213 Match Analysis: Matchmaker Bias, Surprise Findings, Etc

Competitive Discussion
1

Note: Sorry I can’t properly link or embed images. If there’s a work-around (?) I’m happy to edit them in. For now you can copy-paste this Imgur thread url and scroll to the appropriate “Fig” number reference in there: https://imgur.com/a/4DWoM2m

Introduction:

Last October, I posted to the OW University subreddit, lamenting that matches in silver, on my main account, were substantially harder to win than those in high gold, on my alt. (I suspect my skill ceiling is roughly low plat.) I lead the post with this S24 placements comparison graphic - 27% vs 60% win rate (main vs alt) averaged across all 3 roles (solo queue): https://i.imgur.com/imh2KGr.jpg

I speculated about the source of the apparent systematic bias perhaps being a disparity between comp SR and MMR. After discussions and further research, I decided that was most certainly nonsense, or at least the wrong framing - SR is MMR in comp. Particularly helpful was Kaawumba’s fantastic thread(s), referencing everything that’s officially known about the matchmaker. See the latest one for S28+.

In my mind, this left 2 plausible sources of systematic bias: player stats from within comp (which I can’t do much about), or QP stats/MMR, which I could potentially ‘adjust’. So I set out to test the latter, while seeing if I could also quantify the bias (or better still, just climb the cursed account!). Recording 30+ pieces of information per match, chipping away at it when my ME/CFS would allow.

Two sets of placements, on each account, clearly wasn’t enough matches for meaningful comparison with that, either. So I placed a few more times on my alt, as an additional control. Also to prove it wasn’t about to plummet to silver, too.

This Imgur post thread contains all the insightful graphs, plus relevant infographics/screenshots, with full explanation and some discussion appended to each image: https://imgur.com/a/4DWoM2m
⠀

Core findings (summary):

(1) Consistent evidence of matchmaker bias against tank role on my main. Initially stuck for 31 matches at 1900SR, 400 SR below my alt [Fig.1 https://imgur.com/w0qv5HU], with 45% win rate [Fig.6 https://imgur.com/UJEHfuR]. It stayed exactly level on SR due to the performance adjustments giving significantly more for wins vs losses [Fig.8 https://imgur.com/BcyX8Lz]. Consistently strong performance was also seen with even higher average medal stats in losing matches [Fig.7 https://imgur.com/av6cI7B] (i.e. my team mates were far more inconsistent). My stats did not fall away at all when tank SR did finally increase a little [Fig.18 https://imgur.com/bTwadki]. I also estimated my tank QP MMR at 2500-2600 SR equivalent (see point 10).

(2) Matchmaking changed drastically for my tank role at the start of season 27, permitting me a totally unexpected and unprecedented 10 match win-streak! [Fig.2 https://imgur.com/eeG36L9, Fig.1 https://imgur.com/w0qv5HU] I gained 33 SR for the last 2 wins (28 for a couple before that) - a clear win-streak bonus. These matches felt like they’d been balanced for me, instead of counterbalancing my above average contributions (the rest of the time). This period ended, returning to a 45% win rate, stuck now at just over 2100 SR (200 SR below alt).

(3) Possibly circumstantial evidence: Throwing 10 QP support games (not advisable) appeared to unstick my competitive support matchmaking, allowing it to climb back into gold with ~59% win rate, pre season 27 [Fig.6 https://imgur.com/UJEHfuR]. Tank remained stuck, actually losing more matches initially [Fig.1 https://imgur.com/w0qv5HU]. Support match balance felt more favourable [Fig.17 https://imgur.com/cjsSxay], then gradually slid down to feeling negatively biased, as it got stuck again at 2100SR, 300 SR below my alt.

(4) My alt held its rank over an additional 3 seasons of placements [Fig.1https://imgur.com/w0qv5HU]. (101 comp matches now played on it, in total, across the 3 roles.) DPS climbed a little to borderline plat (despite being my weakest, least played role). Support fell a little (but I’d been playing lots of QP support on it with my masters skill friend). Tank feels under-ranked at 2300 SR and my attempts to carry also show up in average medal stats almost as high as my main [Fig.7 https://imgur.com/av6cI7B].

(5) Incidence of leavers (3), throwers (11) and enemy ‘smurfs’ (19) was relatively low (out of 168 matches on my main). But about 50% of lost matches had one of these issues and a far lower percentage of losses on my alt were associated with these issues [Fig.12 https://imgur.com/nF31Ro6].

