Integrity check here devs. Forced 50% real or not real? And silence is not a response.
Alternatively, you could actually go out and read the answer they’ve already given us.
9 Likes
Short answer, not real. Long answer:
13 Likes
49 posts
I’m not an engineer, but I can take a crack at this. The forced 50% win rate discussion is weird and complicated because it’s both true and not true at the same time, and people often inject their own biases very heavily on games to support what they think is happening to them, usually to assign blame for a loss to an external source (in this case the matchmaker) because it’s an easy thing to do (the matchmaker can’t defend itself after all!). First, let’s discuss what forced 50% isn’t.
Forced 50% isn’t the matchmaker putting potatoes on your team because you had a win streak. Frankly it’s hard enough just to find good games that there’s no way we would program such a thing, and there is no incentive for us to. We are incentivized to find matches of equally skilled players in our games because they are the most fun overall, which is what the matchmaker always tries to do.
The way it does this is by assigning a skill level to every player as they play our game called MMR. This skill rating is based on whether or not you win or lose games and nothing else (this is up for debate and some day we may find improvements in the future in how MMR is calculated, but there are many good reasons for this and almost every game uses this kind of system to determine player skill. I’d recommend watching some informational videos on how MMR/ELO works if you’re really interested in the nitty gritty). What the matchmaker does is tries to find games where each team has as close to an average MMR as possible while also having the MMR between each player be in as small a range as possible.
Now, imagine that everyone in the entire game has a MMR of 1, meaning that everyone has the same skill level. If we play an infinite amount of games, then everyone will eventually have a win rate of 50%, since their skill is the same. The 50% win rate of those players isn’t the matchmaker punishing those who have win or loss streaks, it’s just a natural consequence of the matchmaker working properly.
In real life it’s not quite so cut and dry since every player doesn’t have exactly the same MMR, but the fundamentals still hold true. As you win games, your MMR gets higher, which means that you will get into games with higher and higher skilled players until you hit your “ceiling”, where you start to lose games because you are the worst person in the game and bring your team down. Conversely, as you lose lots of games your MMR goes down until you eventually get into games where you are the best person on your team. The matchmaker getting you to a 50% win rate is a natural consequence of the matchmaker finding where you, over a long period of time, actually belong. Your MMR being an accurate reflection of your skill results in the natural consequence of you having something very close to a 50% win rate.
Keep in mind also that even 10+ win and loss streak games are perfectly natural in the realm of statistics, even when the true chance of something is 50%. These win and loss streaks are often what people point to as proof that something’s broken, which is simply not true. If you were to flip a coin 10,000 you would not get heads, then tails, then heads, then tails again over and over. Over those 10,000 times you would get each result half the time overall, but you would see streaks of getting heads or tails. This is also true for players who are at their appropriate MMR but still get streaks. It’s not that the matchmaker is broken, it’s just how things are. Also keep in mind that people have good and bad days, and sometimes decide to throw or to try extra hard to win. These behavior changes on an individual game level are impossible to predict and account for, which is why some games, even when they should be evenly matched, can still easily be blowouts.
This is just scratching the surface of how these things work, but again, for a TL:DR:
- The idea that the matchmaker forces you to win or lose after having streaks is completely false
- Having a 50% win rate over time does happen and is “true” in that it’s the result of the matchmaker finding your proper place in the game over time
17 Likes
I used to think I was the one with no streaks as every game I played was too complicated to win. After I read this looks like I had the luck to get win/loss every game I played without any streak, whether win or lose.
1 Like
This is true overall, but I do have one nitpick with this – win or loss streaks consistently occur far more often than they are statistically probable to (at least for me), assuming the even ~50% win/loss average.
If I did my math correctly, out of 100 games I should typically be seeing about 3 win streaks that are 5+ games long (same for loss streaks). That would mean maybe ~15-20 games should be winning streaks, likewise with loss streaks. That should add up to roughly 30-40 games being part of wins or loss streaks out of those 100 games.
However, I noticed that easily over half (50+) of my games were consistently part of win/loss streaks, and I’d typically see closer to 5 or 6 win/loss streaks every 100 or so games. Those streaks were significantly more common than they should statistically be.
