Star Rating is even more broken now.

Their rating wouldn't be modified on the fly, only at the end of the game. The delta(ratingchange) would get updated with every kill. (On a side note, your rating gets updated if you leave a match early. It already seems to be updated on the fly or at least tracks stats in-game so they can be used when you leave.)

Simplified version: newRating = oldRating + RatingChange + deltaRatingChange.

Initial change is based on hard values that are unchanging snapshots of what the community "should" be capable of based on what has been demonstrated by players to be possible at varyingly reasonable levels. (IE 5-star should represent a high level of skill, 4 high but less than 5, 3 somewhere slightly above the middle, 2 somewhere below the middle, 1 much worse than the middle, and 0 absolutely abysmal.)

The adjustment to rating change (deltaRatingChange) is based on the circumstances that could have affected the stats that decided the initial rating change - mainly, the skill of players you killed/who killed you. The time spent in the game will be used to decide the initial rating change. IE so many kills/deaths/bot kills per minute for an x-star player.

Anyone who performs better than the range this system allows simply won't have that skill displayed, but the idea is to make the range such that very few (if any) will fall outside of it. That can be accomplished with the snapshots, IMO. Of course, if people begin to push in one direction, the snapshots can always be retaken and their values adjusted to fit the current community. Adjustments, however they're made (intentionally in long intervals or by the system constantly) need to be done such that nobody is pushed against a constraint, IE nobody ends up with the minimum or maximum rating as a result of the shift.

TL;DR: Very similar system to what you're suggesting with small differences. I think.

@original topic: I had the same thing happen to me this weekend after the update. I spent hours getting back from the reset 2.5 to 5.0. I took a break for a bit, came back and joined a server running tully's ( I love me some chicky ). Unfortunately the game was near it's completion as the enemy money all was down and at 25%. I rushed to get some kills and barely scraped together a 3-2-0-15 score and the game dropped my rating .7 points

Time simply seems to be ignored in he new system which is pretty bad. Even if it was a horrid performance, which it was not considering the sheer lack of time, why would it ever be acceptable to loose almost an entire point in rating???

You play like the best game of your life and you struggle to gain anything more than .3 yet you can join late and lose twice that and then some. Seems flawed to me.

Exactly the point of this thread. Glad to hear someone else has encountered this problem, but not to hear that your rating took a dive because of it. At least there's multiple accounts of it now.

In any case, I've been tinkering with a model rating system. Posting the basics here just because I feel like it.

Rating Change = (Initial Change + Adjustments) * Change Factor.

Initial Change = Kill Contribution + Deaths + Assists + Bot Kills + Turret Kills + Cash.
(Cut out "contribution" to save space.)

Kill cont. = -1.0 + (kills/(kills per minute * time * rating)) * rating strength
(IE: -1.0 + (kills you got/kills you should have had based on your rating and time played this game)
Death cont. = 1.0 - (deaths/((deaths per minute * time) / rating)) * rating strength
Everything else = -1.0 + (stat/(stat per minute * time * rating)) * rating strength

Adjustments = (1 - rating/avgopponentrating) * adjustment strength
(avgopponentrating is the average rating of the players you killed or who killed you, both are used to decide the final adjustment value.

Stats per minute are values assigned to estimate what a player should be able to earn per minute per star. Note that for deaths, this value is divided by rating because higher rated players should die less often.

Rating Strength is a value that decides how much influence a single stat has on your rating change.

Change Factor is a value that multiplies the final change, can be used to make changes more or less drastic overall.

In general, the system seems to work well enough, but I need to find good values for stats per minute and the various strengths. Each individual contribution formula generates a multiplier from +/-1.0 to +/- infinity (which would be ridiculously difficult to generate) which is multiplied by the stat's rating strength to decide what change it imparts on your rating. The Adjustments prod this value, the Initial Change, a bit higher or lower based on the players you killed/who killed you.

EDIT: The main flaw in this system right now is that a low-rated (1-2 star) player who kills far more than they need to for their rating or a high-ranked player who dies far more often than they should can gain or lose a ridiculous amount of rating if the rating strength and change factor aren't properly set. (2+ stars before rating strength of .05 is applied for kills/deaths in my testing)

Also... adjustments may be impossible to overcome if all the players on the other team are of a low enough rating relative to yours. Perhaps the average rating of your opponents should be looked at as well. adjustment = rating/(avgkilledrating*(avgkilledrating/avgopponentrating))

Fuckin happened to me again tonight man. I joined a game after I got home from work, I saw Dragonash was playing so I joined on him. Ended up getting on his team but he was already working the Moneyball... I managed to get 3 kills before the round ends... and... .8 rating drop... . ******* EIGHT

Sigh, I know rating shouldn't matter but the only response we get to bitching about it is "play more than 10 games before complaining about it" I'm sorry but im sick of playing 15+ games a day just to log on the next day and lose all of my progress and then some because I happened to pick the wrong time to join...

