SMURFS REPORT - W27-30
→ W27-29
→ First report in other leagues !
Hey folks, there are news !
In the first exchanges of the thread, there was feedback like what the reports I made over smurfs ratio suffered from the bias of being limited to my league, and not independent from my gaming habits. Well, several weeks later, I discussed once more with Cheezecake, and while he still hasn’t a knack for nuanced debates, he suggested SC2 API as a possible alternative source.
I inquired about it with the quite friendly creator of the excellent RankedFTW website (who obviously excels at programing), who confirmed it was possible and provided me with greatly useful advice. Go support his site if you can !
I’d also like to thank SC2Stalk’s creator, who diligently answered the few questions I had left.
So, additionally to the usual smurfs report from my league, there will be the first smurfs report from another league, completely independent from my own wanderings. 
https://www.rankedftw.com/ladder/lotv/1v1/mmr/
http://sc2stalk.ninin.me/
SMURFS REPORT - W27-29 (PLATINUM 1)
https://i.postimg.cc/CxGF0yyz/Smurf-data-W27-29.jpg
The usual report from my ladder.
How to read : cf. above post or first post (II).
Quickie :
Over 93 users played :
- 24.73% of confirmed smurfs
(22.58% of freelosers + 2,15% of non freelosing). - 75.27% to 55.92 of regular players.
This is once again not that different from the previous reports, so I think we’re getting to a point where we’ve got enough data on my league, and will probably stop reporting it once a sample of about 500 players is reached.
I’ve got some doubts about the career games criteria though, I wonder if I shouldn’t make it 100 for platinum and 150 for diamond. I’m also wondering is 75% winrate isn’t a too high cutoff, but it would be unsafe to tweak it too much so I’ll leave it at that for the moment. Here are the current criteria for non freelosing smurfs, for the record :
- Total career games too low for MMR ( < 50 for platinum, < 100 for diamond)
- ≥ 2 leagues ≥ 2 times with the same race.
- ≤ 2 leagues of his opponent outside of provisional MMR.
- ≥ 75% 1v1 winrate over the last 32 games.
SMURFS REPORT GOLD1 - PROTOTYPE
So, I was able to get past my programing complete lack of knowledge and OAuth to retrieve the normally inaccessibly Match History of players never encountered by my account. This was made using SC2 API for clanless players (/24 games), and using the ingame rosters menus (/48 games) for clan belonging players. Profiles discrepancies were analyzed using RFTW and Battlenet web or ingame profiles.
The criteria I used were the same than those used to detect smurfs in my own ladder : ≥ 3 freeloses for freelosers ; and those just above for non freelosing smurfs.
SC2 community APIs provides me with the 24 last games played by an user, dated in seconds and with their results. It comes under that form :
{
"map": "Deathaura LE",
"type": "1v1",
"decision": "Loss",
"speed": "Faster",
"date": 1592593370
},
{
"map": "Submarine LE",
"type": "1v1",
"decision": "Loss",
"speed": "Faster",
"date": 1592593314
},
That’s an example taken from an actual user from the very bottom of gold league. As you can see one game was played at 93 314s, while the other one at 93 370, all the other digits being the same. Hence by substracting those 93 314-93 370, I get = 56s, that’s a freelose.
That was also a good opportunity for me to realize I hadn’t set down subtractions in ages. Good exercise ! 
However, compared to checking ingame history, that method is vastly less precise : it can only detect freeloses if the two games are played immediately consecutively, if there’s a game played after the freelose, and if the waiting time wasn’t unusually high. Furthermore, I get only a 24 games sample, compared to the 48 of game’s menus.
So I had to define a somewhat large cutoff for freeloses, which I defined as game lost in less than 60s of waiting + loading time, with the game then left in less than 1’30 (earliest attack timing). So all loses ≤ 150s from another game were considered freeloses.
So, here are the results for gold 1 :
https://i.postimg.cc/hGKKPCgM/Smurf-API-data-W30.jpg
- 32,0% of confirmed smurfs
(14% of freelosers and 18% of non freelosers). - 68,0% to 42% of regular players.
Don’t get carried away, I believe those results are inaccurate. And that for two reasons :
- I did only randomize the players at each MMR sample : using a random number generator set from 1 to 10, players getting even numbers were included, those with uneven results, discarded. But I did still chose manually the MMR samples : the top, the bottom of gold, and then in-betweens. This proved exaggerating the importance of top gold users, and of bottom gold players, and doesn’t accurately reflects their amount in the overall gold population.
. - Second, I included all the gold ranked players. But, surprisingly enough, there are gold players with 3580 MMR (which is near Diamond 2 ), and also some with as low as 1145 MMR (which is Bronze 3
). And of course, 80% of those are smurfs. So, at one hand it overly exagerates the ratio of smurfs, but at another hand it answer to the question « How many smurfs are gold players ? » and not « How many smurfs are encountered by gold players ? ». After due deliberation (with myself XD). I decided the question I want to answer to is the second one, in other words « How many smurfs are encountered at gold MMR ? ». Hence I won’t use the gold filter anymore, and will lose previously spotted smurfs, but will get a more accurate feel of how many smurfs you gold players are being plagued with.
So week 30’s first API report will be considered as a prototype, and not included into the first post’s pooled data. For the next versions :
- The definition of studied samples will be made by MMR borders within the studied league tier, and not by league anymore. This implies that only players effectively playing at gold MMR will be included into the gold sample.
- Every MMR point, for each inclusion, will be randomized within the studied tier MMR borders ; additionally to the players then eligible being randomized a second time. This implies that there will be much more MMR points studied, and that I won’t choose them.
You can expect a reasonably more accurate report of the whole gold league in about three weeks.
Fun facts :
-
So, there are gold players who have as high as 3580 MMR. How is that possible will you ask me ? Well it’s simple, this kind of smurf do only use their profile for a handful of ranked games (1-6) each season, and then only unranked. That way they do stay indefinitely in provisional MMR (with the league that goes with it). And so they can troll P1-D2 players with a gold league frame. As I could see there however, this type of smurf does not freelose, they want to win, and the more overwhelming the better. That’s still rather dumb from them, as the opponent could suspect right away the smurfing.
-
And on the contrary, there were gold users with bronze 3 MMR : simply rank yourself, and then freelose. And it happens that on week 30, one of the most spectacular examples… was an officer, from a significantly large clan (with separate recruiting and official players groups, and even clanwars dedicated groups). The guy was diamond 2 and spent his time destroying bronzes at 1145 MMR.
W30 Officer smurf — Postimages
And the best part, is that that win was earned only because he was matched against another smurf, who left in 0:00s. Also randomized a similar one, who also had a freewin against another smurf, and then proceeds to develop a 130APM play against 30 APM newbies.
Truly inspiring.
I recently had a funny mirror game against a smurf, but this post is already quite long, so I will keep that for the next report. See ya. 