So it must be true!!!
Depends on the actual population… I’ve actually just said this; if we had a HCS playlist it would get rotated.
1 Like
The hidden 20 levels. No lol. As soon as you hit 50 your wins started stacking. You had to achieve a certain amount of wins per level for the most part. You win 5 times then lose 5 times your right back. Not science my friend. Unless you lived and played to compare you wouldn’t know. It’s ok to try contextualize. Hypothesis or not it is far from the same. I’m also not saying this as being at the same level I am from back then because I’m no where near it and much older now. But I do notice a major flaw in this system. There’s to many variables in an actual game to make a true skill system. If you don’t understand that I apologize, but there’s a difference between being book smart, and being street smart. You can’t just research what you may think about something.
This is pretty accurate believe it or not. I have many friends who couldn’t pull it off and they were 50 in everything else. Level locked meant winning more games as your experience in a playlist made it longer to level up. I got a so called level locked account to an mlg 50 for someone and it took darn near 15 to 20 straight wins at 49 just to get it. Others were common to be at 42 ish and 50 in other lists.
LOL as in they didn’t exist?
But they did. Josh was quite open about it. The old ‘50’ was the same as a D3. At least in general play lists.
MLG was obviously a high skilled but low population playlist. So the D3 there may have been a lot higher than the D3 outside.
But that doesn’t disprove their existence in MLG. I suspect they weren’t if the relative MMR of a ‘50’ was so high - but we need more info. Where’s Josh when you need him 
Which didn’t appear particularly onerous? You just kept winning to get the EXP to achieve your military rank (as long as you had the prerequisite skill rank to go along side it). But I guess 500 or so wins is a “hurdle”.
No. Not science. And not much different to what happens now.
People get upset with the variable up and down of the CSR - but for the most part your division and tier are pretty much the same at the beginning of the week and at the end of the week.
You just notice the blips more because of the scale.
The MMR changes are pretty similar. But it’s only the CSR scale and rules that make it look different.
Been playing Halo since 2001. Not that it really matters.
But trying to understand and work out the system is important for going forward.
Otherwise you fall into the trap of just going backwards.
Too many variables as in actual ranking MMR?
Not really. Just beat sides better than you and don’t lose to sides lower than you. And keep up your KPM. Easy.
As in the CSR?
I agree that’s a bit convoluted. But it’s not that hard to follow if you stop trying to describe it’s behaviour with metrics that it doesn’t use. eg. K/D.
As in the current state of the game / match making?
That’s not really the fault of the ranking system is it? What system takes into account desync etc.
Depends on how you define said skill.
Most sports or games are complex. With many nuances. Some out of control of the player. But what is interesting is that virtually all the ranking systems out there boil these variables down to a win vs a loss.
There is. And both complement each other.
The only danger is if you use one as an excuse to exclude the other.
Why not?
And it’s interesting.
Obviously the general skill of that population is higher.
But it doesn’t necessarily explain off the bat if the structure of the CSR was any different. There still could very well be hidden levels. The same number? Fewer?
And it doesn’t necessarily advocate that it would work any better than the system we currently have.
I remember the furore of level locking.
It was ridiculous. So many people had to start new accounts to level up.
But it’s not a problem with TrueSkill2. It has an MMR that is left relatively volatile - and refreshes with each Season.
Thanks for the input
It’s a super fun system to try and understand. I’ve just given up on it. You would think my true winning percentage of 50% would happen consistently but when I don’t play with teammates who communicate it rarely ever happens.
Can we just go back to the traditional ranked number 1-50 system like Halo 2 and 3?
1 Like
I’m curious–it’s getting pretty common to see D1/2 or even platinum folks with Onyx nameplates from previous seasons. How do you think profiles like this https://halotracker.com/halo-infinite/profile/xbl/Disgamer9/matches?experience=ranked&playlist=edfef3ac-9cbe-4fa2-b949-8f29deafd483 are getting stuck at low ranks like D1? This guy and a couple others beat the wheels off me and my platinum/diamond team, and they were playing at what I thought to be Onyx level–and lo and behold, they were both Onyx in previous seasons.
It would be great if you could download HaloTracker’s raw data. From both my own experience (which feels similar as I was Onyx/high diamond previously and am barely able to keep out of Platinum) and from looking at this player anecdotally, they’re virtually never matching “down” (e.g., they’re more or less always the lowest CSR player) suggesting their MMR is perpetually above their CSR–but for whatever reason, they’re nowhere near a 50% win rate over large samples (50+) games and thus can’t rank up.
These players are playing a lot, but their CSR cannot catch up with MMR because their win rates are low over time.
As a solo player I hear you.
Please no.
For a few reasons.