/giveup

you are not alone.

I enter a game where I step out and the moneyball is destroyed. Literally the second I leave the field, the game ends. I lose .4 points.

I get the last shot of the moneyball, in the high teens in assists, 7k to 3d (working the bots and turrets), a good number of bots... and I only go up .2.

Given that people typically leave when their ball goes down and bots keep storming in, it's easy to join at the "wrong time", giving this quite the awful potential.

Just for another data point, same thing has happened to me. I've easily played 30+ games since the patch. Went from the 2.5 reset to low 4s. I ended Sunday night at around 4.4X. First game I hoped into tonight ended within 1 minute of me joining. My team lost, I didn't have anything on the scoreboard and dropped down to 3.88 when we reached the lobby.

Nearly .6 points? Really?

The weird thing is, it seems to me that this just happens once per "session". This weekend I played several multi game sessions, but I don't recall my score dropping a large portion at the beginning of each session if I joined a game that ended quickly. Maybe that's something to look into.

But yes, system needs tweaking. Honestly it feels like the star rating went in the opposite intended direction this patch

Ya just played again... ended at 4.11 before I quit, first game was in for about 5 minutes, we won, about 7-x-4 30ish bots, dropped to 3.7.

I think I know why this was done. Possibly to give some sort of penalty for people who leave early. I guess it cant tell the difference between a short match and a match you leave.

I suppose the best fix would be to implement a few tweaks on how the system interprets performance over intervals of time.

I left a match after 5 minutes of play because I hate sniper duels like there's no tomorrow. My rating didn't budge at all. I was 2nd for cash at the time.

I'm wondering if each time you play (each session) it starts recalculating your star rating as if you'd played 0 games?

Ekanaut's comment of "play 10 games first" and the behavior I've been noticing (each time I start a new gaming session my rating tanks) makes me think it calculates it like so.

It would definitely explain why if your first game of a session is a over within moments and you have nothing on the scoreboard your rating tanks.

Went into my first game today at about 4.6

Only one guy in the server. I stepped out of the respawn room, killed him, then bots immediately finished off my moneyball

Afterward, I was at 3.9

No, the system just does not take time played in the match into account at all. You can play 0 games and join late and lose .6+ rating. You can play 10 amazing games, swift servers into a late game and lose .6+ rating. You can play 20 good games switch servers twice (luckily joining on transition or really early) then join one late and lose .6+ rating.

Let's face it. The system is still broken.

Things that need to happen:
A) % presence in the match MUST be taken into account.
B) The end stats page needs to show what each stat did in terms of contribution towards your rating.(modified of course by time present)
Ie: you scored 10 kills. (+.03)
20 assists (+.04)
15 deaths. (-.04)
100 bots(+.06)
earned 4123 (+.03)
bought/upgraded turrets : 0 (-.06)
Killed enemy turrets : 3 (+.03)
net rating change : +.09
C) support healing needs to contribute to rating gains. I have many friends who can barely break 2.8 and stay above it because of the late join issue combined with the fact that they SUPPORT their team instead of autocannon shotgunning everything.

I agree with your points especially that playing support means your star rating will not budge.

Supports do an important job of filling the cracks in offense and defense and do not do all of the killing of bots and pros for themselves. As a result I personally cant tell the difference between good support play and bad support play by scoreboard alone. You can possibly be kredit to team as a support unless you heal and support.

Also I think there should be some sort of benefit to supports for hacking turrets.

I think you should get extra points to the star rating for landing a GroundZero/HeadCrab from long distances
Was goofing around on SteelPeel earlier and while near the left side Ejector I lobbed a wild airstrike marker at a Gunner on the far end of the upper ring near their jumppad landing
Somehow I did it just perfect and GZ'd him which still managed to come down through the ceiling and smite him

I agree here.

Stars should be calculated based on other players in the server, remove the top and bottom player, average $ income per minute played, compare that to your score and adjust up/down accordingly. Something like that would work fine for a simple system.

I just had three people ditch a game saying stuff on the likes of just getting their rating up and they don't want to try to do it again.

This is what the rating system has done. Made people too afraid to attack the moneyball, made them focus more on k's, and actually promoted raging as a viable and worthwhile option.

You don't get rating just for killing people.

Well you don't get rating for healing, damaging the money ball, or anything of that sort.

Want to reiterate for the devs since the thread seems like it might sound a bit whiny: Rather than just remove the star system, please take some of the input from this thread (and others) and keep working on it. It's gotten better with this recent update, but there are still outstanding and crippling flaws.

EDIT: I just lost .4 rating two games in a row. Both ended within 10 seconds of my joining, and judging by the MVP stats, were full 15-minute games.

That's nothing I just joined a game where the game had ended and lost 0.7 rank.
Ranking system is so broken and this needs to be fixed!