There are lots of possible explanations, some more likely than others, but this is an odd discrepancy that indicates something is wrong. It may be accidental, I’d actually be more surprised if it’s not intentional – we know Engagement Optimization matchmaking is a problem in some other games, and with Activision’s track record there’s no reason to think something similar isn’t implemented here.
Or more like something influences the normal odds. And by something I mean you. Like switching Heroes, changed party sizes.
I personally play solo mostly and I usually don’t have many streaks.
Also good to know that what we consider a streak. Like having the same result for the third time is a streak now? Or the 5+ you mentioned?
What is the math behind the conculsion that it is “unprobable”, because I saw some wild things even in pure luck based coinflips. So a game like HotS can have more variety, since it’s not luck based and has many factors.
5 Likes
I find the last part kind of ironic because they already had given a response, as seen by Phaseshifter’s quote, but you didn’t bother checking.
3 Likes
Yes, that’s what that means. Something is skewing the odds away from what is expected or “ideal.” And it had a habit of being frustrating, so it certainly wasn’t desired, ergo “something is wrong.”
And no, before anyone misunderstands, I don’t think any artificial “forced 50%” is real. As AZJ wrote, “forced 50%” is a natural statistical occurrence.
I considered that possibility, which is why I referred to those games as my games, not games in general.
Perhaps my case’s statistical improbability is just an outlier in the larger dataset that is more statistically consistent.
Or maybe not. I know numerous others have complained about similar things on here, on Reddit, and in-game before, so we can’t say it’s definitely that I’m the outlier.
It’s certainly a possibility, and one I can’t judge one way or another on because we simply don’t have enough info on.
I decided against listing the half-dozen possible explanations for this, since most of them were just variations of “matchmaking isn’t perfect,” but “I’m just an outlier” would have been on the list if I had.
I only played solo, so party size is a fixed variable, and I typically stuck to playing the same hero for days and dozens of games on end, so that’s also a fairly static variable.
The best forumla I could find is N*p^k, where N is the opportunities to begin a streak, p is the probability of victory, and k is the minimum length counted as a “streak.”
With a set of 100 games, a 50% chance of victory, and a streak length of 5 or more, we get this equation:
95 x 0.55 = 2.96875
*Note that the N value is 95 because that is the last opportunity for a streak to start – any later out of the 100 games and the streak will be less than 5, too small to qualify.
If that is the correct equation for this, then we should expect, on average, streaks of 5 or more wins/losses to occur 3 times in sets of 100 games. They may occasionally occur more or less often than that, but that’s the norm.
Disclaimer
Honestly, I’m not sure if this equation is right, but the only possible issue I can spot is that it may be calculating streaks greater than 5 as multiple streaks of 5 rather than single streaks of, say, 6. This is because a streak of 1-6 can be counted as 2 streaks of 5 – as 1-5 and 2-6.
However, if it were doing this, that would cause it to actually overestimate the likelihood of streaks, as streaks greater than 5 are being counted multiple times. That means the actual likelihood of streaks should be even lower than I initially estimated, which would mean something is even more wrong than it first appeared.
If anyone knows a better or more accurate equation, please share it.
Multiple-streak probability is not easy to figure out.
1 Like
Or even being unlucky and getting someone afk, trolling etc.
That’s why we need a better report system, so the MM can work better.
1 Like
Also possibility that there is “something wrong” with the MM indeed, which is the low population which then causes weird things thanks to how team averaging works with the given population.
6 Likes
Yep. That’s actually the lynchpin issue with the matchmaker at this point – low population exacerbates all the other issues, and even if the matchmaker was designed perfectly, it would not function perfectly because it doesn’t have what it needs to do so.
1 Like
For sure. I’ve already said this before, even the most perfect matchmaker won’t work well with low population. Even worse considering you need 10 people in a match.
1 Like
I don’t think every win or loss breaks down to simple luck of the matchmaker, your performance, or lack thereof is the only constant.
Many players who post on the forums will blame anything but their own performance for losses, and never consider if some of the wins they had were really deserved.
Personally I know a few times I made it into Diamond during the old HL were due to lucky placements. At least that was my feeling as I had been matched several times with players in Diamond and Master as the MM had given up making them wait 2 hours and put them with a team of mostly Plat players.
Some of those games I felt like a passenger, or at the very least that my personal performances was only a very tiny factor in the wins. I believe this was a sound conclusion, because each time I didn’t remain in Diamond for long and fell back to my usual Plat rank.