(6) PotG was, predictably, awarded to a player on the winning team about 2/3rds of the time [Fig.14 https://imgur.com/w9k0xmy]. But, beyond this, there was a noticeable bias towards enemy team PotGs, particularly in lost tank and DPS matches.

(7) On average, it appeared I was being pulled down into matches with slightly lower SR players [Fig.9 https://imgur.com/8qjpa3x]. Ally and enemy teams ~10SR below. Which could help explain slightly lacklustre average SR gains, on my main (closer to 23 SR, than the expected average of 24). Like masters tanks get pulled up into GM/top 500 matches, one would expect the player distribution to pull the less played roles downwards the low end, too.
⠀

Incidental findings:

(8) More team mates in voice had only a small association with positive match outcome, despite comms being a personal strength [Fig.13 https://imgur.com/3jrU6wM]. There were noticeably more team mates in voice in high gold vs low gold/silver.

(9) Playing after 11pm (GMT) was a bad idea, with a far more negative win-rate [Fig.15https://imgur.com/b9Ur9WZ]. I’ll also be avoiding Sundays in future. (Note: I have a 26 hour rotating circadian rhythm, so those late night matches were still prime time for me.)

(10) Flex and ticket queues saw the same or marginally beneficial match balance/outcome [Fig.16 https://imgur.com/mlmLHjc]. Which, I think (?) contradicts some popular sentiment. However, checking on this was an afterthought and there was too little data for a firm conclusion.

(11) Fairly consistent Quick Play MMR values can be estimated by averaging the SR, across many games, of the other players with visible profiles, on the role they played. At the end of the study, my tank was around 2500-2600 SR equivalent [Fig.20 https://imgur.com/GW5q336]. DPS ~2300 SR (200 SR below the comp rank of my alt, after its 5 seasons of placements).

(12) My support QP MMR took a very long time to fully recover and stabilise, after hard-throwing 18 games [Fig.19 https://imgur.com/0lEOJXV]. Please don’t try it yourselves. It seems that you only get to stomp low skilled enemy players for 1 game; I think the matchmaker immediately assumes there’s a different user on the account’s role. It gave me a lot of unwinnable and imbalanced/rubbish games for a while. Then, after about 24 try-hard games, a load of too easy games, where I did stomp, winning 17 out of 20 without MMR rising. Eventually my MMR plateaued again, after about 70 games, at 2300-2400 SR equivalent (in line with the comp support SR of my alt).

(13) Players in higher skilled QP games are more likely to have open profiles [Fig.20 https://imgur.com/GW5q336]. The trend is based on extremely noisy data, but roughly speaking, for each 100 SR equivalent, that skill increase, there is one more profile visible.
⠀

General interest:

(14) Rein got the most PotGs on my main, by far! [Fig.3 https://imgur.com/py6hOP3] Also on my alt, but more closely followed by Pharah [https://imgur.com/a/2TEoJOA].

(15) Hanamura came up twice as often as most other maps (gods dammit!). [Fig.4 https://imgur.com/P0xRQGk] Distribution looks random, of course, but are map probabilities weighted to have an even amount of each game mode, on average?
⠀

Discussion:

• Personal context: My top goal has been to climb my main, with all its unlocks and levels, to the modest SR of my alt. That’s close to where I think my absolute skill ceiling is, due to my slow (visual) information processing with ADHD-PI and dyslexia, plus ME/CFS getting in the way and slowing everything right down, a lot of the time.

I dearly love this game, but the competitive matchmaker has perplexed me for years. It seemed like the better my QP games were, with higher skilled friends (high gold, plat, diamond and masters), the less winnable my comp matches became, down in silver. I guessed there might reasonably be a causal link there (as I explained above).
⠀
⠀
• Matchmaker function: I think the most important point to highlight is the developer’s focus on making matches with 50/50 predicted outcomes.

They know players hate one-sided matches (and will stop playing if things feel too hopeless, too often). This is the main reason, that I’ve seen them offer, for not implementing an SR reset - months of horrendously random, messy matches (that wouldn’t achieve much, long term). I think this is totally fair.

But, I think it’s not possible to balance matches closely enough, in mid-ranks, while only considering the factors that they’ve made public. There’s hard stuck gold border accounts vs actual brand new players, smurfs, throwers, etc, all piled together in there. And they’ve admitted there are other (unspecified) data used in certain/niche circumstances, at least.
⠀
⠀
• Speculation: My guess was that performance stats from within competitive and/or from QP (or straight up QP MMR) are used to more finely balance teams in mid ranks.

