PSR, dobar feature ili ne?

[Sajro]

chitajutji jedan topik na S2 forumu pocheo sam da razmishljam da li je PSR dobra ili losha stvar, enivej, ovaj lik i njegov post su me zadivili...

I am tired of reading all the forum posts about PSR and stats, including:

1)They should not be tracked

2)They don’t show true skill

3)Bad players are getting carried

4)It makes it harder to start games

5)Etc etc etc.

First, a definition. PSR is Public Skill Rating and increases/decreases solely by winning or losing. If you play a team of equal PSR, then you gain or lose equal points for winning or losing. If your team is higher ranked than the opponent, you will gain less points for a win and lose more points for a loss. If you greatly outrank your opponents, you will gain 0 points for winning. In addition, if there is a very large differential between your PSR and the PSR of the rest of the players, you will lose/gain a very small number of points (approaches 0 the further away your rating is). This is presumably done so that higher ranked players are not punished for playing with low ranked friends and to prevent low ranked players from being "power leveled" by high ranked friends.

Secondly, if your team is 1700 and you play a team rated 1700 an infinite number of times, you are expected to win half and lose half of the games. If you win more, your rating will gradually increase, and if you lose more it will gradually decrease.

Thirdly, let’s make an assumption that you only play “reasonably balanced” games. By “reasonably balanced” I mean games where either all the players are roughly, equally skilled, or if there are a few significantly worse players, they are split evenly between the two teams. Stacked teams are unfair, not fun for the losing team, and should generally never happen unless by accident (e.g. pro starts a new account and hasn’t gotten his PSR to reflect his real skill level).

I will now answer every complaint against PSR or stat tracking.

Complaint:

PSR does not show true skill. You can be carried by your team or you can lose due to having a feeder on your team.

Answer:

In the short term, there will be fluctuation. This fluctuation could be upwards of 100-200 points, depending on how unlucky/lucky you may get. Thus, anyone who’s true PSR is 1500 could be rated between 1300 to upwards of 1700 solely due to his or her teams.

The key here is that over time, a 1500 rated player will eventually converge to ~1500 rating, because for every bad player on your team, you will encounter a bad player on the opposing team. And similarly, for every amazing player that carries your team, you will encounter one on the opposing team. As long as you don’t voluntarily participate in pub stomps or horribly unbalanced games, this is the inevitable law of large numbers.

Complaint:

PSR does not work because HoN is a team game.

Answer:

A TEAM is composed of five INDIVIDUALS. The better each individual plays, the better the team is as a whole. Assume that an average player contributes +1 to the team and a feeder can contribute less than 1 or even negative. However, a good player can contribute more than +1, and great players can contribute upwards of +2-3. This means that even if your team is not amazing, if you are truly good, you can make up for bad play by your team. Of course, this does not mean you can carry and 1v5 the enemy, but it means you will lose LESS games on average due to having bad players.

Complaint:

PSR is bad because bad players are high rated due to being carried or the "I'm better than him but his PSR is higher than mine" argument.

Answer:

I think the vast majority of players have not broken the 1700-1800 range. The reason of course is that PSR has only been out for some 2 weeks. Give it another week or so and we will start seeing a wider range of ratings, with good players rocking 1900+ ratings.

As I stated above, PSR can fluctuate due to randomness of teams. However, you will never see a 1500 skilled player getting an 1800 rating? Why? Let us assume 5 friends start playing, where 4 players have skill 1800, and 1 player has skill 1500. They all start out at PSR 1500 and will most likely win all their first games. However, after a certain point, this team will no longer be getting points from playing teams rated 1600 and below. The PSR system will give them 1 or maybe 0 points for winning. Thus, in order to increase their rating further, they will have to play other teams of 1700-1800. At this point, their team will win and lose games, and they will probably stall out around 1700, because even though 4 of them are good, the bad player will drag them down. And so players can only be carried so far up the PSR rating before they can no longer progress. Having this slight range of ratings is an unavoidable consequence of team based games.

Of course, you could have the possibility that this original team has 4 players that are skilled at like 2400 and one player at 1300, but I want to say this is probably not the norm.

Complaint:

PSR should use K/D/A or any other in game stats in its calculations or the “I was owning buy my team sucks” argument

Answer:

This encourages play that is detrimental to the team. Including:

1)Encouraging AFKing in the fountain to preserve record

2)Encouraging killstealing

3)Discouraging playing support and doing support actions (buying wards, etc)

4)Discouraging making sacrifices for the team (Vengeful Spirit swap, etc)

5)Encouraging stat farming instead of winning the game

This is a simple case of game theory where playing for individual profit is not necessarily in the best interest of the team. Only using win/loss in PSR calculation forces players to only make decisions that they feel are the best for the team (whether or not these decisions are correct is another matter entirely).

Complaint:

PSR can be abused by stacking accounts

Answer:

Basically, this consists of putting 1 highly rated player on a team with 4 extremely low rated accounts that are being played by very good players, in order to make the average rating of the team lower. This does not work because if your PSR is significantly higher (or lower in fact) than the average PSR in the game, you gain or lose very low points (approaches 0 the further away you are). This not only prevents this tactic, but it also allows high rated players to play with worse friends without risking 30-40 points every game.

I also read something about if you put 2 players on 1 side and 1 player on the other so it imbalances the win% and then having a team concede. This of course will eventually be settled with both automatic matchmaking and hopefully just disabling ratings in games with player imbalance from the start.

