|
On January 07 2013 05:58 Jomppa wrote:[spoiler] ![[image loading]](http://i.imgur.com/4cZ7K.png) [spoiler] What happens that time?
The data shown has been "smoothed" so that bronze is never higher than silver, silver is never higher than gold, and so on. In the raw data, for very short game durations where we don't have a lot of data, sometimes bronze would have a high SQ by luck. So the other league SQs were adjusted upward to bronze level.
If there's interest I'll add a button to the page so people can see the Raw data as well as the Smoothed data. The Smoothed data is what GGTracker uses to compute the Spending Skill for each player in each game. For that purpose of translating an SQ into a league, it's important that Silver not be lower than Bronze, etc 
|
On January 07 2013 03:27 OrbitalPlane wrote:This tool is really great. When i was reading the OP i thought this must be a fake. Until now everyone told me that it is impossible to read out the spending quotient from replays. This webside is indeed awesome and real so! I can't believe i missed it until now. This project needs/deserves a lot more attention. Some suggestions i would love to see implemented: Please add the actual numbers to the spending graph. That the most important feature to me. The league symbol is not very accurate and the league SQ could possibly change overtime. I want to keep track of my SQ in absolute numbers. Filter out non sc2 games. I have acutally tower defenses and other custom games in my profil. :D Easy solution would be to be able to manual delete games. allow several bnet accounts per ggtracker account. A lot of people have smurfs or accounts in different regions. A chrono boost graph similar to the injection graph  Big thx to the creator(s) of this tool!
Hi OrbitalPlane thanks! We are reading Spending Quotient not from the replays, but from the s2gs files.
Chrono boost graph is coming! And something for Terran too.
We graph Spending Skill rather than SQ because Spending Skill is more accurate -- it takes into account the race and game duration also, two factors which we know have a major effect on SQ. However I appreciate that an actual number sounds more precise, and you make a good point that league SQ could possibly change over time. Also, I believe the customer is always right! If several GGTracker Pro subscribers request it, we will add an SQ graph.
|
On January 07 2013 07:08 dsjoerg wrote:We graph Spending Skill rather than SQ because Spending Skill is more accurate -- it takes into account the race and game duration also, two factors which we know have a major effect on SQ. However I appreciate that an actual number sounds more precise, and you make a good point that league SQ could possibly change over time. Also, I believe the customer is always right! If several GGTracker Pro subscribers request it, we will add an SQ graph.
I explained it poorly. I didn't mean you should add a SQ graph. I assume you calculate the Spending Skill from the SQ and some adjustments. So if you get a number after the calculations that is comparable you could display it in addition to the league icon. It's just seeing the master icon in most of my matches doesn't help me to keep track of my development.
But maybe that's not possible if you don't calculate an actual number to place the Spending Skill in a league.
I have uploaded ~250 replays from the last month.The information you get from the s2gs files is missing in nearly all of them. I assume that's because i haven't played a lot of sc2 recentely and these replays are old. In the games from the last days it works fine.
What's the time frame/ last number of games you can get the s2gs files from?
|
On January 07 2013 18:15 OrbitalPlane wrote:
I explained it poorly. I didn't mean you should add a SQ graph. I assume you calculate the Spending Skill from the SQ and some adjustments. So if you get a number after the calculations that is comparable you could display it in addition to the league icon. It's just seeing the master icon in most of my matches doesn't help me to keep track of my development.
But maybe that's not possible if you don't calculate an actual number to place the Spending Skill in a league.
Aha, thanks for the clarification. We will show "flair" on the league icons to show whether your spending was low master, high master, etc. It was supposed to be doing this all along but there was a bug which I just fixed. We do compute a number -- 0 for bronze, 1 for silver, 1.5 for halfway between silver and gold, etc. Now that the flair is visible, have a look and let me know if this works for you?
What's the time frame/ last number of games you can get the s2gs files from?