Obviously, this would cause a kind of ‘stiction’, holding back climbing, because it would pit under-ranked players against equally under-ranked players (judging by their higher stats). Counter-productive to rank mobility. But it would still allow those who improved further to then climb.

Presumably there would be a cap on how much it will counter-balance a player. So, very under-ranked players (which I’m not, mostly) could always climb, too. Although, generally, smurf accounts, and high skilled new players, will climb up quickly, within a relatively small number of early matches on an account. While the account is still getting special treatment…

I think the devs have said that they keep new players away from most of the player base. I think they also make a point of balancing their early comp matches for them, too. To give them as much influence over the outcomes as possible, such that their rank moves in the right direction, initially. This would stem naturally from having fewer performance stats accumulated, to use to counter-balance them.

I think that’s a key question: who is the matchmaker balancing for, and who is being counter-balanced? Long time, regular players, it would take as part of the furniture. Except maybe under certain circumstances. Like, I feel that my high stats on tank might have passed a threshold, triggering the matchmaker to balance for me on that role (for a time), smoothing my climb.

My win streak started right at the start of a season, so I assumed it was a global change. But I’ve not heard of anyone else having had sudden success like that. So either it’s targeted, or randomly allotted…?

It definitely felt completely different. And I was expecting the usual 45% dross, so definitely not a case of expectation bias. I was actually disappointed it was winning while my support wasn’t (given that was dispelling my hopes of having figured out a QP link).

I previously thought the system might only counter-balance players it guesses are smurfing, deliberately avoiding ranking up, whatever. That it misidentified me and I was then seeing more enemy smurfs than other players, holding me back. Or, that it would set me, as an above average player for my rank, with a few other above average players, against super-smurfs. Which were more common than the devs expected, in my rank.

50% of losses on my main did involve a suspiciously high skilled enemy (or leaver/thrower). But I probably failed to identify more ‘smurfs’ on my own team, without the benefit of kill-cam, etc. And I saw relatively few obvious (DPS) smurfs, during this analysis, compared to previous experience. It seemed like mostly an ineffable issue, of overall balance of player skill, initiative, cooperation, etc.
⠀
⠀
• The analysis: When I started this little study, I promised myself that I’d publish it, regardless of outcome. I thought, most likely, it would be a “nothing to see here” debunking of my previous Reddit post.

But, while it’s very far from a proof (which would be impossible, looking at only one account), my results seem wholly compatible with QP stats influencing comp matchmaking. At least in my mid-low-ranks, where we have performance based adjustments (I doubt it’d be used at the top). Even the small drop in my alt’s support role could arguably have been expected, given that I was using my alt to play QP support with my much higher skilled friends.

However, after being 90% confident, at the time I took a break from playing in season 25, I then lost my confidence. The strength of the apparent effect waned, in season 26. It could be that the matchmaker looked only at (or weighted for) stats in recent games/matches. If QP stats are used, then that wouldn’t matter for players who only play comp.

I mostly play casually, given that my ME/CFS makes me too fatigued to play seriously most of the time. Hence this analysis taking 7 months. Hence why I play at least 4-6 warm up games, first, to be sure I’m up to speed. (Which might lower my QP performance, relatively.) And hence why I played a match of each role, together, each time, to control for any personal variation in ability, due to variable ill health.

I also do not recommend anyone else try throwing in QP. For a start, it’s an aweful thing to do to the other players and I felt bad doing it, despite approaching it to make it look as subtle as possible. It was a last ditch attempt to find something to alleviate years of frustration, and then report back for others. If you suspect this is an actual thing, a better plan is to only play comp on the role/account you are climbing (then use another socially).

Even if this study convinces no one of anything (let alone prompting Blizzard devs to tweak things), at least it still helped motivate me to just play more matches. It also took some of the dismay away from negative outcomes; at least I was getting useful data, every time. This has, at least, helped me climb my main roles 200 SR and DPS 400 SR (faster than expected, although it’s still 600 short of my alt). Also, I’ve learned the idiosyncracies of Google Docs spreadsheet (woot, lol!).