Complaint:

Stats (K/D/A ratio) do not show skill or the stereotypical “I only play support” argument

Answer:

K/D/A taken by itself does not necessarily show skill. Stats as a whole, do give a reasonable estimation of player skill. This means taking into account K/D/A, average game time, average creep kills/denies, win/loss ratio, and finally games played. In addition, even if you are playing support, you should never have a disproportionately high number of deaths. Playing support means not having the best ratio, but it is by no means acceptable to go 0/13/2 and say it’s OK because I play support. Good support players do not die unnecessarily.

On the otherhand, this has to be taken with a grain of salt. Ultimately, there is no 100% accurate method of measuring a player skill through in game stats.

For example, let's say a bad player with no map awareness would normally get ganked all the time. However, a better teammate is in lane with him and constantly tells him to back up when heroes are missing, thus preventing him from dying. If you used a K/D/A ratio, the worse player would get a ranking higher than he deserves.

Or what if one hero is the best at strategic planning, and orders his team around for the entire game. By all means, your team should have lost, but thanks to his good leadership and decision making, your team manages to surprise the enemy and thus win. How do you measure that?

In game stats can give a somewhat reasonable idea of a players skill, but in no way show the full picture.

Complaint:

PSR and stats should not be shown. My stats are very bad/very good and people kick me as a result.

Answer:

I have never heard of people getting kicked for being too good, unless it was for a noob only game. If they are, then those are terribad hosts. On the other hand, if your stats are very bad, either join noob only games, host your own games, get better to offset your record, or make a new account. I have the right to not play with bad players. You do not have the right to demand your stats be hidden so that you can sneak into games that you don’t belong in.

Complaint:

PSR and stats make it harder to start games or it takes forever to get 10 players.

Answer:

People do not want to play unbalanced games. It is simple as that. If you and 4 good friends want to play together, then you should only be allowed to play against 5 other random players of equal skill. Due to automatic matchmaking not yet being implemented, this may be difficult, because it is hard for good players to find other good players. However, you do not have the right to play against 5 random bad players and ruin their game experience because you enjoy owning noobs.

If you want to have a quicker game, then split your team up so that the game is balanced. Stat tracking and PSR are one of the main reasons I choose to play HoN over Dota. I’ve had enough games where Sentinel is clearly 4-5 friends that are there to pub stomp.

Finally, this is a BETA. When automatic matchmaking is implemented, it will make finding games infinitely better, as you won’t have 10 games all sitting at 8/10 players. You and your 4 friends can join matchmaking together and you will quickly be paired up against 5 other random people of equal skill.

Conclusion

I decided to add this upon reading some responses. Ultimately, the algorithm behind how changes in PSR are calculated (using ELO vs TrueSkill or anything else) and how you balance teams (taking the average of your teams PSR or some other weighted algorithm) are not what is being debated here. Better players will move up the ladder, while worse players will move down the ladder regardless of your algorithms. However, what is up for debate is what game stats should be used to calculate your PSR. I have argued as to why using win/loss stats estimate true skill, regardless of playing with friends or playing with feeders. I have also shown why using specific in-game stats, while can give a rough estimate of a player, encourage negative playstyles that are not team-oriented and hurt the game. Thus, the current PSR, while not perfect, is the best method of evaluating player skill available.

Source: http://forums.heroesofnewerth.com/showthread.php?t=8105

shta vi mislite o PSRu? dobra ili losha stvar?

[Coyote]

Da lepo objasnjeno, to sam i ja pomislio, zar ljudi ne vide ako je high skilled igrac u ekipi da ce im to mozda samo pomoci ako dovoljno veruju u sebe tj svoj skill

Edited August 13, 2009 by Coyote

[Sajro]

koliko sam ja skapirao taj PSR je idejno smishljen za onaj Quick Game ili kako se vetj zove, da te automatski ubaci u partiju sa ljudima i protivnicima tvog ili slichnog skilla (gledano po PSRu).

[pimpin`]

ma psr ti je, jednostavo recheno, obichna statistika, a statistike se kroz vreme izbalansiraju... i ta srednja vrednost koju dobijesh posmatranjem oscilacija neke statistike, posle nekog duzeg vremenskog perioda, potpuno je nebitna za tvoje shanse da u svakom datom trenutku, odigrash najbolje shto mozesh ili da potpuno podbacish... psr ti daje neku generalnu predstavu za shta si sposoban, a to ne znachi da cesh svaki put da odigrash na nivou svog psr-a, vec moze da se desi da podbacish ili da odigrash mnogo bolje nego shto neko moze da predvidi na osnovu tvog psr-a, shto takodje zavisi i od ekipe s kojom igrash, uloge koju treba da odigrash i mnogo drugih faktora, medju kojima je i mali faktor srece.

sve u svemu, dobro je shto postoji, ali ne treba mnogo obracati paznju na psr. na kraju krajeva random partije su failovi... ako neko zeli da igra igru malo ozbiljnije, treba da igra samo timske partije i tu ce psr sigurno da bude realan. takodje najbolje ce da te oceni tvoj tim, koji moze da ti kaze da li si zasluzio psr koji si s tim timom stekao ili ne...

to je moje mishljenje, ne mora da znachi da sam u pravu ;)

[pimpin`]

ma nisam spavao celu noc, morao sam neshto da radim =))

[Coyote]

pa za to si imao klinke iz lavaboa, a ne hon topic xD

[pimpin`]

ok, izvin'te

nisam hteo nikoga da uvredim =) obrishite sve, nije frka