This forum is in permanent archive mode. Our new active community can be found here.

Rocket League

1568101113

Comments

  • edited October 2015
    It was important to me that I point out that your claim "good lyrics don't help a song much" is objectively false.
    Post edited by Linkigi(Link-ee-jee) on
  • Is it just me or have players been getting better? Admittedly my rank went up slightly (Bronze II! I suck slightly less!) But I find myself infuriated with teammates who are complete garbage less often. Last time I played I won slightly more than half the games I played, but I lost like 3 incredible games in OT.
  • It definitely feels like folks are improving. Unfortunately I've also been seeing a few more trolls as of late, owngoaling and such.
  • I haven't played for like a week or so and got back in last night. It seemed like people were better, but then I had equally garbage games where I was amazed at how bad my teammates were. It's always surprising when a bad player leaves of a 3 man team and we actually end up doing better without that person.
  • Apreche said:

    This tangent couldn't possibly get further from Rocket League. At least we haven't hid Godwin's Law. yet.

    HITLER DID NINE (ELEVEN)!
  • Just played a bunch of games. I'm really frustrated with the ranking system. Not that I care what my rank is, because I suck and can never get a high rank. I just think it would be nice to at least be in the Silver so I can get matched up with higher quality teammates and have a better time.

    The problem is that almost every game, win or lose, I am the second or first best player across both teams. But because I'm winning about 50% of the games I play, my rank doesn't go up. It just fluctuates between like 150 and 225 depending on how many good or bad games I've had in a row.

    Hearthstone has the same ranking problem where it's not actually a ranking, but a grind. Rocket League has it worse because of the team aspect, and only considering wins and losses.

    If we only considered wins and losses that would make JJ Watt an awful football player since his team does not win very much. That would make Geneo Grissom top rank. Who is that? He's a young defenseman on the Patriots who barely plays.

    It should theoretically be possible to have your rank skyrocket if you do well, and still lose, in a game against other very strong competitors. It should also be theoretically possible to have your rank drop like a rock if you lose a game to some scrubs.
  • There have been some serious complaints about the ranking and matchmaking system lately. Like, for example, a win might be +50 toward your rank...but then a loss against the same people five minutes later will be -250.

    Something's screwy, and IIRC, they're looking at it, but no news of a fix yet.
  • Churba said:

    There have been some serious complaints about the ranking and matchmaking system lately. Like, for example, a win might be +50 toward your rank...but then a loss against the same people five minutes later will be -250.

    Something's screwy, and IIRC, they're looking at it, but no news of a fix yet.

    I've never seen swings that big. Every game I play my rank goes up or down 7, 8, or 9. Only once ever did it go down 10. That was bullshit.

    But the problem isn't the numbers. It's the entire design of the system. In real world sports we have a relatively new, but well understood, science known as advanced analytics. It is a system where you can accurately mathematically judge a players individual worth despite what team they play on and what opponents they play against.

    It's possible to use these analytics when all we have are statistics about real world games recorded by hand.

    In Rocket League we have perfect digital recordings of every single game ever played. The potential stats that could be calculated for each and every player are enough to make every MLB scout cream their pants.

    Wins, losses, none of that nonsense should matter unless you're in a tournament with your team. For ranking and matchmaking they should use very advanced metrics to judge a player's individual worth, and sort them accordingly.
  • Apreche said:

    I've never seen swings that big. Every game I play my rank goes up or down 7, 8, or 9. Only once ever did it go down 10. That was bullshit.

    Sorry, I meant points, not ranking. Nobody's ranking dropped by 250, don't worry, it's not that bad.

    But yeah, that's basically it - the entire system is pretty screwy. Like I said, they're working on it, but no news at this time as far as I know.

  • Scott. I've seen same "I could be better, but my team is dragging me down" argument in all and every team based video game with some kind of ranking and match making system.

    And the solution of using more stats to create more accurate system of assessing players is easy to say, but I don't think it has an easy solution. I'm sure that whoever could figure out an perfect algorithm for assessing player's relative skill in LoL, Dota or even Rocket League, would be a rich man.
  • Keep in mind, you may also be suffering from confirmation bias and the rosy retrospection effect.
  • I think if you're #1 on the losing team your penalty should be much lower. You demonstrated that you were the best on a bad team. And I know there are aspects where it changes based on the history of your opponents. That is, if the team you lost against is statistically worse than yours then your loss is much greater and vice versa. However, the problem that I would predict with the approach of making MVP on loser team be penalized less is if your team realizes that they're losing will most likely change strategy to be the top on your team instead of working together to actually make a comeback.

