What sorcery is this?

played a QM, got curious about what kind of potato codes behind QM algorithm made this choice.

this is a game where both team has 4 assassins and 1 support.

enemy team:
medivh(lvl 980) - mmr 2891 (stacks)
samuro(lvl 1034) - mmr 2865 (stacks)
greymane(lvl 378) - mmr 2735 (stacks)
tychus(lvl 2223) - mmr 2890 (stacks)
nazeebo(lvl 1146) - mmr 2627(solo)

my team:
valeera(lvl 505) - mmr 2806 (solo)
abathur(lvl1401) - mmr 2750 (solo)
zuljin(lvl 854) - mmr 2870 (me)
tychus(lvl 37) - mmr 1756 (solo)
tassasdar(lvl 29) - mmr 1923 (solo)

because for the 2 newbie, there arent much records, but i rmb both their winrate are around 37% and 35%.

can someone explain what kind of potato code can match make unfair matches like this?
like even if they swap one of the newbie with enemy nazeebo (which is still totally fine for the team comp), the game will be a much fairer game.

game records is in my heroes profile,
game ID: 37661913
game date: 06/17/2021 6:30:24 am

EDIT: QM IS THE ONLY MODE THAT IS STILL POPULATED IN MY SERVER

Well, it’s not like you did better than that Tassadar, but I’ll concede.
That’s just very unlucky.
At least, you didn’t loose mmr as much as normally.

I’d complain more about this one. https://www.heroesprofile.com/Match/Single/?blizz_id=242454&battletag=MaestrPotato&region=1&replayID=37649866

2 Likes

oh, because half way we gave up, both me and valeera has split opinion on how to deal with both the newbies against the famous stacks that is farming winrate.

you can look at my other games to see how i perform averagely with zuljin.

The Random Potatoe Team Generator strikes back!

1 Like

Here lays the answer I would say. QM isn’t only around rank, but consider also different other parameters as roles and since it tries to find you a QUICK match, it might end up with weird results as this. Like lucifer said it’s unlucky, but well it’s qm.

Low level account doesn’t mean they are newbies though, could also be smurfs.

Likely due to how players were grouped.

The game looked for match for a long time for them, then gave up and matched them with what was available.

1 Like

yea, that’s why i checked their winrate, one is around 35%, one is around 37%.

yea, but still doesnt make sense where why nazeebo is not swapped with one of the newbie

1 Like

This is extreme, but I guess it still depends when you played your match and on which region that the system considered this to be equal… nevertheless another proof that either playerbase is low or most players are just bad, so there wasn’t enough players to equal it out.

ya, my server’s population is low, but this is extreme indeed. still no idea why dont the algorithm swap one of the newbie with enemy nazeebo though haha.

2 Likes

This stuff has been happening to me as well last few months, every few games there is one guy with vastly below 50% winrate, and he is NOT a smurf, he is atrociously bad, worse than probably beginner AI, and unless you are playing a hero that can either snowball or do some crazy combos, it’s impossible to win.
EU btw.

EDIT:
I seriously don’t know why the game does not let you pick between find game ASAP and take as long as you need for a balanced tank & healer comp game. I am saying this because the vast majority of times the game takes from 1-30 seconds to find a game, and what it finds in that very short span of time is usually horrendous.

4 Likes
  1. Heroesprofile does not reflect the game’s matchmaking. It doesn’t match the players, it doesn’t know where players start in blizzard appraisal, it doesn’t know the hero level or group stack mmr handicaps the game uses.

Looking at HP’s mmr does not match with how the game actually does anything. In the game you linked, and the one lucifer IDed, the outlier heroes have 1 game recorded on Heroesprofile, so it has no clue what value they are supposed to have. If more of their games are uploaded, then it will retroactively reflect in the games you played, and it may try to force the mmr to be closer on those matches.

  1. Since HP doesn’t make the matches, but it knows a match was made, it will skew the numbers to reflect its interpretation of why the games were matched. However, since it doesn’t know where mmr starts for blizz, what handicaps it assigns, if there are ‘seasonal resets’ (if people look at mmr spikes on these parsing sites, they adjust some of the mmr to match ranked seasons in non-ranked modes) This is particularly evident for ARAM records as that mode doesn’t have matching/mmr, but HP will act like it does.

  2. we don’t know how long you waited for a game, we don’t know how long anyone else waiting for a game. Most may only appraise the game for how quickly they got matched, but will rarely (if ever) ask how long anyone else was in queue. Since wait time estimates are based on the average time spent in the queue, one person getting a fast match can be to move along other players that were in a long wait. The longer they wait, they further the disparity on how the game tries to match players, especially on low-population servers.

  3. Games only match tanks if both sides have a tank; it only matches healers if both sides have a healer. If neither get a tank, then it will try to match bruisers; if it doesn’t get healers, then it will try to match supports. The game cannot match what it doesn’t have, so if there is a lack of tanks, healers or bruisers to match, then it will make games will assassins instead.

Since the matching is missing several ‘core’ roles that it tries to prioritize, then that would suggest a lack of stuff to match, or rather, that some players are in a longer-than-intended queue.

