PROOF - You CAN rank down & lose SR with Positive Win Ratio

Competitive Discussion
135

It doesn’t have to understand every single way to achieve value–that’s the point. Every hero has a baseline kit that all players will necessarily use. “Value” that isn’t being directly recorded is being translated into wins.

No, the # of offspring does not translate directly into fitness. Period. Natural selection doesn’t produce a dichotomy; there are numerous potential outcomes between any two populations. You can’t distill NS down to “wins” and “losses.” NS is just as likely to “create” an entirely different game without competition for the organism to play–is that “winning?”

“Survival of the fittest” is a phrase you should avoid using. It’s next to meaningless when describing evolutionary mechanisms.

And my message was, “You are misreading the game and under utilizing Mercy’s kit.”

I have never played a game on Mercy where all that could be done is healing. If you cannot figure out how to do more, then you will underperform because other players are figuring it out. I’m not terribly interested in predicating my views on “what if-isms” at this point.

3 Likes
136

I respectfully disagree. To a point, if they had a winning record and was down 100, I don’t think I would respond. And I think the player base agrees as well.

And again your being very exaggerating the issue. The difference is not a grand master chess player or even a platinum in silver, a border line gold (who has to endure new player gold…). And this is based on what we know of the system, and there’s so much we don’t.

I said all I can type to this one. Let’s hold up these strong stance and let the game die…

Let me end with the myst himself, don’t quote me, quote him. Wins and losses are the most important factor… well myst needs to correct themselves.

137

What OP said about their SR difference is a bit misleading. They are comparing their career high from season 25 to their end of season SR from 28.

Despite a career high at 2071, they have consistently ended seasons around ~1750 SR.

7 seasons of end SR-
Mean: 1754
SD: 80
Median: 1795
Range: 236

OP is essentially arguing that a 52% winrate should equalize at or closer to their career high. I don’t think that’s reasonable for anyone at any elo.

And yes, winning relative to the “strength of schedule” of the opponent is the most important factor.

2 Likes
138

Is the myst wrong? I’ll stop quoting him/her if they’re wrong.

139

That response is in culmination of several developer posts. For a more detailed explanation including developer citations. Another member of our community has a very detailed guide.

3 Likes
140

You know… this is all to make the system better, more friendly and to RETAIN people.

141

OP’s “issue” is an intended design feature of all Elo systems–win rate alone isn’t an accurate measure of skill. SBMM + EOMM hybrid systems are designed to match skill level and lower the churn rate of players.

Blizzard is almost certainly using a hybrid system.

2 Likes
142

You could say this, but then this is simply accepting that some playstyles have an advantage over others because all “value” is being translated into wins, but not all value is being recorded for the SR buff/nerf that will come after the win.

The point is win ratio is guaranteed to eventually put you in the the right rank, it is inevitable. In the history of competitions wins and losses have been what determines the ranks. This has been done accurately and effectively since the dawn of competition.

In contrast no competition has ever used pbsr or anything similar as the basis of what determines ranks. The only one that does something to that extent is stuff like boxing, but the priority goes to determining an actual win and loss decided by the competitor, and the only reason it differs to the judges to calculate the win it’s because there was not enough time to get the preferred win or loss.

Yet why do they need multiple judges? Why not count the number of punches landed, missed, blocked, number of times someone’s knocked down? Because everyone knows statistics incapable.

It does not translate into fitness but neither does it matter. Fitness is the reason why they are more likely to survive, equivalent to skill in being the reason that they might win a match in OverWatch. But I’m not testing their skill I’m testing the win ratio. In order to have a win ratio/survival ration you need to have the opportunity to win/live or lose/die thus the number of offspring is being likened to number of matches played because each life/match are opportunities to contribute to their respective survival/win ​ratio.

The one with higher will be the one who was more fit, and this only gets more accurate with more offspring/matches.

Does not require anything fancy, it’s all obvious.

I don’t need the *whole game to prove inaccurate results. Because my claim is not that pbsr is completely inaccurate.

