1. We have added a Gift Upgrades feature that allows you to gift an account upgrade to another member, just in time for the holiday season. You can see the gift option when going to the Account Upgrades screen, or on any user profile screen.
    Dismiss Notice

Do we dare to try to fix QScore?

Discussion in 'Civ4 - Hall of Fame Discussion' started by Denniz, Feb 6, 2008.

  1. VirusMonster

    VirusMonster Quechua General

    Joined:
    Mar 18, 2006
    Messages:
    619
    Location:
    at home :)
    Alright, sorry I don't know much about how you calculate the Qscore, so thank you for the clarification that instead of turns, years are used. Then, most likely the BTS scores for earliest conquest&dominations will be able to compete with Vanilla/Warlords scores, because BTS seems to have a higher number of turns and consequently, for a specific date in the early game, more number of turns would have passed in BTS than in Vanilla/Warlords. These extra number of turns could balance out that BTS is harder for earlier finish dates.

    In any case, to promote the number of BTS players, I feel BTS should not get a disadvantage for Qscore purposes.

    From the discussion in the other thead, I can summarize the following about BTS vs V/W scoring:

    Spoiler :

    edit: I found this thread on scoring discussion between Vanilla/Warlords and BTS. BTS vs. Vanilla/Warlords Discussion

    My conclusion of the discussion in this thread is that BTS and Vanilla/Warlords scoring systems are hard to compare. According to the thread, the math behind scoring in both BTS and Vanilla/warlords seems the same. Base score calculations (land, pop, tech, wonder) seem the same as well.

    BTS is harder to score, because of the better AI, defensive AI poprushing, espionage spending, the new tech requirements, slightly lower traderoute profits according to UnconqueredSun (see his post at traderoute profits are less in BTS, and exponential inflation (was linear in Vanilla). On the other hand, BTS has a few more turns added toward the end of the game, so total number of turns in a BTS game is more than a Vanilla game. Consequently, when you do the scoring calculation (current turn/total number of turns), a lower value is returned in BTS. This value increases the final score by some amount (sometimes more than *1.50) compared to a Vanilla game. Here are some graphs:

    date vs. number of turns graphs on different games speeds for Vanilla/Warlords vs. BTS

    and the original scoring system explained post:

    Civ4 scoring explained

    Let's not forget the random events... I think random events could both increase or decrease the final score in BTS, so I am undecided on their effect to the final score. For example, EliteSwords mission could be very strong if you get it while preparing your initial Praetorian rush. WastinTime thinks that most random events are beneficial and would help for earlier finish dates.

    I also feel the city revolt mission could be increasing the speed the cities are captured compared to Vanilla where you had to produce 10-20 cats just so you could reduce the city defences in 2-5 turns. I still use the siege for the collateral damage, but I wait less time reducing city defenses.
     
  2. azzaman333

    azzaman333 meh

    Joined:
    Apr 9, 2005
    Messages:
    22,877
    Location:
    Melbourne, AUS Reputation:131^(9/2)
    Not really. Space victories are already separated, and the only BtS tech which could adversely affect BtS diplos is Asthetics, but I'm not even sure that's along the path of a MM beeline.
     
  3. Denniz

    Denniz Where's my breakfast? Moderator Hall of Fame Staff

    Joined:
    Nov 28, 2003
    Messages:
    11,092
    Gender:
    Male
    Location:
    Dallas
    I am trying to get my head around this one. :crazyeye: On the surface, it seems that using turns should produce a result that is as consistent as using dates. I don't think we can compare the dates and turn directly, just the curves. The question is which represent relative performance better: Dates or Turns?

    The Curve:
    Spoiler :
    On thing I can tell you, is that in order to mix Min Turn and Avg Dates, it would require replicating/storing the turns to years calculations from the XML for each expansion/patch. :cringe:

    This has been chewed over quite a bit already. In a perfect world all the expansions would have their own tables. The reality is that that would create waaay too many tables and dilute the competition too much. Players are going to use whichever combination of leader, traits, map, etc. they feel will perform best. BTS is just another choice. It is not perfect but nothing is. End of story.

