I am not saying that the implementation will be always bad. I am saying that, in the current situation, if you don’t talk about the implementation, there is actually nothing to discuss. Since the idea was already tested with PBMMR and was rejected with that particular implementation. And implementation is a concern as you can see. If you don’t talk about implementation, you don’t brinkàyrhinf new to the table and the general idea of performance based system was already discussed to « death ».
1 Like
They made it fairly clear that the dual MMR system had been scrapped. Now, they could be wrong, lying, or have changed it since, but I doubt it.
However, there will be variance in the amount of points you are awarded at the end of the match, compared to other people on your team. For me, it is rare that I win or lose exactly 200 points each time. There is the previously mentioned favored team advantage, streak bonuses, and party size adjustments. I really miss the time when they would provide a break down of how your total is calculated.
You can still see that, put your mouse over the points after the game ends in the Summary tab.
Then it shows what your 194, 206 points or so forth, consist of (usually there’s some adjustments like opponent favored, your team favored and others).
It used to show that info automatically. These days we have to manually check with the mouse cursor.
It’s not just bad plays being rewarded that I’m worried about. It’s that your whole measuring system simply wouldn’t work, given how there is no properly definable ‘good play’ that you could measure. At best it would be useless, at worst it would only make ranks even less accurate by having people randomly gain/lose mmr for no valid reason.
I’m aware in the current system ‘bad’ players can also be carried to undeserved wins, or that good players can lose ranks because their allies misplay. I just don’t see that as nearly as much of a problem as you seem to. Your results/rewards being dependent on your teammate’s performance is inherent to team games. If you can’t tolerate that possibility, you’re probably better off sticking to single-player games.
The current system of simply looking at wins may not be perfect, but it does its job well enough in my opinion. It’s also a lot cleaner than having some undefined ‘good play’ requirement for maximum mmr gains. The goal of the game is to win games anyway.
It was measuring your wins. Your 3-0 wins probably didn’t change your rating enough to rank up. Plenty of reasons that could happen; Maybe you got matched against relatively weaker enemy teams (and thus didn’t gain much mmr from it). Or maybe you were only barely into your rank, and the mmr gain only filled you from 100/1000 to lvl up to 700/1000 to level up. Or something like that. Would need to see the actual stats to really figure that out.
Not an expert on how this works though (and apparently it has been changed), so I’m not the best source to ask about this.
Oh, I’m aware those things exist. I just don’t think they’re a realistic solution for this situation. Big Data isn’t some magic spell to easily learn and then gather enough numbers to become all-knowing. It still takes a ton of work and research on figuring out how to properly use those numbers.
Again, I agree that with enough time and resources it would probably be doable to develop a supercomputer that can judge how well someone plays HotS. I just don’t think it’s reasonable to (want to) sink that much resources into merely having slightly more accurate mmr’s.
Or in your puzzle metaphor, nobody has the time to collect all those pieces and put them together.
1 Like
Just like in sports, different games have different levels of team dependency. IMO the over-dependence on your teammates in this game is a design flaw, not an asset. We all essentially live and die by the luck of the MMR. As already stated… only over the course of 150 games will you get an accurate representation.
If true than there is nothing “Placement” about these games. Just give me the same level as last season and let the luck of the MMR do its thing.
My supposition is simple… a multi-variable analysis will almost always give a more accurate picture than a single-variable analysis. That is a scientific truth.
But if you key in on the wrong data, you can draw an incorrect conclusion. Data Science is just that… a science. When looking at a lot of data you need to determine what is the signal, and what is the noise. It can be done and I would bet behind the scenes Blizz already has a lot of this data.
It might be interesting - and HP does have a lot of matches to check - how certain stats correlate with victory. Sometimes it’s complicated, like X or Y stat.
For example, a simply statement would be that a tank should provide, say, 15% of team xp. Do less, and he is either too passive or doing things he shouldn’t, solo camping, AFK, whatever. Do more, and he is probably spending too much time soaking. (Death timer in itself is a stat.)
Similarly, one could argue about damage numbers. I guess it never really turns bad, but above a certain threshold it’s getting less useful.
Also, just like HP, it’s all about balancing it. The concept of a solo laner Jaina… you have lower damage but higher siege. Or having low damage, but a lot of kills. Heck, a lot of enemy death as such.
There is no need to worry about one specific match going wrong. If it evens out on the long run, it’s just fine.
Sometimes even people get it wrong. You could easily get complained, or even reported as AFK for anchoring as a tank. Especially if you are anchoring a boss, the enemy team shows up just in time to contest it and then you somehow lose. If you added your miniscule dps, you had been done 2 seconds earlier and won, but anchoring caused a loss. (It happens fairly often.) So anchoring is not the ultimate strategy, it’s a small plus. Also, some anchoring is good, but waiting 60 seconds in a bush isn’t.