It’s true that you can underestimate your performance, but going by the forums and the players I see who blame in games, the opposite is far more common.
5 Likes
This is an expected value, not the absolute and real come out. The Formular is based on statistics with a big amount of data.
To proof that fomular, you would have to do n-times of 100 games. The result would be, that it will average out around 3. But a single test with 100 games is full of varieties, which have to be eliminated by repeating the test n-times.
2 Likes
Yes, as I said, that’s the average.
However, my games have consistently shown more win and loss streaks than the expected average over months and years, and thousands of games. If it were just once or twice every now and then, those sets would be outliers, and outliers happen. But it’s consistent enough that we can’t just brush them off as outliers. Normal statistical fluctuations don’t seem to account for what I’m seeing, which is my point.
But this formular is just a theory , if all the factors were always consitent and never change.
But there are too many factors that are not considered by these formular.
For example: If you have a very good day and you feel very good it will translate into more wins, because your focus will be better.
On the other hand if you feel tired or stressed, your focus will be worse and you will loose more.
Also the other 9 players will change every game and for them applies the same rules as for you.
Also if i play my 33% winrate hanzo or my 60% winrate tracer will change the result of the formular, because p is not 0.5 anymore.
The formular is just a very simple approximation and maybe will give a good result for something like a coin flip, where every flip is the same, but it will never reflect scenarios with variation properly.
1 Like
It pretty old to be fair. I only know about it because I read it when AZ wrote it.
I don’t know that there’s something abnormal about streaks. If everyone’s games were: win-loss ,win-loss. It would feel rigged. We all have better win rates in some maps than others, that makes a difference. Or your best hero could have been banned, or you could just luck out into horrible or great teammates.
1 Like
Yes, that’s the point.
The formula basically says, “If we always have a 50% chance to win, then out of 100 games you should see approximately 3 winning streaks on average.” It’s an ideal situation.
If what we actually see consistently deviates from that average, then that means the formula, while accurate in an ideal situation, is not accurate in reality.
There must then be some cause for that deviation, which is my point. There is something going on that is skewing the result we see from the result we “should” be getting.
There are lots of hypothetical factors that could account for this. Since it keeps coming up, I’ll just go ahead and list the ones I mentioned skipping over earlier.
- Low player population forces the matchmaker to regularly create skewed matches for at least some players for a period of time, then has to readjust to their MMR being higher or lower than their skill level.
- The matchmaker regularly over- or under-adjusts to MMR changes on its own.
- The matchmaker is just poorly designed and accidentally creates streaks some other way.
- The matchmaker is intentionally designed to create streaks via a “mentor-buddy” system.
- The matchmaker is intentionally designed to create streaks via an “engagement optimization” system.
- My dataset is not just composed of outliers, it is an outlier in its entirety, and most people don’t have this issue nearly as much as I do.
Any number of those could be true at once, and maybe more I haven’t thought of. At this point, we’re just speculating. Something is off (again, at least in my experience), but we don’t know what.
These variables would not be consistent, though, and over enough games would cease to matter.
That would increase the number or length of the corresponding streaks, yes, but it would not explain how the other streaks still occur so often.
I have a roughly 56% winrate with Zul’jin on my main account. Going by the formula, that should mean I should have been seeing ~5 winstreaks every 100 games (which matches what I see in reality), but only about 1.5 loss streaks (which is way lower than what I actually see).
Win or loss streaks are perfectly normal. The abnormality I’m pointing out is their length and frequency.
Out of 2 coin flips, you have a 25% chance of both of them being heads or tails (there’s a 50% chance of them being different). That’s likely enough you’ll see it happen pretty frequently, so if you flip a lot of 2-in-a-rows you shouldn’t be surprised
Out of 5 coin flips, you have only a ~3% chance of all of them being heads or tails, only ~1.5% for a streak of 6, 0.78% for 7, etc. Those will pop up occasionally if you flip a coin enough times, so if you flip one or two of of those out of enough attempts you also shouldn’t be too surprised.
It’s when we combine the frequency and length that things start looking weird. When that theoretical 3%-or-lower chance keeps turning up again and again, that’s a sign something isn’t quite right.
1 Like
If we knew that they were abnormally long for everyone ( And I’m not sure how long constitutes abnormal) we could figure it out. But we’ll never know.
3 Likes