    Space is already in it's own table. Also, BTS has more turns than the others. I was using a factor to scale the turn number when comparing two games of different turn lengths. Turn = Turn * (Max Turns / #1 Game's Max Turns)
     
  4. Shadondriel

    Shadondriel Freak

    Joined:
    Apr 24, 2007
    Messages:
    82
    Location:
    70% Nuremberg, 30% Bangkok
    http://hof.civfanatics.net/civ4/tab...rathon&mapSize=Huge&dsply=0&limit=0&submit=Go

    Some calculations seem to be a bit off, at least from my view.
    In Old2 and Phantom2 Jimmy Thunder would get 103% O.o On Min+60 even 107%

    What's more amazing is, that 1430 BC (BtS) and 910 BC (WL) are only 2 turns apart on Marathon.
    When Jimmy Thunder beat me with his 910 BC I was amazed, but 1430 BC seemed just so impossible to me, until I realized that it is only 2 turns faster, but BtS :lol:

    No matter what, the system should be based on turns, not on dates, as it clearly shows how off dates can be.

    Evaluating the systems:
    -Old1 would give Jimmy Thunder only 90%, although the first spot was only 1% or 2 turns faster. This is just insane ^^
    -Old2 would give the second spot 3% more points than the first one, even weirder
    -25% seems to be ok
    -50% is too soft for my taste. I don't think I would deserve 93% for my game and I don't think a game finished in turn 443 should get 75% when the fastest is 207 (not even half of the number of turns)
    -phantom2 and +60 seem to be broken and way too soft
     
  5. Denniz

    Denniz Where's my breakfast? Moderator Hall of Fame Staff

    Joined:
    Nov 28, 2003
    Messages:
    11,092
    Gender:
    Male
    Location:
    Dallas
    I was aware of the possiblity of that happening. Since BTS Marathon is 300 turns longer than WL, you are seeing the effect of the scaling factor: 209 * (1200 / 1500) = 167.3. The 209 probably should be the FF with the BTS 207 being scored as 207 * (1500 / 1200) = 258.75. I haven't really worked out whether that wpuld be fair or not, though. :(

    I liked 25% but it does leave a lot of games out in the cold with 0.00
    I am still tweaking Phantom2. I am trying to add the ranged Phantom date to the Old2 scoring like Phantom does to Old.
     
  6. FiveAces

    FiveAces Emperor

    Joined:
    Jul 2, 2007
    Messages:
    1,151
    Location:
    Singapore
    You could tweak that though, right? Apply the 25% curve as-is through 15-20% and then stretch it from there to +50% or whatever?
     
  7. Denniz

    Denniz Where's my breakfast? Moderator Hall of Fame Staff

    Joined:
    Nov 28, 2003
    Messages:
    11,092
    Gender:
    Male
    Location:
    Dallas
    The problem is engineering a smooth transition. It is likely that a game just below the cut on the 25% curve would end up with higher score than one just above the cut.
     
  8. Thrallia

    Thrallia Prodigal Staffer GOTM Staff

    Joined:
    Dec 16, 2005
    Messages:
    2,836
    Location:
    Maryland
    ah, I had forgotten that BtS space was in different tables than van/war space...and if you scale for the turn difference, then it seems to work fine.

    So far, my favorite is still Phantom, as it seems to give correct scaling to the QScores...Phantom works as I believe it should...scaling unfilled tables well, and not really impacting filled tables.
     
  9. ParadigmShifter

    ParadigmShifter Random Nonsense Generator

    Joined:
    Apr 4, 2007
    Messages:
    21,810
    Location:
    Liverpool, home of Everton FC
    How about a phantom date same as the worst date? That will balance 2 entry tables a little and hardly affect full tables at all.
     
