A fair system wouldn’t try to make you win or lose games. It would try to match teams relatively evenly so that who wins or loses is dependent on their performance within that match. There’s no point trying to win when the game is purposely pitting you against players significantly better or worse than you are to maintain your W/L ratio.
Players make the most improvement to their game when they are playing against players that give them a challenge, but don’t overwhelm them. That was the point of SBMM when it was first introduced way back in Halo 2. And it worked. Matches felt mostly fair, with the utter pubstomps being the exception to the rule.
In Infinite, pubstomps are nearly the entire game. You get matched evenly sometimes, yeah, but it’s such a rarity that the system needs a rework.
The player is far more binary than they will admit. And that’s the source of the frustration.
Once you hit your skill ceiling you are very binary. Beat teams ranked below you and lose to teams ranked above you.
This doesn’t make sense. You can’t retrospect it.
Your expected kills and deaths presumably come from you KPM and DPM vs that rank. If your KPM against the opponent is 1.5 and the game went for 8 minutes you will have been expecting 12 kills.
How you performed in the game is the bit about actual kills.
The skills of the teams you play are fixed (around your rank). The time of the match is fixed. Your KPM and DPM are fixed (by your skill). So of course the graph is going to go up and down. There is nothing here that can really move out of range.
It’s bound by your skill. Which despite all hopes and dreams, doesn’t make big changes.
There were a couple. They popped in the discussion around Menke’s GDC talk on the same. Where he showed players were more likely to log out of the game (for more than 8 hours) if there was a one sided result. More so than losing streaks or lag. It was that feeling of not being having a chance that outweighed everything else.
I’ve had a little run back through - but can’t find the thread(s) specifically.
Again, they are just using your past data, along the lines of KPM, to work out how many kills you are expected to get (for the length of the game).
You are still in control.
Get better at the game and your KPM vs harder opponents will improve.
People hit their skill ceiling and stop progressing.
They get frustrated.
So they take it out on the system. And the CSR changes, being so poorly implemented, are the easy target.
At least we have found another scape goat in terms of “expected” kills.
It’s one of the worst CSR systems… I’ll give you that.
I don’t get how the game is purposefully setting you up to fail?
You lose the tougher games because you aren’t good enough. You win the softer games because you are better. You break even in the 50:50 ones.
I get it that the lower population is causing some issues. Teams are spread in ranks more than they should be. They are still matched on MMR though. But it does tend to create blowouts (in both directions).
I guess that gives you a feeling of losing control.
But this isn’t deliberate. Same thing happened in H5 when the population dropped. And fingers crossed Infinite can come back from here.
And in the context of this thread - your KPM vs an opponent - shown as an “expected kills” is not data to get upset about.
When you lose a game vs a harder opponent… look at the difference between expected kills and deaths. This is essentially the benchmark of improvement you need to get to that rank.
Where could you have got the kills you needed. Avoided the deaths you should have?
Is the problem more kills (offense) or kills (defence).
It’s potentially a great learning tool to see where you need to improve.
I play a few games before reset each day in the morning. I devastate.
My little bro in law gets home. We play 1 game. We wreck. The next 5 games are as so:
Game 1 is a horrible smear. We loose 18-50.
Game 2: VERY CLOSE 42-50
G3: 10-11 (LANDGRAB)
G4: Win
G5: win but BARELY.
Is it fun? Not always. But the close matches, those atleast don’t hurt .
This system is a pain but it’s a lot better then playing in classic lobbies where clans and try hards of every caliber rack up 87% of the lobbies kills and you were lucky enough to figure out where he was camping ONCE.
Just like with every other aspect of this game it needs refining, but it is a step in the right direction
They aren’t “a bit tougher” and “a bit weaker” they are significant jumps. My CSR puts me around platinum VI. But I NEVER see other plat or even Diamond players. I see mainly Onyx, or Silver/Golds, because the system is so focused on giving me wins/losses it doesn’t put me in games that are close in skill.
If you’re a Plat player, seeing Silvers or Onyxs outside of mixed-skill parties should me extremely rare. Honestly you shouldn’t ever see an Onyx unless the pop is just that bad at the time.
Halo 3’s system was extremely simple. You would ONLY match with players within 5 levels of you, up or down. If you were in your 30’s (est plat) you would literally never see a General unless you were in a party with a 45+ player. And if you saw someone in their low 20’s or below, it was because they were in a party with someone closer to your skill level.
Combination of low pop and squads of friends (or manipulators).
It’s getting worse, I agree. Especially out of peak times.
And while I agree it’s a problem - it’s not part of some conspiracy for 343 to lock you into a rank you feel you don’t deserve (which is the premise of threads like this).
It was better in S1. And it was pretty good at the beginning of S2. Not so much three months in. I suspect things will improve again with the CSR reset and the attraction of ranked doubles in a week or so.
