Search:

Type: Posts; User: ecsplos1v

Page 1 of 3 1 2 3

Search: Search took 0.01 seconds.

  1. Replies
    56
    Views
    2,053

    Poll: This is nothing to do with Microsoft. The...

    This is nothing to do with Microsoft. The researchers that worked on TrueSkill are some of the most well respected in the entire machine learning community and got their work published in NIPS the...
  2. Replies
    56
    Views
    2,053

    Poll: Does it? Picture of what? What measurements? You...

    Does it? Picture of what? What measurements? You just said you want to invent a new measurement system (your weighed 8 dimensional feature vector). Itís analogous to saying when we invent a new...
  3. Replies
    56
    Views
    2,053

    Poll: You still don't understand what I have said about...

    You still don't understand what I have said about the selective Elo stacking. Never mind my patience has more than run out on that one. I couldn't have explained it any clearer this time around so...
  4. Replies
    56
    Views
    2,053

    Poll: Really? I'm extremely sorry to hear these abuses...

    Really? I'm extremely sorry to hear these abuses exist for CTF and my heart goes out to you. Must be infuriating. I only know about the CA ranks, and have never played CTF so can't really comment.
    ...
  5. Replies
    56
    Views
    2,053

    Poll: Yes you are quite right TrueSkill shouldn't...

    Yes you are quite right TrueSkill shouldn't substract points when you win. However you need to take into consideration that what is shown to you is not your skill level, it's probably the...
  6. Replies
    56
    Views
    2,053

    Poll: Mate. You seriously haven't grasped a single...

    Mate. You seriously haven't grasped a single thing I have said. You manifestly don't understand anything about how Elo/Trueskill works. I did link you the online calculator so that you could prove...
  7. Replies
    56
    Views
    2,053

    Poll: Any weighting scheme is arbitrary as I said. All...

    Any weighting scheme is arbitrary as I said. All you would achieve is creating another game-in-a-game which would only serve to distort the main game further. This is why any ranking should be based...
  8. Replies
    56
    Views
    2,053

    Poll: Let's sum up a little bit. You still don't seem...

    Let's sum up a little bit. You still don't seem convinced that TrueSkill should work. I personally think that any weirdness is down to implementation/configuration issues. Let's park that for now....
  9. Replies
    56
    Views
    2,053

    Poll: No you won't lose Elo. I've explained this about...

    No you won't lose Elo. I've explained this about 1 million times already here and in the other thread [2].

    I suggest some of you guys experiment with the online TrueSkill calculator [0] just to...
  10. Replies
    56
    Views
    2,053

    Poll: -2 Elo is nothing. I modelled that game using the...

    -2 Elo is nothing. I modelled that game using the online trueskill calculator (see link below). Of course I had to guess what the uncertainty values were. It said the match quality was 0% (i.e. it...
  11. Replies
    56
    Views
    2,053

    Poll: Yes, thank you :) If you are playing with a...

    Yes, thank you :)

    If you are playing with a load of noobs, then it's expected you will lose, therefore as nothing unexpected happened, nothing changes in your ELO. We need to establish this fact....
  12. Replies
    56
    Views
    2,053

    Poll: This is priceless. I shouldn't even reply. But...

    This is priceless. I shouldn't even reply.

    But I will anyway.

    Quitting is the reason you have such a high ELO. Do you think someone who doesn't know how to dodge with an accuracy of 22% (that's...
  13. Replies
    56
    Views
    2,053

    Poll: I also balance games all the time. Most people...

    I also balance games all the time. Most people do. Your quit percentage is lower than mine (4.3% for CTF and 4.8% for CA), that's below normal in my opinion. It's only the ridiculous piss takers with...
  14. Replies
    16
    Views
    937

    Hi mate Did you read my second post? [#13]....

    Hi mate

    Did you read my second post? [#13]. Yes quite right FFA games converge almost instantly, especially if you beat a highly skilled player with small uncertainty -- this is the dynamic...
  15. Replies
    56
    Views
    2,053

    Poll: It's the same thing mate, completion == 1 - quits

    It's the same thing mate, completion == 1 - quits
  16. Replies
    56
    Views
    2,053

    Poll: Looking at the QL stats, it only tells you...

    Looking at the QL stats, it only tells you limited information i.e. whether you switched or quit. What we really need is the sequence of events. I suspect the QL team do store this information.
    ...
  17. Replies
    16
    Views
    937

    I must admit I was too quick to talk about the...

    I must admit I was too quick to talk about the improved convergence of TS over ELO (sorry "Elo" hah). While this is certainly true in FFA due to TS choosing the correct learning rate (which is fixed...
  18. Replies
    16
    Views
    937

    I think you misunderstood. I was not talking...

    I think you misunderstood. I was not talking about winning over a team of noobs. The way the TrueSkill algorithm works is that it only really updates skills if something unexpected happens. I was...
  19. Replies
    16
    Views
    937

    QLRanks CA ELO more exploits possible

    I'll make this brief...

    There seems to be an exploit on QLRanks where you can win a game, and then have the ELO update applied many times. It doesn't seem to remember that the update has been...
  20. Replies
    447
    Views
    141,533

    Seeing as this will harm QLRanks integration,...

    Seeing as this will harm QLRanks integration, high time iD sofware implements and integrates the ideas in QLRanks.

    iD software could do a better implementation as it has access to more information...
Results 1 to 20 of 58
Page 1 of 3 1 2 3