Question, is it more valuable to revive a dead player or keep another player alive? Of course you might know the answer to it because you would just look at the context of the situation and you can determine easily what would be needed. But try programing the understanding of that in to an AI system, and quickly you’ll realize for something that would appear to be so simple, the task is monumental when using what is basically logic & math to interpret an abstract concept into a factual reality.

143

Yes it is possible.

I think this is a really interesting topic, so here my thoughts. The information provided here is from how I understand the system behind the matchmaking and the SR adjustments so it might not be 100% correct. If I’m wrong pls correct me. And I’m sorry if some things are a bit too complex / confusing.

⠀
⠀
Matchmaking
First, the matchmaker is trying to find a fair match based on your:

  • MMR (Number between approximately -3 and 3)
  • Region including Ping
  • Win probability (must be higher than 40% for every player, otherwise the match won’t even be created)
  • some other small factors

Your win probability is not the same as your win percentage. The system tries to predict how likely you are to win the game based on I think SR (yours compared to the enemies).

  • If a 4000 SR player loses against a 3950 SR player, the 4000 SR player’s SR is higher than his actual “skill?” and the 3950 player’s SR is too low for the “skill”. So this has to change, this happens when he loses.
  • Yes I know, you can’t just say if you are 4000 SR and lose against a 3950, you are worse than the 3950 SR player. It can happen, nobody is perfect. But that’s how the system sees it (I guess) and that’s why you have to play many games to actually climb → Also I guess that’s why placement matches exist.

⠀
⠀
SR Adjustments
SR Adjustments happen after a win or loss. First we need a base SR. If you didn’t play for a very long time, I think they take 50 SR as a base, but this is obviously not the case for you. Therefor we take 25 SR. “Decay” would modify the base as well but in this case we ignore it since you are not above 3900 SR.

This is where the predicted win probability comes into play again.

  • The more (> 50%) the system expected you to win the game, the more they add to the base if it’s a loss and the more they remove if it’s a win.
  • And the less (< 50%) the system expected you to win the game, the more they remove from the base if it’s a loss and the more they add if it’s a win.

Simplified TLDR (Win probability):

  • You get rewarded for winning “less winnable” games.
  • You get less SR for winning “more winnable” games.
  • You get punished for losing “more winnable” games.
  • You lose less SR for losing “less winnable” games.

Now in your case since you are not Diamond or higher, your performance compared to similar players counts as well. With “some SR”, I mean approximately 1 SR - 7 SR:

  • If you perform better than similar players and you win, they add some SR to that number.
  • If you perform better than similar players and you lose, they remove some SR from that number.
  • If you perform worse than similar players and you lose, they add some SR from that number.
  • If you perform worse than similar players and you win, they remove some SR from that number.

Simplified TLDR (Performance based SR):

  • Performing good and winning: You gain more SR
  • Performing bad but winning: You gain less SR
  • Performing good but losing: You lose less SR
  • Performing bad and losing: You lose more SR

At the end you got a number, if you win, they will add that to your SR. If you lose, they subtract it from your SR.

⠀
⠀
Your Season 29
I’m gonna take the SR of your last Season as a starting point. Season 28: 1626 SR
If we take you most recent stats with my format:
Wins / Losses / Ties @Hero (Total games played)

8 / 5 / 1 @Brig (14)
1 / 1 / 0 @Ana (2) [ You wrote 1 played, but have 1 win & 1 loss so I assume 2 played ]
1 / 0 / 0 @Bap (1)
1 / 2 / 1 @Lucio (4)
11 / 8 / 2 @Total (21)

Performing a quick calculation:
11W - 8L = 3 ( We subtract your losses from your wins since they mostly even out in average )
Now we take 3 * 25 SR = 75 SR. ( We take the base of 25 SR and go for a 50% win probability game with no over or under performance )

This would mean that your current SR would be at 1626 SR + 75 SR = 1701 which is 4 SR lower than your actual current SR. Means currently you gained more SR.