Firstly they weren’t as good. Halo 2 was a prototype for TrueSkill - but it was very slow to rank people - and created a lot of mismatched games while you waited for it sort itself out. And Halo 3 wasn’t as accurate (TrueSkill2 > TrueSkill) and had other issues - like the 20 hidden levels and the rank locking.
Secondly - it’s the match making that is breaking down. Low population and people running around in carefully crafted squads to take advantage of the situation. This would have been just as troubling for H2 and H3.
And finally 1-50 or Bronze to Onyx. The scale isn’t relevant to the ranking. They are just different ways to describe your position in the bell curve.
Multifactorial?
The obvious one is Season 1. Most people have this as their best season. The curve was too skewed to the right and then 343 had to reset everyone back a division. When the dust settled everyone seemed to be 2 or 3 tiers below their Season 1 peak.
And now we have different population left playing the game. And this gives a bit of a rank distribution to the left. The system just takes the population it has and tries to spread it into a normal curve.
If you only let Onyx players through to Season 3 - they would be redistributed Bronze to Onyx. Just with a relatively small skill difference between best and worst.
As the population drops - I imagine it tends to be the casual players who fall out the most.
And then there is ranking debt. People jump into ranked when the Season starts to get their rank and then drift back to Social until next season. This creates a debt of rank that the system has to sort through. At the moment this is probably significant.
And finally we have the make up of the squads. It appears that some people get caught in pockets of rank that they can’t escape. In match making they end up either being the carry in a weak squad or the filler in a tough squad. So they end up winning a lot of games that don’t push their rank up and losing a lot that don’t pull it down.
And the end result is that they find it difficult to move their rank in either direction.
TLDR?
- The Season 1 reset down a few tiers.
- Different population curve this Season.
- Rank debt from people not playing all Season.
- Matchmaking and Squads trapping people in rank pockets.
It would be even better if 343 just gave as a Halo 5 like API so we could look and investigate this stuff ourselves.
Can’t believe they just pulled that rug.
As above - I’m wondering if it’s more that they get trapped in match making pockets.
Not so much their MMR vs their CSR - but more MMR vs opponents MMR.
Their MMR is always above their enemies in a win and below in a loss. So their rank just treads water.
And similarly - depending on the Match Maker’s mood - you get more of one than the other - so you end up with a higher or lower W/L than you should. And a CSR that goes nowhere.
But at the very least this disproves the old wives tale that 343 deliberately manipulate your next game to keep you at 50%.
1 Like
Fair stuff. A few points–I think nameplate was season 2 only? Season one just got recon unless I missed it.
And then I’m saying CSR is below MMR because I always gain close to max and lose 5. If I won 50%, I’d be moving up. But I don’t.
1 Like
The win is not the most important factor. If your mechanical skills are bad your win will grant you 5-8 points towards next rank. Your lose however will remove up to fifteen points from rank progress… Holding objectives reduces your skill ratings because you will have more assists and deaths that other players who refuse to hold objectives. So if you win 3 matches gain 15 points towards rank … Lose the fourth and your breaking even. For players playing the objective you have to win better than 75% of your matches to make progress towards the next rank. If you are playing for kills you only have to win better than 25% of your matches to make progress. That is what breaks the point system… Your performance in a single match doesn’t effect how many points you get for that match… your overall performance does. So keeping your skills higher than your current rank means moving up fast no problem. Skills like accuracy headshot accuracy damage out. And all of these are connected to your k/d ratio. So you can solo enter and lose 3 matches just grinding away at your KDA… While your ranking goes down per loss… Once your losing only 5 points per loss it’s time to get your friends and create a party… 15 points per win. 1st win negates 3, 5 points losses. Second is making progress 3rd you might be slowing down … Time to solo and grind KD again for 2-3 matches just done lose too many points.
1 Like
Having low to mid ranked players hold the objective is startegically not a bad idea. However players holding the objective end up in 1v 2/3/4 battles which only lowers their rank further. Immediately following those battles higher ranked players scavenge kills and abandon the objective which keeps their k/d ratio in line and further raises their rank. Maybe a better strategy is to work together to hold objectives? Why is solo kill scavenging so highly promoted in a team game? 4 shots from a br is a kill… When 4 players are shooting the one player playing the objective they are usually dead before they can get a shot off. So the other players can flank abandon the objective and do it again. Holding objectives does not raise your skill. Therefore holding objectives means losing up to 15 points in a losing match and winning as few as 5… While scoring kills raises your skill rating so winning means up to 15 points per match and losing about 5. So as the person who plays the objective because your teammates won’t… You will always be at the bottom of that totem pole.
It is. It really is. I’m not sure how else to emphasise this…
That’s essentially how ELO based systems work.