So, I do recommend match tracking, and analysing replays, to get a more objective perspective on what’s happening and how you are playing.
⠀
⠀
• Debunking: I would have liked to have been able to dismiss those conspiratorial sounding claims that the matchmaker ‘pushes’ people to the rank it thinks they should be; I think that idea came from a misconception about a win/loss adjustment factor that the devs had only in season 3, to correct a maths mistake with the SR distribution curve.

But, I admit, it does look kind of weirdly like my SRs have homed in on specific set-points, and stayed virtually glued there. My feeling is this is probably still just explainable with thresholds of when biasing effects fully kick in. Like glass ceilings.
⠀

• Finally: My Google Docs analysis spreadsheet is shared for potential scrutiny (and can be copied for personal reuse): https://docs.google.com/spreadsheets/d/1QOnzbXvYBWXqbj5X7pgNbp3ycNaUo3CYcFIJs0lRwQc/edit?usp=sharing

I’m happy to answer any quick (or long) questions based on people having read only this post, given that there’s a lot of text to wade through in the main Imgur thread-post: https://imgur.com/a/4DWoM2m

3 Likes
2

I apologize, I didn’t read everything.

Have you considered playing equal numbers of games on both accounts to see if they converge?

As I didn’t read everything, I am not certain how much weight you’re putting on the placements of the alt account. So, I apologize again. But since the placements on the alt account are all around the starting position of 2350 SR, is it possible that the most significant thing about the alt account is that it hasn’t played a sufficient number of games to be accurate?

I’m just asking questions, not trying to be dismissive. I think tracking progress with spreadsheets is very helpful, especially the insights about which times of day are best to play. I tracked for a few seasons, and learned that it was a bad idea for me to play after midnight on weeknights in US Central time. Either I was getting too tired, or the games degraded… Also, I shied away from immediately after school but before adults generally got home: 3-5PM weekdays.

Map troubles also seem to be a really good piece of information, so I would go try and figure out why that map worked poorly for me, and watch more videos on proper strategies for those maps.

The amount of detail is impressive. I’m sorry again for not spending enough time reading it all. I may yet come back to it and try harder to digest it all.

2 Likes
3

So, this was the biggest comment on my first Reddit post, after 2 sets of placements (and loose change), back in October. Then it was basically the only thing talked about in the comments when I tried to post this up again, after 5 sets of placements (and the post got down-voted into oblivion -no bais u just bad thinking).

My alt’s had 33 matches per role. I don’t feel like I can prove that it won’t eventually drift down. Although it already touched plat on tank/support, and is resting 200SR down from that peak (unlucky last couple of accelerated SR matches, my excuse). DPS actually climbed ~75 SR during the study. Support did fall ~75. But if my hypothesis of QP influencing comp matchmaking isn’t complete nonsense, that would make sense, as I played QP on that role with much higher skilled friends, using that account instead of my main.

Bottom line, I didn’t (and don’t) really want to play more on my alt, because I’m not motivated to climb or grind it, at all. I’ve only ever wanted my main (with all its unlocks) to get to the SR I’m pretty damn sure I’m comfortably capable of. Which yes, sadly is only barely above the starting point of 2350. I’m fine being a bearly above average player. Despite gold border and many thousands of hours watching, let alone playing, the game. [End rant, sorry.]

heh. Shockingly similar to me, with the after school times too. Well, that was my feel in the past; I wasn’t able to see that effect, in this analysis. But maybe not enough matches played at that time. And it’s not my own fatigue, here, for sure, as I play at the same time in my wake cycle. But sleep every 26h (broken with ME/CFS).

So it’s interesting to wonder who gets the flip side of that late night bad/losing matches? Who’s taking the wins? Within house account sharers? Big bro coming on late…? Super-smurfs coming out to stomp instead of sleep…?

Actually, that’s a good point. I’d not thought about correlating win/loss with specific maps… Although I did make an awful lot of graphs already. That’s to go on the todo list for any future study.

Thanks for the input. :slightly_smiling_face:

1 Like
4

Hey, no shame in that, I’m a software developer in my 40’s, and getting back up to 2350 would be an accomplishment. I’ve never had an excellent reaction time, even though I’ve played shooters for literally decades. Honestly, though I’m probably better than I’ve ever been at this point.

Ah, sorry.

Actually I don’t even know why they bother with placements at the beginning of every season. At this point, the seasons shouldn’t even exist below Masters. I mean there’s no chance for me to ever be on the leaderboard, so why do I care? I also have no interest in sprays or profile icons…

Oh, that’s another thing… Beginnings and Endings of seasons are way more chaotic than the whole middle of the season… Too many people trying to do their placements just for the “rewards”.