While heroesprofile may try to normalize mmr spread between players, in-game they may have a wider actual spread than seen, which then makes it harder for that matching to find players to make matching averages for the partied players. If you look at the different mmr-types Profile uses (overall, hero, role) you can see some players have a bigger variance than others, which can show part of the issue of the skew HP has in how it evaluates players in ways the game proper does not. Or vice-versa, in that the game is using a value HP is not.

The usual take is players look at the account level and see <100 and >2000 and then ask ‘well why did my side get lower’, and matching doesn’t use account levels, it uses mmr. While players with lower levels could/should have less experience, and thus shouldn’t have the same mmr as a ‘master’, matching based on mmr doesn’t indicate what the other players faced to have that winrate. Having a 35% winrate at ‘grandmaster’ play is different than 60% winrate at bronze. Functionally, people wouldn’t want that sort of disparity on winrates being in those given ranks, but averaged mmr is based on several games, on not snapshots people took of one game.

While people tend to use the information they see to ask 'well of course player is new/smurf/whatever, the computer can’t necessarily make that assumption, and then create a bias around that. It simple takes some numbers, makes a priority of those given the opportunity of the situation, and then tries to make adjustments for the next time around.

People want particular roles, skill range, competency, etc. But there’s only so much of it to go around, and others are stuck with ‘luck’ on what they happen to get since level 9999 accounts could have grounded that out in AI, or any lvl 5 account could be another Fan (streamer that’s a former hgc player) on an alt account.

2 Likes

yea, i know, that’s why i said that one of them have around 35% winrate and the other around 37%. i checked their winrate to make sure they are not smurf.

then why is it impossible to swap an assassin with an assassin? the nazeebo in enemy team could have swapped with any of the newbie in my team and the comp will still be the same, this is something that im trying to figure out.

i think you are misunderstanding something here. a bronze will never have 60% winrate, and a master level player will never have 35% winrate. that 2 newbie is lvl 37 and 29, which have played enough games for the match making to determine their skill roughly.

and a 35% winrate player will never even get close to master games in the first place.
just imagine this, you keep losing games, but your rank gets higher, or the other way around, you keep winning games, but your rank gets lower.

this will create situation where solo queue master pub stomping newbie non-stop or newbie keeps throwing master lobby’s game.

if both these newbie are under lvl 5, then the system might still misjudge their skill rating which is acceptable for me.

Because qm allows partied players, that means the mmr value for what is matched in a game can get very wide. The wider the mmr on a loss, the fewer points lost, but the more gained when it wins. The point of my writing that is you don’t know what ‘level’ of mmr matching those players had in the history of their gameplay. Some losses lose more points, some wins gain more points, but the particular winrate doesn’t necessarily mean what their mmr is going to be. While you can see a percentage and declare you don’t think they are smurfs, the problem there is what games they won may have been in a stack (at high mmr disparity) or it may be that sometimes it’s not the same player on that id.

You can put emphasis on particular values, but that doesn’t mean the matching values those same things.

The one side is a group of 4; the fifth slot is not a tank, it’s not a healer, it’s not a bruiser. If those slots were available, then it would have tried to prioritize those to fill the game.
When you look at the match-up, you see the low marks as the outlier, and thus think those have to be ‘swapped’ with the one vacancy on the group. However, your question does not consider that those two where the priority match to the rest of the team, and that someone else on that side is the ‘fill’ to match the assassin slot. The reliance on HP numbers removes your consideration on what the ‘swap’ would actually be.

As i wrote before, different players have different waiting times, which may influence the priority on which they are matched. The system likely has specific slots it’s trying to match (against parties) where it can, and then it fills afterward. The is likely not a ‘post match’ analysis to just swap things around. You chose to make that assessment because you’re looking for something that is ‘wrong’, but the system does not make that same process because it’s self-analysis does not have that same priority.

If you run a google search, there is a priority to the order that links appear on the screen, and some of those aren’t going to be the ‘best match’ based on how you searched, but the results of the information that process used in a particular order. As an example, when I did a search for ‘wood league’ (For another topic) the engine thought ‘wood’ is close enough to ‘tree’ so it gave me results that had ‘tree’ and ‘league’ in the website and ignored my choice of ‘wood’ until I was specific with uses of + and “” marks.

While the matching isn’t a ‘search engine’ for player input, it has an order for priority on how it matches, so the order on when it sets players isn’t going to consider the ‘well, why doesn’t it ‘x’’ as a post-match concern. Players are looking for a ‘fault’, but the matching doesn’t consider the process ‘at fault’, so it’s using what it can find, and putting it in order accordingly.

A key concern here is you see the positive end of the Heroprofile information as a ‘master’ and you like it, so that’s the information you value. However, since that label is unsubstantiated for the game, it does not have any bearing on the actual matching level that it used for averaging that game. Since you highly rate that, the consideration that lower level players (with a particular win %) therefore can’t be at that ‘level’, so the matching must therefore be flawed, and you stop considering variations that don’t rely on that bias/assumption.