50-50 is harsh but a couple points each match won’t change much, at best it makes me G3 instead of G5 and I’ll have a funny winrate (compensating for the points). However it gives you a small points buff, mainly giving you peace of mind, so that once you get back to 50%, you’ll have slightly climbed. Or dropped, in case that’s the… case.
Talking Heroes Profile score, assume we’d get (MyScore-MatchAverage)/2 bonus points (to ensure the system is zero-balance). That makes it 10-15 points for the MVP, or a bit less even (tends to go 40-80, so average is 55-60).
Last, but not least, if the system allows you to climb (fall) a bit, it will get harder and your stats will deteriorate.
1 Like
Nope, it’s a lot more than that.
(Ignoring losses) the statpadding is this:
- Doing things badly can give you extra points
- Doing things good don’t give extra points because unmeasurable
So it’s possible, that the better player gains less points than the worse, which is what happened when PBMMR was a thing and is the exact opposite what is the goal if an “improved” system.
It’s not that players will hinder their chance to win on purpose for extra points (tho saying no one would fo that is delusional or naive), it’s that playing well, playing better won’t be fairly rewarded.
And if the “bad” can rise quicker in a winstreak than the “good”, I think it’s better to just count the wins.
1 Like
This is the most relevant post:
Spreadsheet stats included in the comments.
1 Like
Doing things badly has to increase the likelihood of a loss… or it wouldn’t be a bad play. Doing things badly on purpose means they lose more often and therefore have fewer points. This is why its like them shooting themselves in the foot. They wont do it.
Not doing those good things also increases the likelihood of a loss. See point above.
5 man… in any full solo situation I doubt medivh would have paid off. Especially any rank below high masters
The only possebility to solve this issue to create more balanced games I see:
- giving people extra points for winstreaks like in HS
- let ranks decay over time OR taking more points from players for losing streaks
Let’s see an example. Let’s say a win is net 100 points, and you can earn extra points by certain “plays” (stats = numbers). A player who earned the win but did nothing numerically measurable will get 100 points. While the player benefitting those good plays and getting high/good stats will get 200.
The “win more” in that system doesn’t necessarily means more points.
Or! If the most points still determined by win/loss ratio, than what’s the point of cherrypicking on certain things and giving extra points for them and only for them?
Your idea was bad, and after long arguments you basically say that the outcome is still mostly win/loss based but we give extra points after things that don’t objectively deserve that?
(For example, I lost games because ppl protected their KDA.)
Lets say points played out like this:
Win = 100 pts
Loss = -100 pts
That is half… now for the other half (Performance based):
Played great but still lost = -100+75 = -25 total for the match.
Played bad and lost = -100 + (-100) = -200 total for the match.
Played great and won = 100+100= 200 total for the match.
Played avg but still lost = -100+(-50) = -150 total for the match.
Played avg and won = 100+50 = 150 total for the match.
Etc.
That is the 50,000 foot view. Don’t take that as a literal algorithm. There’s hundreds of way to determine great/avg/bad performance. And we all know that… No, not every possible aspect of the game can be measured.
Half
1/2
.5
6/12
I don’t know how else to say (again) that half of you points would come from whether you won/lost and half would come from how well you played.
Half is a lot. If you want points, you won’t do things that make it more likely you’re going to lose.
*Played “great” according to flawed metrics.
Getting more/losing less points.
Why wouldn’t ppl try to be “good” according to the algorythm even if they lose? They lose less points if they fail than normal, so they “can lose more”, but if they win, they win even more.
Not the right ppl would rise with this system.
It’s a terrible idea that was implemented before and was pulled back for a reason.
Anyway, I see debating this further is pointless. You’re convinced that it’s a good idea despite history (was done, didn’t work) and logic (things that are good being unmeasurable, but bad things seemingly being good cuz math) telling otherwise.
So I’m out.
I admit I haven’t researched the PBMMR thoroughly… but it seems that system was based ENTIRELY on performance. I am not advocating that.
So you are railing against an idea that no one is promoting.
No need to leave… but 
bump if you do. Good talk.
1 Like
Actually, I suspect the kept the stupid system, and just hid it like the pairing against group thing. They probably think “if the players don’t see it, they won’t game the system!”. Maybe it doesn’t change your on screen score, but I bet you money the mm is not randomly grabbing people of the same mmr and calling it a day. I’m just sick of starting a new season with a high 70% winrate, and then after about a league up or so getting braindead teammates all the way well below where I started.
Why would a 1 post account post in a 3 month old necroed post ? why ?
2 Likes
From what I have seen lately is those 0 post accounts gets a lot of old threads in thier recomended list and then just maks a post in it.
The forum even tell you its necro when you try yo make a post but I guess people dont even care and make a post anyway.
People will necro post on this forum casue forum inself allow it to be done even tho the forum rules clearly says necro is frowned on.
They should just do what Wow forum does today. Every thread that has its last reply 6 months ago would get locked and people would be forced to make a new thread about it if they want to contine talking about it.
2 Likes