  10. Shadondriel

    Shadondriel Freak

    Joined:
    Apr 24, 2007
    Messages:
    82
    Location:
    70% Nuremberg, 30% Bangkok
    Hum "25%" seems to grasp it better.

    Building on ParadigmShifter's idea:
    What would happen if the "avarage" calculates only the avarage of the best 10 games and unfilled spots would be phantom games?

    BTW. what is 25%+adj? It seems to prevent a 0 in the top 10
     
  11. Denniz

    Denniz Where's my breakfast? Moderator Hall of Fame Staff

    Joined:
    Nov 28, 2003
    Messages:
    11,092
    Gender:
    Male
    Location:
    Dallas
    Phantom date and scaling for BTS vs V+W turns are different things. Phantom2 was my first attempt at combining the two by using a turn-base calculation. I haven't got the ranges for the phantom avg turn right yet. Plus I am not sure about the avg turn.

    As for whether the date based qscore needs to be adjusted for BTS turns, I am not sure it needs it. Theoretically, they balanced the dates when they distributed the years per turn in the game. I do know that converting turns to dates is would be a real pain in the you-know-what when you consider all the little changes they made over different patches.

    That would just weight the table toward the worst date. Part of what phantom date is supposed to do is fix the effect caused by one really late date on a table with just a few games.

    You thing people hated the 0 qscores caused by 25%? This would pretty much guarantee bad qscores for anyone outside the top 10. This is the challenge of the Qscore, producing a consistent result if there a 2-3 entries or over 100.

    It is my attempt at implementing FiveAces idea of tweaking the 25% score. It uses 25% score down to 10 points where it switches to the 1/10th of the 50% score. For those that would still be zero it gives them 0.100 points.
     
  12. FiveAces

    FiveAces Emperor

    Joined:
    Jul 2, 2007
    Messages:
    1,151
    Location:
    Singapore
    Think of it like stretching a picture - x-axis is turns (+0% to +50% (or whatever) , y-axis is score (0-100). Overlay the +0-25 curve (after 25 on the x-axis there is no curve since the value is 0). Take a "picture". Now cut that picture into two smaller pictures - say from 0-20 and 20-25. Now grab the right edge of the 20-25 picture and stretch it out to 50, as if you were stretching a 3x3 picture into a 3x5. The score at the breakpoint (+20% turns in this example) remains the same. The scores from 21-25 increase as compared to a pure +25% curve, but are below the +20% score. And the scores from +25% onwards now have a value.

    Don't know how you would code this though, so that could be a problem.

    EDIT: Just saw you're working on this already. Feel free to ignore. :)
     
  13. shyuhe

    shyuhe Deity

    Joined:
    Nov 30, 2006
    Messages:
    6,062
    Location:
    Gone fishing for the summer
    I don't know if this solution works but how about taking the 25% rule and modifying scores that fall in the 20-25% rule against a linear line? So the 25% rule calculation applies for 0-20% (as others have suggested). For the remaining 5% and beyond, take the 5% point and a phantom date of 2050 AD (however many turns that is). Make a straight fit between those two points and assign scores along that line. This will ensure that scores falling outside the 5% rule get some points (probably in the vicinity of 5-10 points).
     
  14. Denniz

    Denniz Where's my breakfast? Moderator Hall of Fame Staff

    Joined:
    Nov 28, 2003
    Messages:
    11,092
    Gender:
    Male
    Location:
    Dallas
    Have you looked at the 25%+Adj formula results? The details are different but every game gets something. Even they are more 50% of the total turns in the game away they get 0.100.


    Anything that far out is not what I would call a quality game. This supposed to be a Hall of Fame. Not the Hall of Everyone gets a Trophy. Which concidering the size of the accronym is probably a good thing. :mischief:

    +++++++++++++++++++++

    At this point, it is down to the 25%+Adj or Phantom and I am leaning towards 25%+Adj. So unless someone can find a major problem with it, we will probably implement it for the next update.