So again, for you, you see the low marks as the outlier, so those may have been ‘last’ to switch the nazeebo (last for that side) instead of seeing them as ‘first’ to already be matched against that team. Granted, it is harder to make ‘sense’ of newer players thus being matched like that (with a high level difference + mmr handicap) but a key part of this still boils down to priority (queue times over mmr parity) and role/player availability.

Keep in mind you already doubt the system (‘potato code’) so you’re not really interested in what the system does since anything that isn’t “fair” is a fault of something easy to blame.

1 Like

agree this to a certain degree, but it is still impossible. lowest winrate i have ever seen in games are around 35%, i have never seen much lower winrate that dip below 30%.

and as you said, they can be stacks, but the higher mmr his teammate is, the harder it will be for them to win too. since most of the game it will be 4v5 and as we all know hots is a moba that are even much more dependent on temmate if you compare it to Dota2 or LoL

the solo carry potential in Hots is extremely low, hence 4v5 will be extremely hard, and your example of 60% winrate on bronze is very extreme. and much more impossible the other way around too.

and for the opposite case, a master with 35% winrate is extremely impossible due to enemy has to be constantly 5 stacks that has really good teamwork almost every single game. which from my personal experience, this have never happen before.

im a solo queue player which heroes profile think im at master level in QM and my winrate is around 49.2% at the moment. if im having 35% winrate, i pretty sure i wont be playing around master level.

take the tychus with 37 level as example, im not sure how many games that is, but just lets assume he played 37 games. he has to face 5 stacks enemy with same MMR on both team every. single. game. to maybe even get 35% winrate.

still doesnt explain why an assassin cant be swap with another assasin though.

this too, when 10 players are matched into a game, isnt it suppose to average out the MMR again? like, isnt this common sense?

your conern here is queue time, which i think doesnt matter as queue time only dictates how fast a player can get into the game and broader skill rating tolerance across all players in the same game, but it should never pair bronze and master level player no matter what right?

this i dont agree with, as i said, i play against enemy stacks alot, and i rmb their name, they have extremely high winrate, if im not mistaken, some of them have around 80% winrate.

and the stacks keeps farming more and more winrate, how are the stacks not defined by match making system as maybe “grandmaster” or maybe top “500 players”? with like maybe 70% average winrate?

quite the opposite, im interested in why the match making system is trying to do this. your assumption is that im wrong about what the player’s actual skill rating is, which i dont agree, since i have been into this situation quite alot in my server.

not to mention, that a player with higher winrate but lower skill rating, or lower winrate with higher skill rating doesnt even make sense in the first place as this is what makes the game broken.

no player wants to play in games like this at all. if you are good at the game, you still lose more games? doesnt make sense right?

there are some game records i have here, take a look, especially the last comment.
this is a norm in my server. i know i cant avoid 5 stacks but sometimes it is too extreme.

nobody should have that high amount of winrate even in master games.

and i also dont agree with you on winrate shouldnt define skill rating. you win game, you have skill, you lose game, you dont have the skill, simple as that. the higher your winrate is, the better your enemies are next game.
of course, when the server is filled with smurf, that’s another case, but still, it shouldnt dip that low.

but examples like these aside, the 4 stacks game in this topic is another kind of extreme imo.

maybe blame extended qm search

1 Like

Ye that thing is why i lost like 60% of all my matches yesterday. Its stupid to watch qm goes into extended que after only 5 sec at 1700 in the afternoon on EU server. Then you know some stupid comp will pop up and yes I was right about 80% of the time.

Solo tank vs tank/bruiser teams.
Full melee team vs a team who has 3 range.
Solo tank and two bruisers againts solo Garrosh
Full mage team againts two strong auto attackers with Aba on thier team

Should not surprice me if matchmaker is bugged aswell since it create so uneven and unfun games. Its basicely 100% stomps 70% of the time and new players learn nothing of it. They will uninstall when they see enemy team was a full 5 stack with a GM on thier team while his own team was all unranked and maybe a few silvers.

Its not long time ago I just got stomped in qm late night at 2 pm againts a 5 stack with 4 masters and unranked Aba on thier team while my team was full mage team and they had Aba hatted Zera/Samuro to bully us all back to base lol.

5 stacks should only get matched up againts other stacked teams but that train is gone long ago lol. Now solo players and duo teams has become training dummies for 5 stacks and that will contine untill all solo players has uninstalled and left this game.

7 Likes

If this is true, then even EU population is getting low. That’s sad.

First extended que was activated after 10 minutes (year 2016-2017). Then after 300 seconds or 5 minutes.

In 2020 it was 20 or 30 seconds. Now 5?

1 Like

No idea but I just had 2 games yesterday at late afternoon and game went into extended after only 5-10 sec even when i qued as healer with bonus.

1 Like

should have been the case no matter what tbh. 5 stacks normally means they have better comp and teamwork than the other team by default.

especially frustrating when the 5 stacks picks bunches of AOE crowd control heroes. like, there isnt even a single decent fight in that 15-20 minute time span which most of the time result in a lost.

1 Like

I dont know my mmr. What sorcery do you use?

1 Like