1 Like
5

Right?! I swear they even cap the amount of SR change, per placement match, to be closer to 24 SR than a regular match (having won a 5v6 a few seasons back, then interpolated the changes).

I think there are still seasons mostly to motivate players to come back and play comp at all. For the competitive points.

Yeah, that’s my standard experience. Though with fewer players, at the moment, and low-ish rank, end of season’s not that remarkable. And there was someone placing in 90% of my recorded matches, through the season, anyway. Without those, I think the player base might struggle even more. And the role distribution.

I’ll always avoid the first day, or two, of new anything, though; I tend to get these ridiculous loss streaks, 10-15 games in arcade and QP, on the first day of any new event that’s at all hyped. I think the matchmaker favours retuning players, under the technical pretext of higher skill uncertainty, but the intent of hooking them back into playing with wins (not off-putting losses).
⠀
⠀
Edit note for main post: I’ve just made all the sections collapsible, to make navigating the wall of text easier. (By borrowing the markup code Kaawumba used in their post, thanks.)

1 Like
6

Very illuminating writeup

I testify to this

:astonished:

2 Likes
7

“You are hardstruck like me too! Must be the matchmaker and not our terrible skill!!!”

3 Likes
8

Confirmation bias

2 Likes
9

I’m really impressed with the amount of W/L data you’re showing! Your graphs are very impressive! It seems like you weren’t able to come to much of a conclusion though about your hypothesis, which is a shame.

Just to make sure: You were doing “4 or 5” QP games just donking around/“throwing” before going into comp and things seemed to feel better overall, right?

I’m a bit confused as to the purpose of the second account in your experiments. Was it just a control or were you running the same experiment twice?

The most I’m able to make of your data is that you’re on an about 3 game cycle of W/L with a few outliers. From my experience that is the typical bounce between skill peak and floor. While I’ve been under the impression that a skill variance of 500 is considered normal - I’m open to the option that player statistics can carry over between gamemodes. You’ve clearly put forth effort into your study of the concept.

I’ll need a little time to see if I can replicate your finding in any meaningful way.

From my personal experience, I’m normally placed with Gold players in QP on both my alt and my main (both are in mid/high silver) but I don’t honestly feel like I’m good enough to break Gold.

In regard to your question about maps… na, there doesn’t seem to be any specific correlation with what map you play at any given time. Last I checked, around season 12, specific map ratios were around 25% chance for any given map type in the rotation.

1 Like
10

The date & time of the QP games are all in the “QP MMR estimation” sheet of the google doc. The first QP 10 support throws were in the days before the first comp match of the analysis. Then there were 8 more, total, in the first 2 seasons, either accidental flex queues or checking up, topping up.

We’re talking virtually zero heal or damage zen/lucio. And, to be clear, I don’t really recommend others trying this, as it’s disruptive of other player’s games (and feels aweful). Although, after the 6th game, I was no longer able to cause losses. The matchmaker had adjusted, in some capacity.

The matchmaking did initially feel better on support, then. But as [Fig.17 https://imgur.com/cjsSxay] shows, that feeling trended back down, as that role came to rest 200 SR higher. For whatever (or no particular) reason.

So, this study was a kinda of follow up to my Reddit post, that I mention in the introduction, where I compared the two. Matches on my alt feeling easier to win, despite being significantly higher SR (400-800 above).

I played a bunch more on my alt, yes, as a kind of control (nothing zany tried there). And to verify to myself that it’s not about to plummet down to silver. Not that the “ur just bad” crew would admit there’s any meaning in this.

You mean the zig-zag pattern, on the SR progression graph of my main [Fig.1]? Is that the accepted pattern, of a player at their (current) skill level?I see.

Well, given a consistent matchmaking bias, you’d expect to see the same pattern, just offset from the SR the player would ‘naturally’ be at, if the matches were balanced for them. (As is probably the case for new accounts/roles, more so.)

As Cuthbert says, you’re fighting your own shadow, and need to play better than yourself, your matchmaking counter-balance, in order to climb.

Just tracking the data, without any experimental interventions before hand, could be interesting. See if any roles show you over-performing. Although, to be honest, I am very curious to see if anyone else can unstick themselves via QP adjustments. But I think it might be a somewhat niche effect, if it exists at all…

Oh right. If it is definitely 25% per game mode (map type), then that would mean a lot more Hanamura, Volskya and Anubis for everyone, not just me. That’s all I was wondering.