You beat sides ranked above you and you rank up. Lose to sides ranked below you and you rank down.
The magic is in working out how much to rank up or down to rank someone as quickly and as accurately as possible.
The CSR change at the end of the game is decided by the result, the skill gap between you and your opponents, and the pull of your MMR (your CSR resists moving away from it).
Nothing else.
I suggest reading Microsoft’s TrueSkill2 discussion paper. Or any post from 343. Anything on the forum by Josh. Moserware’s maths articles.
And last Season I kept track of 500 games - and recorded CSR change and as many metrics as we could. There wasn’t a significant correlation of CSR change with K, D, A, KD, K-D, KDA, Objective score, KPM, DPM, etc.
Nope.
Playing the objective properly increases your skill rating because you will (hopefully) win more games.
You might get a few less kills… and a few more deaths. But not enough to worry about. eg. you should get just as many kills winning the oddball. And if you are carrying it you should drop it to help defend it.
Not sure why you would end up with more assists.
No. You just need to beat teams ranked above you.
No.
If you are just starting, early in the season, your KPM may drive your rank up faster.
But playing only for kills will eventually bring you undone. If you want to rank as high as you can go you need to be playing the objective as a team.
Partly right. Your performance that match means nothing to the CSR.
But a solid KPM and DPM can push your MMR up - which will then pull harder on your CSR in the next few games.
But here is the caveat. Try pushing your KPM up vs players your own ability or higher. It’s very hard.
No. Metrics such as accuracy, damage, and KD mean little.
The only significant metric is KPM. And that works to rank you up faster - but not higher.
Which will just cost you rank as your KDA will little to save you.
It’s KPM.
And it’s an important distinction.
In a ten minute game you can have a high KD by getting 5 kills and 1 death (KD of 5) but that’s only a KPM of 0.5
As a rough guide you need to a KPM of 1.2 to 1.5 vs the rank you are trying to achieve.
The only important bit here is to create a party. Team work will get you more wins in objective games.
Solo can only take you so far…
You need your best 1v1 player’s protecting.
But you also need the lower players to drop the objective and fight back.
Only to the point of maintaining their KPM vs the rank they are trying to achieve.
Yes. 100%
It’s not. It’s a lie that people keep perpetuating.
No thanks to 343 and their lack of communication.
The key here is to have the other three players helping defend the fourth. And the fourth to join in.
Essentially if 4 players are attacking - 4 should be defending.
Nope.
The CSR changes on losing are based on your skill vs theirs and how far your CSR is from your MMR.
It is true that a bunch of people playing together will slowly separate in ranks. And that is based on the weighting for your KPM.
For example if I place with my sons I will come out high Platinum and they come out mid to high Diamond. We’ve won the same number of games - but their higher KPM pushes them up.
But it’s not because I’ve held most of the oddball. Which I have. It’s because they are better at 1v1 and they will get more kills. Not just the few extra by not holding the oddball - but more because they are better at it, and spend less time respawning than I do.
You’re right about the start off process of getting ranked.
What do you think about if you have ten games of ranked played with a placement, but if your friend only has 9 you can no longer play together?
I wish I could post screen shots. You keep referring to kpm or kills per match. And stating that falsely. The true rank system as described in the page takes information from unranked matches as well. This is two separate systems working together… If your a gold 1. But your overall KD ratio is that of a bronze player the two systems will try to come together. You will recieve 5 or 6 points towards platinum for a win and 12+ points for a loss. Against higher ranking players. This is because the true rank system is trying to "fix’ those statistics. Depending on how badly broken your account is there is a remedy. My account is new. And I spend most of my time in platinum and diamond sweat lobby’s due to our teams average mmr. So I recently spent a few hours in unranked matches focusing on kills only… It’s was a blood bath. Made my teammates have to sweat the objective but skewed my KDA or kill death average. I now gain 7 points per win. But I have to do this periodically.
Because if my dying once means 3 kills for the team. Then it’s a worthy trade. The games about winning.
Sorry. KPM. It’s kills per minute. Or your “kill rate”.
https://www.microsoft.com/en-us/research/uploads/prod/2018/03/trueskill2.pdf
It’s described in Section 8.
It’s true you have a global MMR - which contains all the bits on “recent” games. This helps reduce the data footprint of the system.
And then each playlist has an offset MMR.
So if you are playing well in one playlist, your global MMR goes up, and all your other playlist MMRs will bump up as well.
Think of it as a “form” rating.
But each MMR has it’s own curve. Including sigma values. So those lists you play a lot in (and the system is confident in your rating) shouldn’t move too much.
Your KPM is a weighting - and we don’t know exactly how strong.
