Streaks in Overwatch, Simulation and Analysis and Data

How would they make more money from a fixed system?

You realize your argument is less warranted than a fair system?

Also there are other reasons to keep the matchmaking algorithm hidden. The largest being that if it was released then that gives players the ability to abuse it. If the algorithm is fully transparent, players can manipulate queueing up games to fix matches themselves.

If you trust in what a company tells you, then you don’t need transparent matchmaking.
It is not a conspiracy theory, the rules by which the games are built and the players are rated are secret. That is a fact.
For what reason, if not for money, should someone do such a thing? This is not just about Overwatch, almost every company does that.

This mechanism is the whole secret, so they didn’t say anything at all.

I never said it was fixed. It is overly controlled and not transparent.
I suspect that in matchmaking more attention was paid to keeping as many players in the game as possible instead of evaluating them correctly and honestly.
This does not mean that you can’t climb, but it is intentionally slowed down if you are not a world class player.

The matchmaker does much more than putting teams together. Players and game behavior are profiled, game sessions are analyzed and players are put into a flow state. (this can have similar effects to an addiction, an addiction can develop from this)
Streaks serve an important purpose in this context. This has nothing to do with luck at all.

It is definitely an advantage if you know how many players will play, with which probability, and at what time of the day.
( => e.g. scalable server costs )

If you know how the distribution of players to the different ranks will change within the next 7 days, matchmaking can be optimized accordingly.
( => reduce computing power, shorter queuing times for players)

Most likely there are other factors that play a role in matchmaking. For example, how likely Blizzard considers it is that you will not follow the agreed rules.
Such players could be subdivided so that they do not disturb “good” players too often in the course of the game, so that they can stay in the flow state.
( => similiar to Valves rainbow of trust
https://www.gs4u.net/en/news/1070-valve-fights-against-counter-strike-cheats-with-artificial-intelligence.html )

Allegedly, the developers have worked on the matchmaker component for 2 years. So we should assume that just about everything that can be considered useful has been implemented.

Implying that blizzard intentionally rigs matchmaking to create flow states is ridiculous though, because they naturally occur. You naturally go on streaks. Take a player with any record, x wins and y losses, they win p=x/(x+y) percent of their games. They’ll have a distribution of wins and losses, where the player will win any given game will p probability. You can measure each streak of wins and losses, and the empirical percentages of each streak should match up with a geometric distribution. You would have to compare those two distributions otherwise it’s just speculation.

1 Like

no you can’t thats wrong. what you describe can be seen with randomly distributed data over a quantity of several million records.

this has absolutely nothing to do with streaks of players that change from winning 10 games in a row to losing 10 games in a row every 2-3 days.

Also it is not rigging if it is a core function of the system.
That’ s why no developer has lied about this yet.

I mean like where’s the proof right? You would have to show overwatch players go on constant back and forth win streaks.

And yes you can, that is actually exactly how you would test your hypothesis. Compare the empirical results you collected to a geometric distribution. If they differ, then you are right and blizzard matchmakes flow states. If they are the same, then it was all just speculation.

1 Like

As a player I am not in the right position to independently and reliably test something like this. A few thousand records are not enough, pursuit.gg was
probably banned for this reason. The data sets would have revealed too much.

I know what you are getting at with the geometric distribution, but I don’t know why many people think that it would apply here, because the conditions are not given. Just because something has the same form, it does not mean that it was created in a definite way. One of the few things we know about matchmaking is that it is not completely random.

Any development studio that claims to be serious about matchmaking should use a fully transparent process. There would be plenty of opportunities to do so, after all, even currencies and contracts now work on the basis of transparent and decentralized networks.
For the reasons mentioned, no efforts will be made in this direction.
And till then I will make every speculation, which comes to my mind and seems logical.

This post is centered too heavily on averages. The reality is that it illustrates a possibility for a fair system to create these results, and nothing more. Wins/losses are not a coin flip system, if you assume skill to be accurately represented then each loss should increase your chance of winning the next game by a small amount(as it’s moving you into a lower bracket where your personal influence is inherently higher). By the time you’re at a streak of 8-10 losses, you should be playing with a subset of people that’s notably easier to play against.

The ‘data’ used in the post seems to be sloppily self recorded, parts of it are missing and it has comments like ‘undocumented loss streak’ and notes that are generally indicative of the person recording it not being very clever. I am sure it’s not Kaawumba’s own data.

