• This is one reason why XG is a piece of garbage.

    From MK@21:1/5 to All on Mon Dec 26 02:52:40 2022
    First, my understanding of how XG does things. Please
    feel free to correct/clarify if/as needed.

    1) According to XG's Dice Distribution, average equity for
    opening rolls, (excuding doubles), is 1642 / 30 = +0.0547

    2) XG keeps "luck rate" as a +/- zero-sum number for both
    players.

    3) XG keeps track of equities as gained/lost compared to
    the best move (even if the best move also loses equity).

    4) My comments follow ## signs.

    With that said, let's start by looking at the 5-move game at:

    https://www.bkgm.com/rgb/rgb.cgi?view+68

    ======================================
    Player1
    Dice-D for 31: +0.237 - +0.109 = +0.128

    Luck after roll: +0.234(0) -0.234(0)

    ## XG doesn't subract the average eqity from the opening roll
    ## Luck should be +0.237 - +0.0547 = +0.182

    Hint:
    1. Book 8/5 6/5 Eq.: +0.234
    5. Book 13/9 Eq.: -0.031 (-0.265)

    Equity after play: -0.265 +0.000

    ======================================
    Player2
    Dice-D for 44 : +0.446 - +0.026 = +0.420

    Luck after roll: -0.181(0) +0.181(1)

    ## It seems it added +0.415 not +0.420
    ## Luck should be -0.234 + +0.420 = +0.186
    ## This will ripple through all subsequent calculations

    Hint:
    1. 1-ply 24/16* 13/9(2) Eq.: +0.441
    12. 1-ply 24/16* 13/5 Eq.: +0.150 (-0.282)

    Equity after play: -0.265 -0.282

    ======================================
    Player1
    Dice-D for 32: -0.049 - -0.165 = +0.116

    Luck after roll: -0.067(0) +0.067(1)

    ## It seems it added +0.114 not +0.116
    ## Luck should be -0.181 + +0.116 = -0.065
    ## Luck should be -0.186 + +0.116 = -0.070 per prior correction

    Hint:
    1. 1-ply bar/20* Eq.: -0.046

    Equity after play: -0.265 -0.282

    ======================================
    Player2
    Dice-D for 31: +0.018 - +0.041 = -0.023

    Luck after roll: -0.080(0) +0.080(1)

    ## It seems it added -0.013 to Player1 for some reason??
    ## Luck should be +0.067 + -0.023 = -0.044
    ## Luck should be +0.065 + -0.023 = +0.042 per prior correction
    ## Luck should be +0.070 + -0.023 = +0.047 per prior correction

    Hint:
    1. 1-ply bar/22 6/5* Eq.: +0.059
    4. 1-ply bar/24 16/13 Eq.: -0.076 (-0.135)

    Equity after play: -0.265 -0.416

    ## Should be -282 + -0.135 = -0.417 a minor inaccuracy

    ======================================
    Player1
    Dice-D for 43: +0.140 - +0.075 = +0.065

    Luck after roll: -0.005(0) +0.005(1)

    ## Luck should be -0.080 + +0.065 = -0.015
    ## Luck should be -0.047 + +0.065 = +0.018 per last correction above

    ## If we tally luck figures manually after 5 rolls, we have:
    ## +0.234 +0.420
    ## +0.116 -0.023
    ## +0.065
    ## -------- --------
    ## +0.415 - +0.397 = +0.018 like magic, we end up with the same
    ## number above, after all my suggested corrections in between

    Hint:
    1. 1-ply 24/20 13/10 Eq.: +0.151
    8. 1-ply 20/13 Eq.: -0.107 (-0.258)

    Equity after play: -0.523 -0.416

    ======================================
    At this point, we have recycled back to opening position
    Now let's look at Nasti's opening 53 again

    Dice-D for 53: +0.085 - +0.109 = -0.024

    Luck after roll: +0.019(0) -0.019(1)

    ## This time XG subracts average eqity from "/opening roll/"
    ## But luck should be +0.018 + -0.024 = -0.006
    ## Instead it's calculated as +0.005 + -0.024 = -0.019
    ## and thus the error carries on...

    Hint:
    1. Cubeful 0-ply 8/3 6/3 Eq.: +0.083

    Equity after play: -0.523 -0.416

    ======================================
    And what about not-so-humorous luck ratings after each
    play in Nasti's example of 4 rolls? Here:

    After Nasti's 53
    Average Luck It's in the wrist (+0.0829)
    Average Luck Not your day (-0.0829)
    Total luck +0.83 points

    After Gnubg's 42
    Average Luck Quite lucky (+0.0147)
    Average Luck Quite unlucky (-0.0147)
    Total luck +0.29 points

    After Nasti's 66
    Average Luck It's in the wrist (+0.2132)
    Average Luck Not your day (-0.2132)
    Total luck +0.64 points

    After Gnubg's 62
    Average Luck It's in the wrist (+0.1726)
    Average Luck Not your day (-0.1726)
    Total luck +0.69 points

    ======================================

    I tried to be as sparing with words as I could.
    If anything isn't obvious to you, please feel free
    to ask and I will try to explain.

    Happy XG-mas...! ;)

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to All on Sat Dec 31 01:49:23 2022
    On December 26, 2022 at 3:52:41 AM UTC-7, MK wrote:

    .....
    At this point, we have recycled back to opening position
    Now let's look at Nasti's opening 53 again
    Dice-D for 53: +0.085 - +0.109 = -0.024
    Luck after roll: +0.019(0) -0.019(1)
    ## This time XG subracts average eqity from "opening roll"
    .....

