Wednesday, June 30, 2010

Screw winning, I'm in it for the PSR

Every since DOTA existed, people have been begging for a way to keep track of their score. It only seemed fair- after all, I'm better than everybody else, why shouldn't I have some sort of concrete number to reinforce my argument?

When S2 was looking for ways to make Heroes of Newerth worth buying, surely stat tracking was an idea not to be ignored! But where do you start? What is more important, kills or actually winning? And so it began, a vague question which would eventually spawn a debate even more useless then that of normal mode versus easy mode.

Hilarity did not ensue

Every since PSR was willed into existence, it remains the only stat displayed publicly every time you join a game. This has created increasing amounts of criticism, especially from players with low PSR but comparably high kill-death ratio.

Advocates of KDR argue that KDR does a better job of judging individual skill- after all, a completely useless player can still win a game if his 4 teammates are skilled, right? Stats, after all, should be a personal judgement of skill, completely independent of outside influence.

There's just one problem with arguing this: it's impossible to do because Heroes of Newerth isn't a 1v1 game, it's balanced around 5v5.

Sometime after Arenas were introduced into World of Warcraft, Blizzard stated that 1v1 duels were not balanced. Many people questioned this statement- was Blizzard incapable of balancing a few classes? In truth, Blizzard made it clear that it was virtually impossible to balance team PVP with a set amount of classes while also making 1 versus 1 balanced using the same roster.

HoN suffers a similar fate. KDR is in fact not a perfect system of personal skill ranking. This derives a lot from the fact that certain heroes simply are designed to score more kills than others- in fact, in competitive Heroes of Newerth games, certain heroes are literally discouraged from scoring killing blows on enemy heroes unless they are the only one capable of doing so.

PSR on the other hand suffers from it's own unique flaws. As already stated, it is theoretically possible that a player who is lacking in skill can be carries by 4 other people. This theory, however, simply doesn't stand the test of time. No matter how good your 4 teammates are, eventually you will reach the point where 4v5 simply wont cut it anymore. I do not have any statistics to back up any points I make, you'll simply have to take it or leave it.

It all boils down to one simple statement: Neither systems are perfect; KDR is flawed due to hero roles while PSR is flawed due to teamwork- however, PSR is less flawed overall in determining personal player skill.

So where does KDR fit into this? Is KDR useless? Does PSR > KDR? Short answer: Yes. Long answer: Both count for something, PSR is the lesser of two evil ways to describe personal skill. Trying to create a "personal" skill ranking for a team based, teamwork dependent game simply isn't 100% possible.

...I for one am confident S2 will make "Wards placed" the only publicly displayed stat come next patch...


  1. Yeah, no one wants to play when no wards are placed by their team for one or another reason :>

    Most of the time, when the teams are even, the team with best placed wards and most ganks win.

  2. To be honest, the most important stat will be team match making rating when it comes out. As it will show how well you do when facing organised teams of 5.

  3. True that, Gimble.

    PSR>KDR, definitely.

    In the future, though, HoN will have numerous Achievements, I'm hoping that that would be a good thing to gauge a player's skill on.

  4. I don't usually hear the argument that bad players get carried up so much as good players get pushed down by a team of fools. I for one play with my IRL friends all the time which as a random British teammate we once had put it "is like watching retards fuck a doorknob."

    I agree that a truly skilled player could just take mid and carry his team almost completely independently of his team, but there is a bracket below very skilled where just skilled players PSRs suffer from bad teammates.

    Anyway, in conclusion when judging player skill I look at all the stats, including PSR but it really comes down to how well they play after the game starts. Skill is something a person has, not a scoreboard.

  5. Im sure you can muster up a profile of individual skill in HoN, but the fact is this profile wouldnt be very accessible to humans, but useful for machine matchmaking. Think of the PSR, you play with within a certain range of your own PSR. Now imagine k/d ratio, you want to play within a certain range of kill death ratio right? You can use this concept of range searching, but instead of just 1 dimension (PSR), you can do multiple dimensions of range search, each dimension orthogonal and adding to a final total score.

    Imagine having a number of values measuring your ability to win, total wins, win speed, win acceleration, time played, time of day usually winning, probability of disconnect, carry to support ratio etc. These things can be measured with ease, using them for matchmaking is trivial, choosing what to use or measure is hard.

  6. Then people will immediately start to spam wards at the end of every game to get their wards placed stat up.

  7. I was joking about wards placed, guys :)

  8. Oh, damn, actually I saw in 1 game our carry place fucking 4 ward at the very end where we finally won. :D

    Other then that: TSR ~ True Skill Ranking ...which is a mod, that tries to combine both, KDR and PSR with all other stats. I know, from personal experience it has also flaws, like in my chase it says i have 6.4 (green), which is almost pro like due to only few players are going beyond 6.0, but is accurate most of the time.