    I don't think I've ever left a match. For one, getting experience against real players vs the AI is much more valuable skill-wise. Another is that I've had games where we're down by 4 and tied it up or broke above in the last minute and a half. It's real easy to get cocky and let a couple goals landslide into a loss. Specifically I was on a team that was down 3 or 4 points. Both my teammates left. It was just me, the other team started fucking around and was sort of playing but thought they had it in the bag. I ended up winning the game because they didn't get serious until I was one point behind them and I bested them with another two points to take the win.
  • Apsup said:

    Scott. I've seen same "I could be better, but my team is dragging me down" argument in all and every team based video game with some kind of ranking and match making system.

    And the solution of using more stats to create more accurate system of assessing players is easy to say, but I don't think it has an easy solution. I'm sure that whoever could figure out an perfect algorithm for assessing player's relative skill in LoL, Dota or even Rocket League, would be a rich man.

    This would actually be an easy thing to program. I could easily create a system to simulate and test different ranking algorithms. Genetic algorithms also might be a good way to discover a good algorithm. But I do not believe there is big money in it.

    It goes something like this.

    Populate a database of players. Give each player an ID and a skill ranking. That skill ranking is their true skill level. You could even divide the skills into things like defense, attack, etc.

    Now create a simulation that has these players "play" games and track statistics for each game played. If you wanted to go all out you could actually run games with AI players at the given skill levels. It would be a lot less CPU intensive to just calculate the odds of each team winning, and player performance to generate some stats with some randomness that is affected by the player's "true" skill.

    You also have to simulate players quitting games. Some players playing a lot, and some playing rarely. Players with lag and those without. Throw in as much real world as you can. You also have to simulate player's skills going up over time, or going down if they have been away from the game for awhile, or just having a bad day, or a good day.

    Now you have enough stats you can pass them off to a ranking system. It can't see the "true" skill of the players. Just the stats generated by the game simulator. You know the true ranking of the players from 1-N. You just have to create an algorithm that can get as close to that ranking as possible based on the generated statistics. Also, after the initial ranking, provide that algorithm with a feed of new data coming from the simulation to adjust its ranking accordingly.

    While the math part is hard, this is actually something that only takes coding skills from Computer Science 101.
  • Am I just unable to read, or did I miss the part where you take the quite abstract concept of "player skill" and somehow transform it to a variable that computer can easily handle.
  • Apsup said:

    Am I just unable to read, or did I miss the part where you take the quite abstract concept of "player skill" and somehow transform it to a variable that computer can easily handle.

    Just use a set of heuristics from the player's inputs into the game. Start small and escalate.

    I can think of reasonable ones to generate a synthetic number easily enough.

    Also, on the team element, CounterStrike has done a very good job (relative to Rocket League so far) in terms of ranking individual players who interact solely though teams.
  • edited October 2015
    Apsup said:

    Am I just unable to read, or did I miss the part where you take the quite abstract concept of "player skill" and somehow transform it to a variable that computer can easily handle.

    This is a thing that can be done. It's called advanced analytics. It's real and it works. I don't understand why so many people do not believe this can be done.

    For example, in baseball there is a system called Sabermetrics. It consists of many different metrics for judging a player's overall worth. My favorite metric is value over replacement player, or VORP.

    https://en.wikipedia.org/wiki/Value_over_replacement_player

    Consider that at any time a baseball team can call up a minor league player at pretty much any position, pay them the minimum salary, and get them on the field. There are tons of such "replacement players" out there.

    But you don't have a replacement player, you have a professional. How much better is that professional than a hypothetical average replacement player? Are they twice as good? How many more runs can you expect to score with this player instead of a replacement one? How many fewer fielding errors will you make with the pro?