But it will drop off again after a short while.
And Infinite holds the same premise.
I’m sure the match maker would love to pull players closer in rank to each other.
And we know it’s 343’s wish as well - they have already started the process of putting limits on who can squad up together.
But at the moment everyone is the victim of low population.
Noone is claiming that they are locked in a rank in this thread. The complaint is that the system chooses whether or not it’s your turn to win, or lose a game before it finds players to give you that win or loss. It has nothing to do with players trying to push past their CSR, the problem is that your CSR is used incorrectly for matchmaking because the system is trying to give players a 40-60% won/loss rate by stacking the teams, as opposed to attempting to evenly match, and distribute them onto each side.
Even if the issue was just population, matchmaking should at least be trying to put the “top players” in a match on opposite teams when they aren’t partied up. It doesn’t so this though. It assigns you a win (or loss) finds 3 other players who’s turn it also is to win (or lose), then finds 4 unlucky (lucky) souls who’s turn it is to lose (or win) that form a team that’s supposed to lose (or win) against your team. All in the name of giving even low-tier players, and the creme of the crop a near-50% win/loss ratio.
Where your W/L being somewhere around 50% because of how evenly matched most of your games were before. It’s 50% for the opposite reason, because matchmaking sets your team up to win or lose games by putting you against significantly stronger or weaker players.
The incentive to care about your wins and losses is completely removed from the equation, because no matter how hard you’ll try you’ll never win a game you’re supposed to lose, and the games you’re supposed to win won’t even require you to try. That’s how large the skill differential is in the average match in Halo Infinite.
At it’s core - it’s the frustration of feeling that the system is conspiring to stop you from progressing.
But we are in this exact situation because there aren’t enough “top players”.
It can’t make them wait forever… so it throws them into matches the best it can.
Tin foil hat protocol engaged.
The alternative is that it just finds you the best match available. It continues to match balanced teams albeit with unbalanced individual ranks. The results are still 50:50 - just with the frustration of results that may not end up being as close as you would like. The variance of skill levels of the players increases the variance of the team output.
To some degree you are probably right here.
The increased team variables can dilute your individual performance.
But you can still move.
Last week I mucked around with my aim settings and struggled badly. Over the course of a few days I dropped from mid D1 to mid P6. It was an appropriate drop off given I was performing badly.
With some more tweaking and a bit of practice I started to improve. It took me about four days to get back to mid D1. Again - an appropriate climb given my performance.
I was starting to play well - and I’m sure I could have climbed to D2 if I hadn’t had to go interstate for work!
The ranks in my game are wide as well. But it all comes out in the wash. Just stop micro-analysing the CSR after each game. Stop looking for a conspiracy in the match making. And definitely stop trying to find demons in something as simple as using KPM to calculate an expected number of kills.
No, you cannot dismiss something that isn’t a binary distribution with a binary counter example and move on with the argument like you’re more correct. You’re example literally means nothing in this context. You can have a great KD and lose or a terrible KD and win. We’re not talking about wins and losses, we’re talking about individual performance.
Predicted kills can be any non-negative number. Predicted deaths can be any non negative number. A Binary analysis and counter point is meaningless.
If it were truly random then EK and ED then there would be no pattern in the predicted kills and predicted deaths values. It would be all over the place, with occasional streaks, but mostly just an obvious independent scatter of values.
There is a clear and perpetual wave pattern over time. There’s at least enough of a pattern to be worth looking more into and not dismissing.
(Bolds were typos)
The point of this thought, is that with 8/12/24 players in a match and a few hundred in the eligible match pool, the system can probably configure a game where the chance at winning is close to 50% and also enforce the EK and ED streaks
You have a skill level. You will beat people with a skill level lower than you. You will lose to those higher than you. You will break even with those at a similar level.
It’s even predictable between divisions. Gold players will go 25:75 against Platinum players. The divisions are all one standard deviation from the mean on the population curve. The six divisions covering 98% of the population (plus or minus three standard deviations from the mean).
You can.
But you can’t pluck out an outlying result and give it significance.
Overall the system is self correcting. If you are improving your KPM and DPM the chances are you are ranking up as well. Your KPM and DPM will then drop.
And don’t fall into the trap of trying to interpret the trends with KD. That’s not the metric. It’s KPM and DPM - which are much better reflections of your ability to win a 1v1 at that level.
The “predicted kills” is simply your KPM vs that level of opponent - extrapolated from past game data. And at the end of the game it takes the time of the match to work out your “expected kills”.
I am not putting any meaning on it then that.
It is simply a reflection of your skill rank vs the opposition skill.
Again the expected kills and expected deaths are a reflection of the opposition’s skill (relative to yours). If the opposition’s skill goes up and down - so will the expected number of kills and deaths.