We retrieve s2gs (spending skill, etc) for your last 25 to 50 games, and of course all your new games from that time onwards.
|
On the GGtracker site, an SQ of 80 is considered high grandmaster, while SQ of 101 is mid grandmaster. I assume this wasn't intended 
Other feedback:
I'd like to be able to view my average SQ in categories such as: 1v1 SQ, last 10 games SQ, Overall SQ etc. Since almost all of my 1v1s are in the GM range, it means my SQ could be anything 80+ and I have to view each game to check.
2nd piece of feedback is we should be able to see the SQ of a game in the profile summary page without having to click on each individual game by hovering over the icon just like in the game summary page.
|
Australia1191 Posts
Korean SQ will put them over the fiscal cliff!
|
On January 08 2013 01:12 Najda wrote:On the GGtracker site, an SQ of 80 is considered high grandmaster, while SQ of 101 is mid grandmaster. I assume this wasn't intended Other feedback: I'd like to be able to view my average SQ in categories such as: 1v1 SQ, last 10 games SQ, Overall SQ etc. Since almost all of my 1v1s are in the GM range, it means my SQ could be anything 80+ and I have to view each game to check. 2nd piece of feedback is we should be able to see the SQ of a game in the profile summary page without having to click on each individual game by hovering over the icon just like in the game summary page.
Your SQ decreases as the game progresses. So infact it is possible that a Sq of 80 ist high grandmaster and a sq of 101 is mid grandmaster.
You see that the first game took 12 minutes to finish whereas the second game lasted for about 20 minutes
|
On January 08 2013 01:12 Najda wrote:On the GGtracker site, an SQ of 80 is considered high grandmaster, while SQ of 101 is mid grandmaster. I assume this wasn't intended
Whoops thanks for the bug report! It's fixed.
I'd like to be able to view my average SQ in categories such as: 1v1 SQ, last 10 games SQ, Overall SQ etc. Since almost all of my 1v1s are in the GM range, it means my SQ could be anything 80+ and I have to view each game to check.
On your player page, you can click "1v1", which will make the list show your 1v1 games and the charts and averages will change to show only your 1v1 games as well. For example, here are your 1v1 ladder stats vs Protoss. BTW great Spending Skill!! :D
Your last 10 games average is shown in two ways -- it's the last data point on the chart, and it's also the "Recently" value shown below the chart to the right side. Your overall SQ is also below the chart, on the left side.
2nd piece of feedback is we should be able to see the SQ of a game in the profile summary page without having to click on each individual game by hovering over the icon just like in the game summary page.
Agreed! It's on the list
|
On January 08 2013 01:26 aidosae wrote:Your SQ decreases as the game progresses. So infact it is possible that a Sq of 80 ist high grandmaster and a sq of 101 is mid grandmaster. You see that the first game took 12 minutes to finish whereas the second game lasted for about 20 minutes
Very good point aidosae, and I was going to say the same thing, but then I looked at the Spending Skill Stats and an SQ of 101 is high grandmaster for any game length and race. It was just a bug :r
By the way happy birthday! 
|
On December 19 2012 14:09 VPVanek wrote:
My spending skill is grandmaster apparently.
AWesome! :D
This seems to be true for most master players, it doesn't seem to be that of an effective stat.
|
On January 08 2013 01:38 Zheryn wrote:Show nested quote +On December 19 2012 14:09 VPVanek wrote:
My spending skill is grandmaster apparently.
AWesome! :D This seems to be true for most master players, it doesn't seem to be that of an effective stat.
I was worried about this too, so I looked. Currently there are 92 GGTracker users who are ranked 1v1 Masters and have at least 10 games in the system. Their average Spending Skill is slightly sub-Masters. To put it in number terms, if 0 is Bronze average, 1 is Silver, ... 4 is Diamond avg, 5 is Master avg, and 6 is GM, then the average GGTracker 1v1 Masters player has an average Spending Skill of 4.8.
