If you or anyone you know is only gaining 20 SR for a win and losing 30 SR for a loss, then this individual is being HARD BOOSTED!!
Competitive Overwatch performs a statistical analysis on a player to determine if they’re exceeding the performance of their peers (this happens every time you play competitive).
If the system determines you as an individual are exceeding your peers, it’ll give you PBSR (performance based skill rating) and you can gain anywhere from 25 to 150 (the most I’ve seen smurfs get) SR for a win.
If you’re receiving less than 24-26 SR for a win, you are under-performing for your rank. If you’re losing 30 or more SR for a loss….
He plays exclusively tank, because that’s all he could queue as. 1k damage/min, 1.6k damage blocked/min as Sigma. Similar with Zarya, but less blocked. I’m seeing this as “if it’s good enough for us, why not for him” kind of situation.
I’m confused have we determined the player is being Hard Boosted because the system performed a “statistical analysis” on a player to determine if they’re exceeding the performance of their peers" despite also knowing stats don’t exactly win games?
So a computer is a able to predict the chance if a games out come using stats?
The same stats that don’t win games? If they don’t win games and how can you predict games being won?
I do agree that stats can show your capability/capacity to do a task but. But my capacity to get elims on the easiest targets, or pump damage into roadhog, or consistently healbot tanks, or simply just stay on the cart as it moves the longest, or play safer while just shooting the safe targets with no thought on their value would purely statistically show what exactly.
How does this apply to pbsr which is calculated with the statistics that must show more than just simple capacity but effectiveness to justify previous or gains and loses?
What? This side debate seems very off the mark to me.
PBSR doesn’t have to show anything. It’s an arbitrary metric that exists simply to allow players to have some individual influence over their rank that is independent of wins and losses. This is partially to try to move people more quickly to their proper rank, and partially because the playerbase gets very agitated if they are ranked purely on wins and losses.
It’s intent of course is to be based on metrics that are correlated with good gameplay, but for practicality it’s also limited to easily measurable numbers chosen by humans. Yes, those humans had a lot of data to help their decision, and they are looking for numbers correlated with better chances of winning, but PBSR is not and is not intended to be a data analysis of your chances of victory.
This is a joke, right…?
Because the denominator is irrelevant, performance-over-time can be converted to any time unit. The only other meaningful alternative is performance-per-match.
You said it yourself players want to have an influence over their rank (which is a redundant point because that’s what winning/losing is for and will eventually prove) One would assume that this influence is due to good performance & not brainless stat farming.
The other user admits and I agree, the stats only show what someone is capable of but that does not define value. Just because someone’s capable of getting mega damage just pumping led into a pocketed hog doesn’t translate into mega impact. Since PBSR uses these same stats that we just agreed do not inherently show value, then it cannot be said that pbsr is an accurate judge of value.
The reason that this matters is because the user had claimed that the other player was being “hard boosted” and did not actually have skill like the OP said and rationalized it with the pbsr system. But if the pbsr system uses stats, and we just concluded that stats can not measure value, then has just contradicted his own rationale and thus the claim of the player underperforming would be invalid.
No because someone who had 10 minutes on Ashe might have 28-32 elims/10 mins when really that number should be closer to 3-4/10 minutes.
The more time spent on a hero, the more “10 minute snaps” of time the system has to observe/analyze that players performance.
10 minutes of data does not paint any kind of comparable picture to 120 minutes of data. That’s literally 12 ten minute blocks of data as compared to 1.
What in Chris’s name are you even saying here?
I didn’t make any implication of this whatsoever.
What I said is, if you’re gaining more than the average amount of SR for a win, you’re receiving a PBSR boost.
Conversely
If you’re losing more SR for a loss than you gain for a win, then the system has determined you’re under-performing in some way or another.
This is theoretically true, and in fact it used to be actually frequently true especially for supports and sometimes for tanks. That’s why PBSR was removed, there were simply too many players who had consistently high win rates but never climbed due to PBSR.
However, while PBSR was gone they reworked which metrics to emphasize for supports and tanks, and when it returned, it was made a much smaller factor. That means the chances are much lower that someone has a successful play style that simply isn’t reflected in the stats, and if they do, they are much less likely to be held back by it.
So, I mean, it could happen, but I wouldn’t hold my breath waiting for it to happen.
Well, that seemed to be what the other guy was saying you were saying… the whole exchange seems very strange to me.
Yes, it’s determined that your measurable values of a fixed set of statistics is below average for players near your SR.
With this,
You seem to be suggesting there is some calculation of the chance of victory involved in PBSR. I mean, maybe you aren’t, it’s frankly a confusing sentence, but if you don’t mean that, I don’t know what you do mean.
Measuring in per-10 doesn’t mean you measure in discrete 10 minute blocks… you take all of the performance over the period (typically lifetime totals for Quickplay/Arcade and Season totals for Competitive), and divide appropriately to get it in ‘per 10 minutes’ units.
E.g., if they have played for 120 minutes this season, you take their total damage, and divide by 12 to get their damage/10min. You could just as easily measure in per-minute or per-hour… or per-23 minutes. Whatever. It doesn’t change the underlying statistic being represented, just the units it is represented in.
I assume per-10 was chosen as being a round number that is also roughly the time a single round in an Overwatch match takes.
He’s not the strongest at picking up what I’m putting down strong
Bro it was midnight and I was half asleep.
When the match maker creates the match it tries to achieve 50% chance for either team to win (as you know) and there are favorable/unfavorable edge cases.
This is the system calculating the likelihood of a win based on stats from players being pulled into the match.
He was writing something stupid and I was half asleep trying to shut him off.
No of course not. But if you look at your stats at the beginning of the season (when you do your first placement game) the stats are all over the place.
So you’re telling me that in the 4 mins I spent on Pharah that I was putting out 15k damage/10?! If that were true I’d legit be playing Ace Combat.
what I really mean is the first 20 to 30 mins someone spends on a hero isn’t accurate enough to determine their long term trajectory of performance.
Whereas someone with 2 hours on (hero) has won games, lost games, died a lot in one game, not died at all in another.
The data becomes a consistent profile of player performance.
Where as in the first 10 mins of the season, you could have 14 deaths/10 mins even though you died once, respawn’d and got staggered before you could escape. And then swapped to another hero.
If I boop a Hog with Brig for 70 damage, the AI system will register that, but what exactly was the value of that particular boop within the context of the the current situation, a far more complex question that humans can determine easily but can be immensely complicated to program AI to understand, even if it could assigned value can’t actually be accurate just like you can’t define what exactly is hot and what exactly is cold.
Now there’s many ways which you can make a pseudo system, but these ways do not guarantee accuracy, can still be abused and and we’ll have flaws that affect certain characters more than others.
There’s a dev update from Jeff where he says that AI isn’t in the blizzard match making software but he thinks it’ll be a great addition when their systems are more modernized.
