any idea how I can get actually get that permission, I can see how much of a load it takes on my end.
|
United States12240 Posts
Well all we would need is one solitary snapshot, it wouldn't have to be continuous like SC2R. Regardless, I'll forward along the permission request.
|
So I ran the bronze top 500 semi active players by rank to see what happened to them. Snapshot is about 25 hours.
The first thing I noticed was there is much less volatility within the top 500. The top 122 players moved around spots, but the first newcomer wasn't until spot 123.
To compare, though the plat snapshot is 2 days instead of 1, there are new people at spots: 2, 15, 19, 23, 25, 30, 33, 35, 37.. the list goes on and on throughout the top 500.
11 people were promoted within this snapshot. 10 went to silver. 1 went directly to gold.
Since I took a look at the plats to compare to the bronzes, here are their stats as well. Snapshot for this is about 36 hours, give or take.
47 promotions, all to diamond, in these divisions:
+ Show Spoiler +
Archon Romeo: 17
Antioch Sierra: 2
Dropship Mars: 2
Kerrigan Lima: 2
DuGalle Epsilon: 1
Moria Epsilon: 1
Nagglfar Victor: 1
Prelate Iota: 1
Corruptor Alpha: 1
Moria Omega: 1
Anselm Indigo: 1
Tarsonis Alpha: 1
Warp Prism Kappa: 1
Ladranix Alpha: 1
Moratun Omicron: 1
Dylar Indigo: 1
Tarsonis Charlie: 1
Duke Xi: 1
New Folsom Alpha: 1
Khalis Foxtrot: 1
Augustgrad X-Ray: 1
Muadun Pepper: 1
Norad Sierra: 1
Jorgensen Theta: 1
Halcyon November: 1
Halcyon Kappa: 1
Arbiter Lambda: 1
Dragoon Whiskey: 1
|
My terran account Khrone on the NA server is at 3017 points in platinum, 396-390, which'd make me currently in 9th in the NA server, and still not promoted, the system makes promotions after many games way too hard.
UPDATE: Finally got promoted. My stats after promotion were 2843 398-393 in the diamond divison Taarim NU, an old division.
|
On February 21 2011 04:39 LetoAtreides82 wrote:
My terran account Khrone on the NA server is at 3017 points in platinum, 396-390, which'd make me currently in 9th in the NA server, and still not promoted, the system makes promotions after many games way too hard.
UPDATE: Finally got promoted. My stats after promotion were 2843 398-393 in the diamond divison Taarim NU, an old division.
From what I've observed, the adjusted rating needs to be over 250 points to go from platinum to diamond. That doesn't mean eclipsing it will promote you, but your 3017 - 2767 or 2768 = 250 or 249, which is right at that number. This makes sense since you were at the time one of the highest rated platinum players in the world.
If you can PM me your character ID i can take a closer look with your match history. Or you can IM me in game since I'll be on NA for the next few hours.
edit: saw your post was 5 hours old, which would put your bonus pool at 2767-8.
spoilered to avoid clutter
+ Show Spoiler +Okay after talking with you and seeing your match history, here is what your data looks like: ![[image loading]](http://imgur.com/DSKfx.jpg) The Yellow on the left is using the data you posted. The light blue represents bonus pool numbers I guessed at. Since you played actively, it means there could be some skewing by +/-1 or 2 points somewhere. del-B is the change less bonus pool skewing for wins. P+U-B = adjusted rating (Points + unspent - total bonus) dMMR = displayed MMR From the limited samples I've looked at regarding specifically plat to Diamond promotions, 250 adjusted rating seemed to be roughly the observed area for your moving average. Clearly your data shows 225, which I find interesting and I'll look at some of my other samples (namely my own, since I was just promoted from plat to diamond on SEA and have a good sample). Thanks for sharing and congrats on your promotion.
|
|
|
|
|
|
EPT
)