The waves you are describing is largely due to the length of the match (expected K = KPM x mins). If you were to plot KPM vs the skill of your opponent it would not change. There is no manipulation here.
I’m sure it’s doing it’s best. Within the constraints of population and wait times.
It’s not enforcing them.
They are not made up targets to set you up to fail.
It’s a measure of what YOU have done vs that level before. It’s a bench mark that YOU have set yourself.
Use it as a tool to guage your performance. It’s likely to be a better tool than KD etc. And that’s probably why 343 have put it up there.
If you are getting more kills than expected (and less deaths) then you are perfroming better - and probably in the process of ranking up.
On it’s own, and as a generalisation, not overly useful.
You have to look at all the different parts.
The actual ranking system. ie. TrueSkill2 / MMR is solid. Plenty of evidence to back it up. If you want to replace it then you need to present a better option. And just saying the MMR is too complicated is a very silly reason to replace it.
Match making sucks a bit now. But that’s because it is trying to balance lower population with wait times. Again, if you want to fix or replace the system - you need to propose something that deals with the current issues. Good luck.
The CSR system, as implemented by 343, does suck though. I agree. I’ve been very critical of it myself. Zealously. And I think I’ve been very clear in what I would like to see instead. I think it’s important though to make sure you understand what it is trying to do. It’s frustrating as hell - but it’s not a deal breaker. You are still being ranked accurately in the background.
And as for this actual thread - the expected kills and deaths? Definitely doesn’t suck. It’s just your current KPM applied to the time the match took. No more. No less. It will likely be a valuable self evaluation tool to guage where you need to improve. But it’s not a conspiracy to force you to get ‘x’ number of kills and trap you into a pre-definied rank.
They keep adding features to make the game feel more competitive than it should be. Halo first and foremost was always about couch co-op and fun multiplayer.
FUN. In their bid for success and recognition, they have forgotten the core aspect which makes games actually desirable. How tone deaf can a company get to forget the main reason the product they create exist?
If the variable being estimated is binary then it’s a binary outcome. One outcome may be more heavily weighted under the influence of any number of variables, but it’s still binary. You can argue that it’s a bad measure of success or one outcome is more likely than another, but it’s still binary.
What I’m referring to is years old and made the rounds roughly at the same time as Activision had their controversial patents spotlighted.
Aren’t you actually proving what’s being claimed in the OP?
If your skill is being used to predict kills and deaths during a match against other players based on their skills, and the result of those predictions end up in a wave over time, then an assumption can be made that the game does indeed seek out players who you most likely will get lower kills and higher deaths against, and vide versa.
Not if it’s designed not to, but with different goals in mind set by those who are in charge.
I.e, more player engagement, longer play times, etc.
Reminder that this is the developer under suspicion of having the active challenges directly influence game mode picks and weapon racks in the Matchmaker.
The company claiming “players first” and having done basically anything but that.
Or you could assume that the game is seeking out a mix of opponents harder and softer than your rank.
And that your KPM is better against softer opponents and lower against harder opponents.
And finally that the “waves” are then exaggerated as they also depend on another factor - the length of the game. And this is only known after the game has finished.
I don’t get how a random flow of softer and tougher opponents can be interpreted as deliberately seeking out players to manipulate the result?
You would expect a natural correlation between the quality of the opponent and both the expected result and your KPM (vs that rank). The expected number of kills is calculated directly from the KPM so would have exactly the same correlation.
The same suspicions that certain people get different weapons in Fiesta, or that some people have “homing” grenades that seek them out, or that lag and desync are specifically ramped after after a run of wins.
It wouldn’t be hard to collect and present hard evidence for a lot of these things. I mean it would be very easy to show that your challenge weapons spawns less in Fiesta. I’m sure people have tried - but we don’t see anything posted. Because there isn’t any.
Can’t believe I’m doing this… but let’s literally flip a coin 10 times;
H T T H T T H H H T
Interesting tid bits;
I had to ask a colleague to borrow a 20c piece. Who carries dirty filthy coins in the middle of a pandemic.
I was actually getting nervous on the possiblily of the 4th H on the final flip.
Now, let’s imaginge you are D2 and the match maker can pick either a soft game D1 or a hard game D3
So your opponents are now going to be;
D1 D3 D3 D1 D3 D3 D1 D1 D1 D3
Your KPM is graded against your opponents. Let’s say it is 1.5 vs D1, 1.3 at your level D2, and 1.1 vs D3
So your KPM goes
1.5 1.1 1.1 1.5 1.1 1.1 1.5 1.5 1.5 1.1
And because you are playing consistently your results will tend to go;
W L L W L L W W W L
So you can clearly see that the flux of opponents will correlate beautifully with expected KPM and the result. They just happen to be inexorably linked.
The correlation just looks noisy because your KPM is multipled by the length of the game after the fact to give you the “expected deaths”.