    +++++++++++++++++++++

    The next thing on my list is balance. No matter what we do it is going to be complicated to implement if we want to make balance an entrance requirement to the EQM. I think that is the only way to keep EQM from being just another check off the list competition like QM.

    A couple ideas I am toying with (these are possible alternatives):
    • Limit the number of games that can come from any one table to two.
    • Set a maximum number of games for any one Victory condition. (I figure that no more than 1/7th of the total games should be of any one victory condition. 1/5 for Van+Warlords. i.e. Time and score not limited.)

    I was also thinking of a 30-40% max of one speed and a max of 40% of one mapsize. Of course the table or vc filter might make these redundant.

    RoA and Guantlet would not count towards those limits. RoA would probably have to have limits of it's own.

    Finally, the thougt of having a minimum score requirement (non-RoA) is also a potential filter that might work to exclude cheese. Of course, that assumes that there aren't to many single entry tables scoring a 100.

    Edit: These are potential alternative ideas, trying to do them all would not be practical.
     
  15. ParadigmShifter

    ParadigmShifter Random Nonsense Generator

    Joined:
    Apr 4, 2007
    Messages:
    21,810
    Location:
    Liverpool, home of Everton FC
    There are only 7 victory types in vanilla & warlords ;)

    1/7th seems a bit low anyway, surely alot of wins are going to be dom and space.

    I play all my games with HOF mod on and submit any I win. Space and dom are my bread and butter VCs.
     
  16. WastinTime

    WastinTime Deity Supporter

    Joined:
    Jan 27, 2006
    Messages:
    15,238
    Gender:
    Male
    Location:
    California
    Denniz, you need to stop, take a deep breath and look at the confusion you're creating. I can't even keep up with the formulas/suggestions and I've been reading along the whole time. And for what? What is really getting fixed? QScore isn't that far off, one SIMPLE fix and it can be forgotten. The current scoring doesn't deter people from playing anyway. You should focus on attracting people to the competition (<==most important) and making it a noteworthy acheivement.
     
  17. ParadigmShifter

    ParadigmShifter Random Nonsense Generator

    Joined:
    Apr 4, 2007
    Messages:
    21,810
    Location:
    Liverpool, home of Everton FC
    Whoah, 75% of my games are on standard sized maps and 60% are at normal speed as well.

    I agree with WastinTime also.
     
  18. Harbourboy

    Harbourboy Deity

    Joined:
    Jun 27, 2003
    Messages:
    2,998
    Location:
    North Harbour
    I'm completely lost on what has happened in this thread. The only part I understood was that many of my games would count for nothing. The rest of it has gone way over my head.
     
  19. Shadondriel

    Shadondriel Freak

    Joined:
    Apr 24, 2007
    Messages:
    82
    Location:
    70% Nuremberg, 30% Bangkok
    Yeah, 25%+Adj looks fair imho

    -The first point would be all right, I guess.

    -However, the second point really confuses me since I would never know what games count and what don't. Besides that, if only one game per table counts, that would limit it to a maximum of 20 games that can be of any victory condition anyways. (per difficulty level, of course)
    If you aim for 100% games that would be even worse, if you already have 1/7th filled with 100% games and don't realise it. For example: "yaay, fastest conquest on [game #1], hey my 100% Borealis game is gone :nuke: "

    Or do you mean, that for example if you have 200 dom, 100 dipl, 100 conq and so on games, that only the 100 best domination games would count? I guess that would still be all right :confused:
     
  20. Ozbenno

    Ozbenno Fly Fly Away Moderator Hall of Fame Staff

    Joined:
    Apr 5, 2006
    Messages:
    11,442
    Location:
    Sydney, Australia
    I've also completely lost track of what is going on here. I really don't think there is that much wrong with Qscore as it is now (possibly only a problem with tables with 2 entries only). I think things are getting too complicated and I'm not sure it is going to make things any better.

    I've actually stopped playing HoF games as I have no idea what will count and what won't. I'd be happy to let things stay as they are, as I then know what I'm up against.
     

Share This Page