Spending Skill is calibrated to not just those 92 GGTracker users but to >17,000 1v1 Masters accounts, and the thousands of 1v1 Ladder games they have played.
I'll keep an eye on this regularly, as the calibration is important to everyone.
|
On January 08 2013 01:38 Zheryn wrote:Show nested quote +On December 19 2012 14:09 VPVanek wrote:
My spending skill is grandmaster apparently.
AWesome! :D This seems to be true for most master players, it doesn't seem to be that of an effective stat.
OK, I took a closer look and my earlier numbers were not right, although you'll see that the effectiveness of the stat remains strong.
Restricting ourselves to 1v1 Ladder matches for which Spending Skill has been computed, there are 72 GGTracker players ranked 1v1 Masters who have played at least 10 1v1 Ladder games. Their average Spending Skill is 5.08, so slightly better than Masters, where 6.0 is Grandmaster average Spending Skill.
Looking at all 1,493 battle.net accounts that are ranked 1v1 Masters and for which GGTracker has at least 10 1v1 Ladder matches, their average Spending Skill is 5.01, so the calibration is pretty darn good!
It's pretty interesting to look at the players who are ranked 1v1 Masters but have sub-Plat Spending Skill. They seem to be specialists in doing enough damage with an early army to make up for their relatively poor macro. "Winning ugly"?
|
|
|
On January 08 2013 02:33 Bratsche wrote:Maybe this is a bug, maybe it isn't, but... The two games linked below, the icon ascribed to my spending skill is bronze, while the corresponding number is above bronze. http://www.ggtracker.com/matches/617359
This match was 5 minutes, 51 seconds. Although on the web it shows that as 6 minutes, for the purposes of computing Spending Skill we throw out the seconds and call it 5 minutes. I agree that's totally confusing and we'll fix it. For 5 minute games, Spending Skill is kinda borked -- as a Protoss you'd have to get a 91 SQ, at which point you get labeled as GM for that game. Probably I should eliminate Spending Skill entirely for 5-minute games.
This is a 6 minute 8 second game. According to the Spending Skill Stats for AM Protoss you gotta hit an SQ of 83 to be Bronze *average*. On the plus side, you won this game! :D :D
Just an FYI. Your site is awesome. Keep it up!
Thank you! The encouragement means a lot to me.
|
|
|
On January 08 2013 02:52 dsjoerg wrote:This match was 5 minutes, 51 seconds. Although on the web it shows that as 6 minutes, for the purposes of computing Spending Skill we throw out the seconds and call it 5 minutes. I agree that's totally confusing and we'll fix it. For 5 minute games, Spending Skill is kinda borked -- as a Protoss you'd have to get a 91 SQ, at which point you get labeled as GM for that game. Probably I should eliminate Spending Skill entirely for 5-minute games. This is a 6 minute 8 second game. According to the Spending Skill Stats for AM Protoss you gotta hit an SQ of 83 to be Bronze *average*. On the plus side, you won this game! :D :D Thank you! The encouragement means a lot to me.
Ah I see. Thanks for the info about the super short games!
|
Inspired by this problem, I've updated the stats, as we have nearly double the number of matches since they were first posted. I also fixed a mistake that affected short matches -- I was excluding from the stats all matches with a resource collection rate (RCR) less than 600. For 5- or 6-minute games, this is not good because the average RCR for a 5-minute game is 414!
You will now see much saner stats on the short end of the curve.
In other news, I'm gathering Korean games... in about two weeks will probably have enough to present sensible stats.
|
I love it how you guys are turning this into a science, very interesting stuff! Nice to see some data to back up some common assumptions, such as EU being slightly stronger than NA, and macro being crucial to advancing leagues.
|
|
|
On December 19 2012 14:18 Thienan567 wrote:
You know what that means. Terrans have better macro -> more skill -> other races are for noobs :p
I think you dont know what youre talking about
|
|
|
|
|
|
EPT