11

Your probably had a lot of work with this, but sadly its too hard to read, I head trouble to even finish it and its very confusing. At one point you talk about throwing QP games which i dont understand as QP and comp are completely separated.

There is so many pictures in thex text, I cant bother myself to even watch them. I would aprretiate just numbers a lot more than that.

3 Likes
12

No really, as someone with serverl alts i can say that this is not true.

Matchmaker is balancing match for teams, not for individuals. There is no YOU in eyes of matchmaker , you are just number to fill average MMR vs average MMR teams.

3 Likes
13

Sorry. I guess I do take for granted the ability to make sense of graphs, given my science and engineering background (in my life before chronic illness).

You’re welcome to look at the raw data, in the Google spreadsheet itself, but I doubt that will be helpful…

The post, above, with the numbered findings, was supposed to be more concise summary. But has ended up quite cumbersome, especially without being able to link directly to the specific graph images…

They should be, as far as we know. But I’ve felt like my comp matchmaking has gotten harder, in the past, in proportion to how much better (higher skill) my QP games got, playing with more talented friends. So I decided to try and test it directly.

It could make sense, for the devs to include stats from QP, especially for those who play little comp, to e.g. try to match you with a fellow role team mate who doesn’t one-trick the same hero. Or make sure each side is equally disadvantage, in that regard.

Or, I supposed, to see if one has a higher QP MMR than their comp MMR, to make sure the matches are balanced in that regard. Or if not MMR, then raw damage/healing/etc stats.

Getting 50/50 odds balance is very hard to do. I’m sure they use more than just comp MMR (and uncertainty) numbers. (And group sizes.)

You mean that your alts all felt the same to play, in mid-gold, as your main (or an old alt)?

To me, my alt felt like a breath of fresh air, first season, plus. Completely different and easier than my main, ~500 SR or so below. Shrug.

It’s trivially easy for a system to just balance the MMR numbers on either side. I’m sure there’s always millions of possible combinations of team line-ups that would fulfil that basic criterion, even late night.

But the matchmaker goes way beyond that simple maths, in aiming for 50/50 odds. Taking a long time comparing a lot of different dimensions, I think. So it’d be interesting to know (for sure) everything it does consider, under all circumstances. I’m sure there must be helpful things in there, that would sound bad to players, so the devs don’t disclose them.

Yes. But if it doesn’t know much about a new player, for example, it can’t take into account their specific strengths/characteristics, to counter-balance the match with someone like them on the other side of the equation. The side effect is that a match will be more fundamentally balanced for them.

Which is arguably very helpful, for a new player, to have their relative strength/weakness be the main decider on if they climb/fall.

1 Like
14

Well i am looking at it now, spreadsheet is a lot better info than majority of your topic here , sorry man :smiley: If you cant post pictures, you should just shared numbers instead (you need trust level 2 or 3 to post picture btw)

Can you summarize in like 4 sentences what was your goal? To prove that matchmaker is giving older accounts different games than to new account?

Sorry cant read more now, i am super tired and i am going to sleep , i will try to read rest of your post tomorrow.

4 Likes
15

Btw, you wrote a lot after every match in that spreadsheet, I do similar thing and it had to take a lot of time.

One last thing today, I might have idea why you are not climbing. You play way too many different heroes and roles, you should specialize and focus one role at one time.

3 Likes
16

I see, so we’re talking no standard interval or trigger for you to jump into QP. I understand why you put in the disclaimer for people to not really try it out but it is a key component of your hypothesis. =\

I refuse to believe anything that cuthbert puts on the forum. It’s conspiracy nonsense and proven incorrect. The only thing you’re, realistically, “fighting against” is normal human inconsistency.

I take you 1000% more serious than him because you’re willing to bring actual data to the conversation while he and his cult bring nothing but hearsay, twisted developer statements, and disproved theories.

I’ve tried a few things to “unstick” accounts in the past, including methods of MMR manipulation, but nothing does it better than actually getting better at the game.

4 Likes
17

I had multiple goals. (I generally need 2 or more reasons to do anything substantial.)

Excuse me if my above introduction had meandering sentence structure; I’ve rewritten parts of this 3 times, at this point. A second time for Reddit, to remove the (most interesting) controversial aspects that got my first post downvote nuked.