When Josh used to look up people’s MMR for them in Halo 5 - he used to talk about their KPM. Specifically vs the rank they were asking about. He would say you have a Diamond KPM vs Platinum players - but only a Gold KPM vs Diamond players - and that’s why you are struggling to rank up.
And it’s bound to a playlist. If you keep track of your expected kills and deaths on Waypoint (which are calculated from your KPM and DPM)- you’ll find that it’s specific to each playlist. eg. I have a 1.5 KPM in King of the Hill but only a 1.2 KPM in Slayer.
I’m not sure what you mean.
Your wins will carry your rank to a certain level. And you could probably use KPM to squeeze a little bit more.
But I doubt you are going much further.
And remember, it’s about matching the KPM of the rank you want to be. If you want to be an Onyx player you need to have the same KPM as other Onyx players vs Onyx players.
Isn’t every ranked game supposed to be a “sweat” lobby?
No you don’t.
Speak to another person and they will tell you that the best way to get your CSR up is to smurf their account in Social or bots…
It’s all a big old wive’s tale.
Any affect it has on your global MMR and then ranked MMR is minimal. And short lived.
So just get out there and work hard at getting better.
So your solution is to sacrifice winning team strategies to boost a single players “Kills per match”. But you are outspoken that the win has a greater effect on csr / mmr? Sounds like your talking in circles? Or would you prefer an open arena where everyone first spawns with a single weapon and there are no alternative weapons no objectives other than killing and no spawning (because if there were spawning after that spawn killing would dominate the game) everyone plays to a single team strategy shoot better than the guy across from you and see what happens. As I already stated the use of non ranked matches to raise KPM has to be done regularly to work. And it does work with no sacrifice to team wins or deviation from winning strategy. And only a time commitment of killing sprees and some lost matches. -_(0.0)_/- .
Incidentally I have been awarded an achievement for being outnumbered in battle. So whatever I’m doing they made an achievement for it. Mind you my account is … 30 days old?
Not my solution.
Microsoft’s
And you have to keep in mind that TrueSkill2 has to be functional. It can’t waste data space and processing time on chasing every minute strategy or play skill.
When they crunched the data - the only metrics that were useful over and above the result were KPM and (to a lesser extent) DPM.
And you can’t argue with data from millions of matches.
But, as with any system, there will be outliers that are hard done by. And some that will benefit.
Apologies if I haven’t emphasised clearly.
The result is the key. And specifically who you beat. Wins vs higher ranked opponents and losses to lower ranked opponents are the ones that shift your rank the most.
There is a “weighting” for KPM. But we don’t know how much.
I believe it’s main function is to identify big fish in small ponds - and rank them up faster. Smurfs for example. But if you think about how KPM works - it will quickly drop as you approach your own skill level - and that’s why I think of it as ranking you up faster but not necessarily higher.
But when you have a group of friends who play reguarly together - and so pretty much have the same W/L pathway - the effect of KPM is exaggerated (at least relative to each other). It’s likely that KPM is the influence that spreads those ranks within the squad.
Not sure what this has to do with the discussion?
Sorry. I disagree.
I have been involved in a lot of ranked threads over the years - and nobody has been able to provide any hard evidence that it works. Zip.
And when I did a bit of data dredging… over a period of five weeks… where I played ranked six days a week and then spent Tuesday nights mucking around in Social… I found no statistical evidence that the average MMR of my opponents changed the following day.
People are still feeding off the urban legends that followed from that player who used Bot Camp to artificially inflate his MMR into ranked and got slaughtered in Onyx 1800+ type lobbies.
Cool achievement. Getting a new one is always good value.
And look. You have do you. Halo is all about having fun.
But if you want to rank up a bit more - you have to play the system.
Playing in a squad with better players and dying a lot is going to make it hard for you to rank up. And your friends are probably going to rank a bit higher than they should as well (on the back of your noble sacrifice).
But that’s just a reality.
There isn’t an AI judge sitting there and going - give that Spartan more CSR for being so brave.
You never acknowledged his actual response again
Which bit?
Edit; sorry - just saw the highlighted bit;
Bottom line…
Playing the objective does come at a cost to your KPM. And probably DPM.
But if you are playing the objective to the point where your KPM suffers enough to drop your rank… you’re doing in WRONG.
It’s a weighting.
But anyway… let’s look at the OP’s last five ranked games; 8-13, 4-15, 3-11, 7-21, and 10-18.
I clicked on the game they only got 3 kills. Their team mates got 15, 16, and 17 kills. Which leaves the player’s KPM as FIVE times lower than everyone else.
And yes. Once again. It’s only a weighting. But that’s a lot of weight.
A team that shares the Objective roles - will end up ranked higher in the end.