⠀
⠀
Important Notes
You mentioned, you got the stats based on the win percentage in your career profile. The win percentage for some heroes might not be accurate since I think they calculate it for every second you played a hero in a match. So if you played a hero in a match and switched to another hero, this might be influenced. If you want, you can recalculate it by using the numbers from the Career Profile’s “Statistics” Tab. But as mentioned in the first sentence and explained in this reply, it is possible to drop SR even with a win percentage > 50%. Depending on the matches and your performance.

Always keep in mind that most games are balanced due to MMR and the system got your back if you get matched against a higher average team. See such things as an opportunity to farm SR : )

Keep climbing and try to improve: Watch Streams, look at your Replays and try to figure out what was going on and why you lost games. You will start climbing really fast!

1 Like
144

Sure–but winning is still a direct measure of the non-recorded value. Additionally, the recorded stats are an indirect measure of that non-recorded value. E.g. CCing will proc your Inspire.

That isn’t true though. All wins resulting Elo are weighted by the difference in the players’ skill level. Chess, for example, calculates the expected probability of winning given the difference in Elo between two players. If the player performs lower than expected, they lose Elo. If they perform higher than expected, they gain Elo.

PBSR is designed to converge on the player’s skill level more quickly. Because Overwatch is a team game, we don’t want players artificially hard stuck due to bad teammates. We also don’t want players artificially boosted due to good teammates.

By analogy, chess uses an Elo modifier called the “K factor.” The K factor is a coefficient for the difference in actual wins versus expected wins. New players have a higher K factor for 30 games to find their skill level faster. After 30 games the “bonus” is dropped.

Dropping PBSR after a certain number of games has also been suggested–but that doesn’t address the issue of being stuck/boosted due to teammates.

And I’m saying that NS doesn’t translate into a win ratio because it’s not binary. “Winning,” in evolutionary terms, does not mean having a greater survival ratio relative to the competition. “Surviving” is an equilibrium function relative some environment. See Lotka-Volterra equations for examples.

No, that isn’t the case.

Right, it is situational. But, making the most correct decision will increase your chance of winning while also filling the on-fire meter. Increasing likelihood of win = value. In either case, your healing or res stats will be cataloged–indirectly measuring non-recorded value.

2 Likes
145

Just curious, You don’t think people could be a force for a win… despite not performing the metrics that measure in the PBSR system?

Like if someone comes up with the strategy that forces the win, that is immeasurable in terms of performance based algorithms right? I am purely against extra measurements for competition, and all competitive aspects should be measured exactly the same way. I think PBSR should either be measured in all ELOS, or none. Having it in one, makes people who get carried to Diamond able to be carried in all ways.

And to add to this, if there is a 48 percent winrate, There is no conceivable way that they should be rising in SR, obviously a 52 percent winrate should raise SR IMO too. I don’t care how good you were, you still lost. I understand people don’t agree with this. But oh well.

I say this, before I dropped to this current profile, My team and I had stopped playing competitive because despite ALL of our wins together, we didn’t end up in the same place. We six stacked, and maybe some of us got carried, but we ALL still won. With teammates, and then it forced us not to be able to play together anymore. All the practice, the plays we built that worked on… wasted. That doesn’t seem like a way to force a team game… to be a team.

I don’t know why I’m typing this. I know I’m just going to get hit with why a team playing together shouldn’t be rated the same despite whatever thing that someone says is more important.

146

Sure–but, performance is about consistency in gameplay. It isn’t about a few good or bad games. If you’re consistently forcing wins through strategy, you will climb.

People often forget what PBSR does: it gets players to their true rank faster. Your performance is still always being measured through MMR–regardless of Elo. Even if you do get boosted into Diamond, you still have to perform like a Diamond player or you will derank.

PBSR is a modifier that helps you compensate for bad teammates and avoid getting “hard stuck.” However, MMR is always present; you just don’t get extra SR for playing better than your teammates in Diamond+.

Why? If you have a winrate of 48% against opponents 800 SR higher than you, shouldn’t you climb?

Okay, do you think boosted players should be rewarded? By analogy, should that person that did 10% of the work on a school group project still get an A+?

3 Likes
147

That’s why they removed PBSR for players higher than Plat. Because they want them to focus on winning the game as a team.