Right. I’m surprised that you’re the first to pick up on that. :slightly_smiling_face: I’m definitely thinking of focusing more on one role at a time, if/when I get to a position that I want to try to push higher than my alt account is currently sitting. Or maybe now, to try and lift my main’s DPS to gold, too. If that’s going to be possible.

To be fair, I’m virtually a Moira one trick, on support, in terms of play time. And I’ve been very focused on Zarya during the course of this analysis. Despite having a lot of time on most of the tanks, previously, and generally wanting to fill.

DPS is more counter-pick oriented anyway, right? I’d like to play more Pharah, but rarely get a healing comp to make it viable against chip damage. And I shied away from my best DPS, Mei, because she’s been too team play oriented, when I need to be just shredding enemies by myself, with higher damage output, to climb in silver reliably. Hence the Soldier, which I’m doing well with, but have a fair bit to learn, e.g. in terms of positioning, like when to flank or not, when to drop or not.

I’m also pretty ADHD in needing variety of play, to keep fully engaged and focused. So alternating matches between roles (as an experimental control measure, here) suited my mental.
⠀
⠀
[Edit: can someone else reply to this thread (with anything!) so that the forum will let me respond to @Eyevou, too? So many ridiculously suffocating, arbitrary restrictions on this forum. Stifling any new enthusiast contributors. :rage:]

1 Like
18

They most certainly do not. I’ve gone 4-1 and gone down nearly 100 SR from where I started.

19

To be absolutely sure I’ve not misled your understanding here: I threw only 18 QP support games, in total, ever. Not some before each session; after that, I only warmed up in QP on tank, DPS and Mystery Heroes.

Yeah, it’s bad enough that I might essentially be putting out misinformation (if people run too far with what I’m saying). I don’t want a lot of throwing on my conscience, too.

And QP stats related bias is still only one potential cause (or contributor) to possible comp matchmaker bias. But sure, it’s the only one I know how to test in a somewhat controlled fashion.

I think Cuthbert alludes to a second option, here (or I’m projecting my thoughts):

I.e. if having good stats works against you, as a strong player with the matchmaker counter-balancing you… Then it would be best to throw as hard as possible in unwinnable matches, to reduce your average stats.

I don’t like that at all; I always play to win. And double-guessing the outcome, early enough for this strategy to help, is going to be distracting and prone to becoming a self-fulfilling prophecy. Too many players seem to do this by default, anyway. For psychological reasons, I’d guess (wanting to be in control of events, even when they are not).

Anyway, talking of the devil…

Yeah. I think his polemic against MMR is closest on the business and political/philosophical side. But selective on the technical side, in touching on some possible truth of experience, but glossing over the limits of these. In order to pitch a more rousing call to arms.

Part of my reply to his thread:

Right. That’ll always work. Even in a paradigm where one is getting counter-balanced in matchmaking. If you play better than yourself. And that’s going to happen naturally, for (new) players, as they learn the game.

… Provided that the matchmaker’s player metrics are somewhat slower to adjust than the rate of your improvement. Or the degree of counter-balancing is capped.

But for those of us who know as much about how to play well (in theory) as is reasonable to know, who are mostly limited by physical ability to process the information in real time and execute on it… There’s maybe more chance of being stuck, a little way short of one’s ability.
⠀
⠀
⠀

Good call here. I’ve added a win percentage per map graph to my main sheet, and wow, heh: https://imgur.com/4yBWlc5

The maps I hate most, I have low win rates on (I wonder which came first). I felt like I was getting Hanamura and Volskya all the time too. Horrible chokes to break, 2CP at it’s best. It seems to be across all my roles, so no one hero strat to go try out.

In these matches, it usually feels like a lack of team co-ordination problem. And a lack of initiative. Compared to enemy team rolling together more.

Route 66 and Junkertown, too, (but the latter with relatively low play counts) - both I dislike as their first points are very open, sniper friendly maps (not that I have to worry about enemy snipers hitting me in low gold/silver, it’s just different, to most other maps).
⠀
⠀
⠀

Oof! Was that within the last half dozen seasons? There used to be a lot more weirdness like that years past… No disconnects/leaver penalty, on your end, of course?

Oh, and thank you for replying, so that I can now too. Gotta group all my responses into a minimum of posts though, apparently… :confused:

2 Likes
20

It was within the last 2 seasons.

1 Like