While a 40-60 chance may be predicted, that does not mean the individual match in those circumstances is really 40-60. I imagine most stomps would play out the same way if the teams played several games in a row; there are identifiable players who are holding their teams back or carrying. Thus, any factor that can be used to make it more likely you get those poor players in the interest of maintaining a perceived 40-60 range, is rigging.

As ANTCRSTM continues to mention, Blizzard is a company with their greatest interest in profit. Promoting engagement at the expense of fairness is a logical decision for them to make, and they’ve been very tight-lipped about every aspect of matchmaking. Let’s not forget that they banned pursuit, added private profiles(with a default to on), and things of this nature.

When you get grouped with someone who has a 0% winrate over 20 games(has happened to me), it’s pretty obvious that person is doing something to force their team’s chance of success far below 40%. How can matchmaker be confident a game is in the 40-60% range while placing that person on a team? Surely the best they can do is give them 5 teammates that are seemingly excelling, and even then the match is likely not going to reach 40% win chance. Fortunately, in most cases you have absolutely no way to observe things like this, because profiles default to private and most people don’t even notice it’s a setting to begin with.

2 Likes

I think the game is trying to tell me I belong in mid Gold or something… because any time I hit 2.8K on tank I just drop back to 2500 or so… even now at 2600 Tank I seem to win 1 for every 3-4 games I play… it’s bad, and not fun. Guess I’m not at the rank I need to be to win 50% of my games at a minimum? My support is 2805 I just placed but it’s hilarious because I basically only play Moira or Lucio.

My Tank is way more versatile and I definitely deserve to be 2.8 on Tank and not support. DPS… ugh started 1-8 the other day and ended up 5-10 for the day. Game just isn’t fun. At least I haven’t had leavers or throwers. My DPS placed 2452 but I’ve dropped to 2370 something.

@kaMMakaZZi9 I’ve personally experienced the SR system to be quite accurate. I think, if you always play with the same quality, a skill area of 100SR is normal but I feel, it’s rather 50SR. That would mean, when you have 2700SR and you drop to 2650SR when you did no mistakes you usually don’t do, you shouldn’t interpret too much in these losses. You should always use your in-game performance as the measure and not probabilistic abstractions. That would also mean that when you drop 200SR or even 300SR, you did do significant mistakes in-game you haven’t done before when you were stable at the previous rank. Now comes the trick, a mischievous question tries to mislead you: I don’t feel like I played worse, how could I still drop that much? How you play comes from your state of mind. The mistakes you did in-game went under your radar. It might come from a lot of things. You could be distracted by a mistake a team mate did and then you tilt. You could just be tired after playing for 1 – 3 hours. You focus less on your own game and don’t realize your mistakes you usually would’ve realized or even wouldn’t have done in first place. I know that problem from myself and I fixed a lot of that. I’m still not perfectly over that after months to years because behaviour can be difficult to change. That you say these things “aren’t fun” is a strong indicator you have similar issues that keep you from the game.

There are situations where Moira is much better than Ana, so playing Moira or Lucio or placing 2805 with them isn’t hilarious. If it was that hilarious, how would you explain Moiras in way higher ranks?

To conclude all of that, Overwatch is a highly complex game where you don’t always see everything. Understanding needs a right mindset. You shouldn’t get hung up with irrelevant thoughts.

The original version of this post was on the old forums. It had a fair amount of discussion, but unfortunately Blizzard nuked it along with the old forums.

Try rereading the original post. It looks like you got tired and stopped in the middle somewhere.

As said, the data came from Des and Porkypine, not me, though I appear to have made a mistake and only linked to one of the datasets. Yes, there is some missing data, but “generally indicative of the person recording it not being very clever” is unnecessary.

And here is where you follow him into conspiracy theory. Really, if you believe that game companies in general and Blizzard in particular are lead to evil by the profit motive, then it is up to you to not play games.

The game makes its best effort to put players together fairly. It does not do a great job with new or toxic accounts. See https://www.reddit.com/r/OverwatchUniversity/comments/aatezy/why_match_quality_is_frequently_poor/ for more.

1 Like

It’s dismissive and childish to call something like this ‘conspiracy theory’. Similar things are currently in use by other gaming companies, by social media, by basically any business who’s profit depends on keeping users engaged. Blizzard has multiple patents of the same ilk. It isn’t a conspiracy as much as typical behavior in this field.

While you’re clearly very talented in the realm of statistics, you’re still overlooking and refusing to acknowledge this very basic fact:

Your statistical models decisively illustrate that a fair system will have bad games and SR swings. They do not, in any way, prove or support the idea that the current system is fair.

