This topic contains 36 replies, has 0 voices, and was last updated by  naka 9 years, 1 month ago.

  • Author
    Posts
  • #6628

    BOY
    Participant

    @ Peanut:

    I rearanged your equation to analyze what it means in detail:

    1000 * { (Kill-per-Rnd) * 10 (Gm-Wins/Shots + 1) } I think that’s right.

    Part of this was good, part bad. I realized that Shots per game-win was a really good idea. The innacurate shooter who is only missing so that he can pound out the win and because he survives to shoot farther and farther oponoents should show an increased game win percentage. Shots per game win shows that.

    However, a player who does not win much but does tend to play a few rounds here and there doesn’t get the bonus. This is where round win percentage comes to play. Round w ins are the most significant factor in game wins, so a player who has little time on his hands for a full game needs consideration.

    Okay so we may need Game Wins and Round Wins as well, with Game wins as the important factor.

    A good caclulation will have little or no arbitrary weight factors (such as (K + 10 * GW) It would instead use an operator such as + or – or * or /.

    I am still exploring a way to factor in all this without a significant arbitrary numbers added.

    Try this on for size. Kill/Rnd + GW*RW/Shots^.5

    The first term is the basic efficiency, the second term does the following. By multiplying Game Wins and Round Wins the skill of the very good player is compounded if he can win it all. Dividing this second part by square root of shots does this: it allows the most efficient shooters to have an advantage in thier stats, but only if they win. A winning player will often shoot more, but if he can shoot less and still win, he gets a larger skill score. By using the square root, all players a penalized a bit less for shots and also evens out both of the terms in the equation and demonstrates that shooting more does not always mean less skill.

    The resulting number can now be multiplied by any factor to make the score appear better. Such as 1000 or 100. But no since stat is given any arbitrary weight.

    I will post some calculations when I have more time.

    #53462

    PeanutsRevenge
    Participant

    I did not use shots, but shots per kill.

    If using shots, then people might start to skip more if a kill is less than certain and there is no threat.
    A way to get around this might be to start splitting round win points between all players that survive a round, thus encouraging people to really try to be the last tank standing. (Which I think should happen anyway).

    I have been inching more towards your previous method than my own with more and more thought.
    However, I am thinking of adding opponent rank into the equation.
    With the new systems being discussed, it would show the additional skill in killing better players and also stop people only playing newbs/bad players in order to gain a higher rank.

    I had not been thinking of this as at the moment rank does not show a players ability.
    Ideally, I would like to add average player skill per round into the equation (per shot or kill doesn’t mean very much and per game is too up n down).

    I see what you mean about going my round wins, as many players don’t always have a full game, however, those that join late in a game will suffer anyway due to lack of finances (even tho many people gift late arrivals), so I do not think it would help much.

    BTW, could u move these posts to the ranking thread and keep this one more about whether it should change than about the changes themselves. 🙂

    #53463

    Raden
    Participant

    @boy wrote:

    First of all, using that many nukes or sandhogs is impossible with the current economic settings. It is all about the economy Players rarely use nukes because of the price.

    It is not entirely impossible but can happen when 2 or 3 players only play.
    @boy wrote:

    Second of all, what it leads to is a higher kill ratio and more kills. Which is what every player wants. Good players use baby missiles to save money, or because they want to look impressive.

    🙂 it can’t lead to anything if it is impossible to happen as u say…but I’m just kidding with u BOY. I understand what you’re trying to say…..
    @boy wrote:

    I really don’t mind if the money awarded for higher arms levels is less, but I don’t think it should affect skill. The economy does the work to demonstrate skill. No arms levels are needed in skill calculation if economy is right. (I know I’m a broken record, but there are many readers and only one of me :P)

    Hmmmm…..OK, you’re right. Good point BOY, didn’t see it this way before, when I thought up my calculation method.

    #53464

    BOY
    Participant

    Okay, perhaps not totally impossible, just extreemly rare. 😉 Thanks for understanding me Raden.

    #53465

    BOY
    Participant

    Okay, split the topic as requested. 🙂

    As to the idea of using shots per kill, yes, you did use that (F2), but you DIVIDED by a fraction in which kills was the demoninator. It is no different than muliplying by kills per shot (the kill ratio)

    =SUM((C2/(E2/10)/F2)+(B2/E2))*1000 <—- from your spreadsheet

    I rearanged by dividing by factoring out Kill per round, which appears twice because your term F2=Shots/E2 where E2 is kills.