    You just have to ask these kinds of questions for Rocket League. If you have Scott on your team, how many more (or less) goals do you expect to score than if you had an average player? How many more or fewer goals can the opposing team expect to score if Scott is on your team? How much do you expect your teams time of possession to increase if Scott is on your team? How many more shots on goal will you have? How many scoring opportunities will you miss because you had Scott instead of someone who would actually hit that ball?

    If you ask the right questions, and measure the answers properly, ranking is not all that difficult. Think of it as reverse RPG. The character already exists. You watch them fight dragons and shit. You have all the data from many dragon fights. Now tell me, without looking at the character sheet, what are their STR, CON, DEX, INT, WIS, and CHA?

    TL;DR: You can absolutely assign empirical and extremely accurate numerical values for a player's skill when it comes to sports.
    Post edited by Apreche on
  • Apreche said:

    You watch them fight dragons and shit.

    STOP BEING A LAZY FUCKING ASSHOLE AND HELP KILL THE DRAGON YOU TWAT.
  • Just as guide, what would your numerical value actually measure? When it comes to measuring individual players, I can think of:

    Positive:
    - How many goals scored.
    - How many goal assists.
    - How many goal assists that don't end up scoring.
    - How many goals by the opponents blocked or saved.
    - How many opponents blown up.
    - How much touches to the ball in a row without it touching the ground or touching an opposing player.
    - Other bonus things.

    Negative:
    - How many goals allowed (closest player(s) to the goal when opponents score?)
    - How many goals on target of your own team deflected from the goal.
    - How many assists missed with an open goal.
    - If you let a ball in the air touch the ground after time and your team is behind.

    So even in a winning team, you could have a negative overall value, and when on a losing team you could have a positive overall value.
  • For more complex "individual player skill" heuristics, figure out a way to determine whether or not the player has:

    1. Ever
    2. In the last 20 days
    3. Regularly/reliably

    A. Performed an aerial maneuver and struck the ball while in the air
    B. Scored while performing an aerial maneuver
    C. Assisted while performing an aerial maneuver
    D. Performed a save while jumping or boosting

    That gives a score metric between 0 and 3, which can be used in aggregate. There are some things only skilled players can do at all, and only very skilled players can do reliably.
  • You can make a lot fancier metrics than that.

    1) How many times your cars accelleration vector is on a trajectory clearly intended to make contact with the ball, but does not hit the ball.

    2) How many times you make contact with the ball and the ball is next contacted by an opponent, as if you had passed it to them.

    3) Of all the times you make contact with the ball on your own end of the field, how often is it cleared or not.

    4) When a goal is scored by either team, of all the energy that set the course for the balls final trajectory vectors, how much of that energy was contributed by contact with your car?

    5) How much time does your car spend on the wrong side of the ball?

    6) How many fewer goals can you expect an opposing team to score while your car is positioned in or near your own goal?

    I could go on like this forever. It's all about asking the right questions.
  • Unlike sportsball, rocketball analysis has the full dataset. They know every input and result. Objectively.
  • It's not just asking the questions and having the data. The weight also matters.
    Which is more important, passing ball to person who makes goal, making a goal or maybe preventing opponent from making an goal?
    The person who just kinda nudges the ball to right direction is obviously less important than those people, but how much less, 50%, 20%, not important at all?
  • The relative significance of (and even complex interactions between) various metrics can easily be evaluated via statistical means, when you look at how much they actually impact on winning.
  • Apsup said:

    It's not just asking the questions and having the data. The weight also matters.
    Which is more important, passing ball to person who makes goal, making a goal or maybe preventing opponent from making an goal?
    The person who just kinda nudges the ball to right direction is obviously less important than those people, but how much less, 50%, 20%, not important at all?

    You know how when you are playing an RPG, like say Diablo, and you've got to choose your inventory? How do you weight the different metrics? The stabby knife hits more often, short range, can attack rapidly, does light damage. The magic sword hits less often, it's slower, longer range, and does crazy damage. Which is better? Which of these metrics gets more weight?

    Sometimes the numbers are just so far apart that it's no contest. Despite the swords slowness, it's just clearly superior. Sometimes even though each metric individually has a clear winner, it's a tough choice. Despite the sword being way slower, it's still a tough choice compared to the knife because of the difference in damage per hit.

    Someone who is insane on defense, and hardly ever allows goals, but couldn't score worth a shit. How good do they have to be on defense to make up for the fact that their offense capabilities are negligible? Well, imagine that sword. How much damage does it have to do to make up for its slowness compared to the knife?