1 Like
148

[Quote} People often forget what PBSR does: it gets players to their true rank faster. Your performance is still always being measured through MMR–regardless of Elo. Even if you do get boosted into Diamond, you still have to perform like a Diamond player or you will derank.

PBSR is a modifier that helps you compensate for bad teammates and avoid getting “hard stuck.” However, MMR is always present; you just don’t get extra SR for playing better than your teammates in Diamond+. [/QUOTE]

so it shouldn’t hurt that everybody has it right?

not IMO,

[quote] Okay, do you think boosted players should be rewarded? By analogy, should that person that did 10% of the work on a school group project still get an A+?
[/quote]

back in the stone age when I went to school, this is how it happened, regardless of work put in by others. You all got the same grade, so some people had to fill in for the lacking person.

149

I get that. I said either everybody or nobody. (I’d prefer wins or losses as the metric) but if everything is PBSR toop, I’m fine with that. Two sets of rules IMO ruin it for me.

But after this… I’m not responding. I’m just going to read anyway. Like I said I don’t know why I even popped in here.

1 Like
150

Personally, I don’t think it hurts. But the devs were concerned that higher elo players would game the system for bonus SR instead of trying win. It’s especially a concern in games you know you’re losing.

It’s also been suggested that PBSR be a temporary bonus modifier until X number of games.

At the end of the day, it makes it easier to climb and makes the player less reliant on their teammates until Diamond+.

Why? If a novice chess player at 1200 Elo wins 48% of their games against 2500+ Grandmasters, in what world does that novice not deserve to climb?

I’m not asking if it happens–it clearly does–and many instructors specifically safeguard against students that are getting “carried.” I’m asking you if it’s fair that someone doing 10% of the work gets 100% of the reward.

2 Likes
151

this isn’t chess. But being as I don’t know the Chess world, I would guess the 2500 chess player doesn’t play a 1200 game. Or that the 1200 player is under ranked. The 1200 player should play 1200 players, and rise until he reaches 2500. Also, it’s a solo game. not team based. Your analogy is fine, it just doesn’t fit to me. Maybe because you seem to understand how competitive chess works better than I.

and if it was worth the A for you, then it doesn’t matter who did the work. That’s an individual project then. IMO. You’re fine if you don’t agree, I understand your argument. I just don’t agree. I’d rather get the A, carrying the guy… then a C if I didn’t.

152

I’m aware it’s not chess. I’m drawing parallels to other common games that use Elo to rank players.

Under any Elo system, consistently beating better players causes you to climb.

I agree that a 1200 player should be playing someone closer to their Elo–which is exactly what Overwatch does. However, if the Elo range is 1200-1299, the 1200 player only needs to beat a player 12 Elo higher than them 48% of the time to gain Elo.

Why is that?

  1. There’s expected win odds for the 1200 player against the 1212 player is 0.48
  2. If the 1200 player exceeds the odds of 0.48 with a win (worth 1), then they are rewarded with Elo.

That’s how Overwatch’s MM works too. If you’re predicted to lose with a 47% odds, but you win 48% of those games, you climb.

You still get the A, but the other guy that did nothing gets a C.

2 Likes
153

once again. That’s not how it worked on group projects when I went to school. one grade. Maybe things are different when you went to school, or where. Group projects, group grade.

and I know you are aware this is not chess. But if I compared it to another team competition that the game is more akin to in my eyes, I’ll get told it’s not like that. SO. if that’s the case as has been put to me, sorry it trickles back. But still… if a 1200 played a 2500, and won, yes. They are underranked, OR the 2500 is OVERRANKED. to even be in the same league if that would happen. But that’s not what happens. I am bowing out now.

I get your arguments, we are on different sides. Have a good day.

154

Okay. I have 3 degrees and I’m working on my next two. I TA an undergrad genetics lab with group experiments and assign grades on the basis of work produced by each individual.

Again, is it fair that someone doing 10% of the work gets 100% of the reward? Another analogy: you and two friends invest in some stock. One friend puts down 10%, you put down 45%, and your other friend puts down 45%.

Of the total investment return, what percentage should the 10% contributor get?

3 Likes