    According to Excel you have the equation set up as I rearanged it. Are you sure it is set up as you wanted it?

    My latest thought is this:

    FIRST TERM = THREAT RATING
    (Kills – SK + 25)/(R+100)

    THis term allows for all new players to have a threat rating of 0.25 This is also so that a single good game cannot allow a player to have a super high rating. A good players rating will be around 1.5 or 2. and it will take a few hundred kills at least to reach tha high level as the original numbers of 25 and 100 become insignificant.

    SECOND TERM = SURVIVAL RATING
    (GamWins X RndWins)/Shots^0.5

    This term shows the survival and winning skill of the player. A player with zero game wins will have a zero rating added to their overall skill. Shots are used as a basic. A player who wins half thier games might get a rating somewhere around 0.1 or 0.2 and this is added to the first rating sort of like a bonus.

    I have a little more to do on this idea…. 😀

    #53466

    imported_gcamp
    Participant

    Not sure all of these algorithms are going to be easy to implement but I am willing to give it a go.

    When do you think you will have something to trial/test?

    #53467

    BOY
    Participant

    after some actual spreadsheet testing, the above idea does not work. It is actually better to simply do away with the second term and then it does work quite well actually.

    Since the skill system already awards for round wins and for game wins, the new sorting system wouldn’t really need a second term anyway, it would be redundant.

    the two options on peanuts spreadsheet are pretty good too.

    But before you do anything I should ask what you mean by algorithm. Is it not possible to simply resort the server rank by a new simple statistic such as some kind of ratio? Isn’t the server rank queried by the skill algorithm determined by total kills?

    The easiest thing I could think of was just resorting the ranks for the skill calc.
    ie. leave everything as is, except don’t use kills as the rank considered by the skill system and possibly rework the values awarded for wins and such.

    it is getting late for me… I’ll work on this more tomorrow 😛

    #53468

    Thrax
    Participant

    @boy wrote:

    The easiest thing I could think of was just resorting the ranks for the skill calc.
    ie. leave everything as is, except don’t use kills as the rank considered by the skill system and possibly rework the values awarded for wins and such.

    Ie.. The method currently being tested on the Mod-servers?
    That is what I did; changed the Rank sorting to count by the existing “Skill” trait, instead of the existing “Kills” trait.

    It still counts each Kill, but applies the change to both the Shooter (+) and
    the target(-), with the ammount changed set by a weapon level factor.

    In effect, killing 2 players with BBM’s will gain you more “Skill” than killing 10
    players with Death’s Heads. While at the same time, Lowering the targets Skill;
    proving that you out-smarted them and caused them to decrease Skill.

    Hence, if you want a Top-rank player to be removed from his post… Kill him
    a lot..

    #53469

    BOY
    Participant

    I have another question, exactly how much is the difference in points for killing a high player versus a low ranked player?

    Your system is an improvement, I’d say, by partly fixing the issue of time played problem. Does the algorithm tend to favor players with a lot of game wins and round wins regardless of total rounds played?

    #53470

    Thrax
    Participant

    @boy wrote:

    I have another question, exactly how much is the difference in points for killing
    a high player versus a low ranked player?

    Your system is an improvement, I’d say, by partly fixing the issue of time
    played problem. Does the algorithm tend to favor players with a lot of game
    wins and round wins regardless of total rounds played?

    Gavin’s current Skill code does not descriminate on high-rank vs low.
    Each player involved has his rank added or removed equally. If 15 points
    is added to the victor, then 15 points is removed from the target.

    It does still favor those that collect a larger number of kills, if they also
    haven’t been killed a substantial annount. Since a defeat lowers your points,
    long-time players may even drop rank if killed in return.
    Eg.
    Stats on mod-servers by Kills; top-4
    Name


    Skill


    Kills
    MichaelMr….1915……..419
    2 Thrax……1545……..340
    3 scar……..2089……..213
    4 (Ai)Fred…336………198
    I have a lot of kills. But since I also get killed, my ‘Skill’ isn’t as great.
    Scar killed the bot often. Hence his skill rose, and the bot’s dropped.

    side-note: Also means.. I should make the bots smarter so they can take
    out scar and Mike.. 😀

    #53471

    imported_gcamp
    Participant

    @thrax wrote:

    Gavin’s current Skill code does not descriminate on high-rank vs low.
    Each player involved has his rank added or removed equally. If 15 points
    is added to the victor, then 15 points is removed from the target.