Further, the simulation is still relying on the basic assumption that it can decide a win by comparing only SR(with inconsistancy modifiers) against a beta rate. You model smurfs very strangely with a chance of SR reset across the board at a rate of 1/1000. Not only is this number obscenely low, but taking smurfs from all SR ranges is poor representation as there will be an inherent bias in smurfs from ranks above or below the normal. Why make an alt just to continue playing in the same SR?

In reality, this is not the case. Smurfs are not just resetting their SR and playing as intended until it balances out, they are cycling intentional throwing with hard carrying. Someone throwing in an effective manner is going to reduce winrate by far more than your prediction allows. They can give effectively negative value by adding useless damage between encounters to ensure the enemy healers get ult faster, starting fights in voice chat, or actively feeding health on a hero such as roadhog or ball. Even if they are given a huge inconsistancy, this is not likely to account for how much negative impact they are able to produce. Further, from what I understand, you apply the inconsistancy indiscriminately, rather than assigning a greater inconsistancy to smurfs.

Finally, it seems that your simulation indicates games are worse than they actually are, while simultaneously underplaying the factors that result in the worst games. This could be interpreted as a situation where rigging is actually making the game more playable rather than less.

I do not accuse such companies of evil motives.
It is quite plausible, it is worth doing it, it is exactly what a successful company has to do because everyone else is doing it. I would like to see more transparency on this subject, but I also understand if they prefer never to talk about it seriously.

Please take a look at the definition of a conspiracy theory. I have no agenda, I do not care about my rating, nor do I expect it to be raised as a result of this discussion. :laughing: And I am also willing to change my opinion on this subject.

For the sake of discussion, it is not very helpful to call others conspiracy theorists, as this is a derogatory term.

Take a player’s sequence of wins and losses.
If that sequence looks like a geometric sequence then it was randomly distributed. Like: WWLWLLLWWWWLWLLWWWWLLWLWLWLWLLLL
If that sequence looks more like what you said, like: WWWWWLLLLLLWWWWW, then blizzard is doing some shenanigans.

Test it out on your own win-loss record, I could be wrong, but I bet you it looks more like a geometric sequence rather than what you are implying. Streaks naturally occur in a geometric sequence, which is what the OP was getting at. So the presence of streaks does not indicate blizzard is intentionally matching you up for streaks. This is why its a conspiracy theory, because it’s all speculation.

Also being on a hot streak and being tilted are real things. If you get pissed off from losing and keep playing you will probably keep losing because you have low morale and your head isn’t in the game. If you are on a hot streak you have high morale and your head is in the game and you will probably win more. This happens in all games and all sports, not just Overwatch. This is also why good players keep their cool.

1 Like

No, it doesn’t work that way.
This point is not thought through at all, I have the feeling that you are simply repeating something someone once suggested in the forum.

No interest in correcting every naive misconception. If you want to get a real answer, maybe try https://math.stackexchange.com/ https://stats.stackexchange.com/

Mind building on that a bit more? Rather than just saying no it doesnt work that way lol.

1 Like

You don’t toss a coin when you play, and no one wants you to toss a coin.
Everything important has already been said in this thread.

Never said you are tossing a coin.
Probability distributions, like a geometric distribution, are used to model events in nature.

The problem is you aren’t building your theory on any evidence or backing it with any concepts. You are just denying what everyone is saying, even when they bring up evidence and concepts observed in nature.

1 Like

It is mindblowing actualy. Some people will always believe crazy conspiracy theories instead of actual evidence.

1 Like

:man_facepalming:

There can be no evidence until it is fully released. That is in the nature of things.
Where is the evidence?

I do not deny anything completely, I just say that within a closed system it is quite possible to create patterns that you perceive as random. These patterns can be constructed.

The results are not so important, what is important is how they were produced.
That’s much more important in this case.

The whole approach is not valid in this way, one cannot draw any conclusions from it.

It is a pity that it is not possible to talk about it objectively.
Here comes the next one talking about conspiracy theories.
At this point, I don’t really care what you claim to know.

You do not know how to create a scientific evidence.
And certainly you wouldn’t know how to evaluate it.
Anything beyond discussions by hero-nerfs is probably too much for this forum.
I’m out.

More of the same. Just denying what people are saying without contributing to the discussion. You ask what evidence, but have been ignoring the evidence presented by the OP or that every game and every sport has streaks. What you are saying are general things that can be applied to any game. Yes it’s possible, yes there are probably games that do it. The only way to test it is to write out your sequence of wins and losses and see if you consistently see it playing out that way.

1 Like