    Once you've measured everything properly, questions like this become very easy to answer. As gamers you already answer them all the time. You just have to encode your own decision making process into a more strict algorithm.
  • Scott, are you sure it's just because of your teammates that you're losing? Having a high individual match score in Solo 3s, especially in Bronze, doesn't mean that, if you had better teammates, you would win more. Try playing a lot of ranked 1v1 games; how well you do there will definitely show you whether or not you are actually as good as you say you are, or not.

    The best part is that once you become real good at 1v1s, that skill transfers to Solo 3s where you can 100% carry your team to victory (even if they are potatoes).
  • Banta said:

    Scott, are you sure it's just because of your teammates that you're losing? Having a high individual match score in Solo 3s, especially in Bronze, doesn't mean that, if you had better teammates, you would win more. Try playing a lot of ranked 1v1 games; how well you do there will definitely show you whether or not you are actually as good as you say you are, or not.

    The best part is that once you become real good at 1v1s, that skill transfers to Solo 3s where you can 100% carry your team to victory (even if they are potatoes).

    1v1 and 2v2 are garbage games. I'm not going to play them. If I was Rocket League, I wouldn't have even made them an option. If you were making a baseball game, would you even consider making a play mode that wasn't 9v9? Or a soccer game that wasn't 11v11?

    I'm not upset at what my rank is. I'm upset that it only goes up or down based on whether my team wins or loses, and not based on anything else. I've had games where I've scored 5 goals, but we lose, so my rank goes down. I've had games where I score the only goal for our team, we go to OT, and we lose because one of my teammates makes an own-goal. I've had games where I score an own goal, but AutoPele happens to be on my team, so we win and my rank goes up. I've had games where everyone in the game is scrubs who are worthless, but the other team has AutoPele. I make some epic saves and score a bit, but it's not enough. My rank goes down, but the two scrubs on the other team who did nothing get rank up!

    Ranking should be an attempt to rank the players from best to worst. 1 through N. It shouldn't be a grind where you are trying to level up.

    Also, the separate ranking system based only on experience (rookie,pro,etc.) should be removed from the game entirely.
  • Apreche said:

    1v1 and 2v2 are garbage games.

    Those two are way more dependant on your own ability to play the game well than 3v3; the fewer players there are in the match, the more your own skill matters in determining the winner.
    If you were making a baseball game, would you even consider making a play mode that wasn't 9v9? Or a soccer game that wasn't 11v11?
    The sports analogy is tennis for 1v1 and 2v2, not baseball or soccer. Would you make a tennis game that wasn't 1v1 or 2v2?
  • Apreche said:

    Ranking should be an attempt to rank the players from best to worst. 1 through N.

    No, a ranking from 1 to N is not sufficient. At a bare minimum, you also care about the sizes of the gaps in between those players, and so you need a numerical score rather than just an index from 1..N.
  • Apreche said:

    Ranking should be an attempt to rank the players from best to worst. 1 through N.

    No, a ranking from 1 to N is not sufficient. At a bare minimum, you also care about the sizes of the gaps in between those players, and so you need a numerical score rather than just an index from 1..N.
    Yes, there's a numerical score obviously, and then you rank everyone based upon that score from 1 to N. Just like Tennis. But you are best off not showing players that numerical score, and just their ranking. Not only because it's cooler to know that you are the 2534'd best Rocket League player in the world, but because seeing the other number will affect player behavior in perhaps negative ways.

    For actual matchmaking you can use either number. With a middling player base of just hundreds of players, large skill gaps in the ranking are going to be very rare. Even if they do exist, say the 5th player is way better than the 6th, it's still fine for them to play with each other. Otherwise players near a gap will have a very hard time matchmaking. Also, players below a gap won't be able to experience that higher level of play they need to achieve to rank up. And the player with the big gap under them will be unfairly protected from falling in rank if they don't keep their skills up.

    The other big reason to only display the actual rank is that it makes rank a zero-sum game. For one person to rank up, another must go down. Right now there is no limit on how many people can be "Bronze II" or "Gold I". If everyone is a god at the game, then the game is too damn easy.
Sign In or Register to comment.