    Not quite true. It currently calculates the difference in skill between the killer and victim and uses that as the change factor. If the killer is lower in skill than the victim, the killer recieves more skill than if the killer is of a higher skill than the victim. The bigger the difference the more the change.

    The actual algorithm is this:-
    = / ( 1 + pow(10, ) / 1000 )

    So yes it is affected by the number of kills, but it does use both skill difference and weapon used.

    Note: This type of skill information is recorded when the shots are made and cannot be calculated after this event has taken place.

    The added skill for rounds and overall winners etc. was added by Bob (I think) and put directly into the php stats pages code. I probably would have got the server to calculate and store it instead.

    #53472

    Thrax
    Participant

    @gcamp wrote:

    @thrax wrote:

    Gavin’s current Skill code does not descriminate on high-rank vs low.
    Each player involved has his rank added or removed equally. If 15 points
    is added to the victor, then 15 points is removed from the target.

    Not quite true. It currently calculates the difference in skill between the killer
    and victim and uses that as the change factor. If the killer is lower in skill than
    the victim, the killer recieves more skill than if the killer is of a higher skill than
    the victim. The bigger the difference the more the change.

    The actual algorithm is this:-
    = / ( 1 + pow(10, ) / 1000 )

    So yes it is affected by the number of kills, but it does use both skill
    difference and weapon used.

    Note: This type of skill information is recorded when the shots are made
    and cannot be calculated after this event has taken place.

    The added skill for rounds and overall winners etc. was added by Bob (I think)
    and put directly into the php stats pages code. I probably would have got the
    server to calculate and store it instead.

    I stand corrected. 😀
    Great. must have missed that part on my initial scan.

    Was curious about that tho. If a pro kills a beginner, does the beginners skill
    drop greater? Or, if the beginnner kills the pro.. does the pro’s drop more.
    Then if both parties are similar skill, the skill change is less?

    I would have thought the reverse. Equal parties would have a greater
    impact on eachother, and a beginner killing a pro would show more skill
    than the reverse.

    Definately a tricky calculation to fiddle with.

    #53473

    imported_gcamp
    Participant

    Yeah, thats the one.

    If a beginner kills a pro the beginner gets more points and the pro loses more points.
    If a pro kills a beginner then the pro gets less points and the beginner loses less points.
    It’s the diference in skill used. So yes if they are both the same skill an average change is affected.

    When this doesn’t work is when a new player joins who is actualy pretty good. He will take a lot of points of the others until it evens out.

    You also get more points the small the weapon used for the kill.

    #53474

    BOY
    Participant

    Okay, so must be a ratio of the skills, if it were a subtraction it would become a rather large exponent and the skill change would wind up being nil most of the time.

    I was also under the impression that the “difference” was kills ranking. If this is not the case any more all the better. I must have been asleep when the change was made or was confused all along.

    However, here are some problems with the equation as it stands IMHO.

    1. Players do not regard it as an accurate rating.
    Although the trends in the skill points to point generally toward the better players they don’t seem to make a lot of sense between known players. I believe this is because players have always tended to gravitate to understandable ‘hard’ stats such as kill ratios to determine skills.

    2. The system favors players who play a lot.
    Take two players of close skill level who both get more kills than deaths thereby gradually increasing thier skill rating, They also have similar win ratios. Although the higher in rank they get, the slower the gains are, they always increase, with the usual back and forth. The player who plays less has a disadantage.
    -Points gained from Game wins and round wins should be divided by the amount of time played. The basic unit of game play is rounds.

    3. unless consistently 1 v 1 – The idea is flawed.
    Players should not be penalized for dying. Why is that anyway? I see little reason why dying demonstrates a lack of skill. Kills say a great deal about the ability of the shooter, but deaths are a result of kills and therfore also only say much about the shooter. A kill MAY show that player A is better than Player B, but it is not a reason to deduct skill from player B.

    I have more, but I figured I should end it there for now.

    #53475

    Laptops Daddy
    Participant

    seems to me that any skill calculation should be based on shots per hit. thats really all it comes down to. couldnt you just change the code to reflect hits rather than kills?

    no doubt a skillful player will stand more chance of popping a bubble by selecting appropriate weapons (this is also a reflection of skill to some extent), but having your projectile go where you want it to is the key to doing well, and is the only real measure.

    so, ‘shots per hit’? maybe with a few allowances/deductions for aimed weapons?

Viewing 15 posts - 1 through 15 (of 38 total)

You must be logged in to reply to this topic.