    I am following up to my own article, in order to
    make it clearer and conclusive. It's a little long
    but easy to read and you will be well rewarded
    at the end. :)

    Initially I assumed that XG was calculating the
    luck rate per the common, (including Gnubg's),
    definition, i.e. adjusting actual equities of rolls
    by average equities of positions because it did
    so at the recycled starting position.

    Discussing this in the Gnubg thread, I thought I
    should verify it and also try to find out why my
    manual tallying never matched XG's figures for
    each individual roll.

    Sure enough, XG's definition in the glossary of
    its manual is different:

    "Luck: Luck is determined as the equity after
    "the roll (assuming the best move is played)
    " - equity before the roll.

    I also decided to recycle to the opening position
    twice to check if luck equity and skill equity will
    will zero out as they should.

    So, here are the same sequence of 5 rolls twice,
    omitting skill equity figures since those are okay.

    Please feel free to duplicate this for yourself on
    you own XG if that will convince you better.

    ======================================
    Player1
    Dice-D equity for 31: +0.237 (Dice-D = Dice Dictribution)

    Luck after roll: +0.234(0) -0.234(0)
    ======================================
    Player2
    Dice-D equity for 44 : +0.442

    Luck after roll: -0.187(0) +0.187(1)

    ## It added +0.421 not +0.442 (off by +0.021) ======================================
    Player1
    Dice-D equity for 32: -0.041

    Luck after roll: -0.071(0) +0.071(1)

    ## It added +0.116 not -0.041 (off by +0.157) ======================================
    Player2
    Dice-D equity for 31: +0.030

    Luck after roll: -0.051(0) +0.051(1)

    ## It added -0.020 not +0.030 (off by +0.050) ======================================
    Player1
    Dice-D equity for 43: +0.151

    Luck after roll: +0.014(0) -0.014(1)

    ## It added +0.065 not +0.151 (off by +0.086)

    ## At this point, we recycled back to opening position

    ## If we tally luck manually after 5 rolls, we have:
    ## Player1 +0.237 +0.000 = +0.234
    ## Player2 -0.237 +0.442 = +0.205
    ## Player1 -0.205 -0.041 = -0.246
    ## Player2 +0.246 +0.030 = +0.276
    ## Player1 -0.276 +0.151 = -0.125 (XG is off by +0.139) ======================================
    Now let's continue to reroll the same 5 rolls again,
    with Player2 on roll at the current position.

    Player2
    Dice-D equity for 31: +0.234

    Luck after roll: -0.108(0) +0.108(1)

    ## It added +0.122 not +0.234 (off by +0.112)

    ## Lo and behold! +0.122 is XG's average equity
    ## for the opening position and this time it did
    ## subract it from the equity of the 32 rolled. ======================================
    Player1
    Dice-D equity for 44 : +0.442

    Luck after roll: +0.313(1) -0.313(1)

    ## It added +0.421 again not +0.442 (off by +0.021) ======================================
    Player2
    Dice-D equity for 32: -0.041

    Luck after roll: +0.197(1) -0.197(1)

    ## It added again +0.116 not -0.041 (off by -0.157) ======================================
    Player1
    Dice-D equity for 31: +0.030

    Luck after roll: +0.177(1) -0.177(1)

    ## It added again -0.020 not +0.030 (off by +0.050) ======================================
    Player2
    Dice-D equity for 43: +0.151

    Luck after roll: +0.112(1) -0.112(1)

    ## It added again +0.065 not +0.151 (off by -0.086)

    ## Here, we recycled to opening position a second time.

    ## If we tally luck manually after 5 rolls, we have:
    ## Player2 +0.125 +0.237 = +0.362
    ## Player1 -0.362 +0.442 = +0.080
    ## Player2 -0.080 -0.041 = -0.121
    ## Player1 +0.121 +0.030 = +0.151
    ## Player2 -0.151 +0.151 = +0.000 Ta da!

    ## But XG is off by -0.112 which is the average that
    ## was subtracted from the 2nd opening roll above. ======================================
    If we continue to reroll the same 5 rolls once more,
    with now Player1 on roll, will it subtract again? Yup!

    Player1
    Dice-D equity for 31: +0.234

    Luck after roll: +0.234(1) -0.234(1)

    ## It added again +0.122 not +0.234 (off by +0.112) ======================================

    There you have it folks! Conclusive demonstration of
    your worshipped XG being a "garbage generator"... :(

    XG doesn't adjust equity of rolls by the average equity
    of positions, including the rolls for the initial opening
    positions, but some mysterious reason it does adjust
    them for the subsequent "recycled" opening positions.

    I could speculate on the reason for this one but I'll let
    you guys take a first shot at it... ;)

    Oh, also, XG doesn't just add the equity of the roll from
    the Dice Distribution table at the same ply either but it
    seems to be using some secret formula...(??)

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to All on Mon Jan 2 17:16:24 2023
    On December 31, 2022 at 2:49:25 AM UTC-7, MK wrote:

    XG doesn't adjust equity of rolls by the average equity
    of positions, including the rolls for the initial opening
    positions, but some mysterious reason it does adjust
    them for the subsequent "recycled" opening positions.

    I could speculate on the reason for this one but I'll let
    you guys take a first shot at it... ;)

    Come on, bozos..! Has someone duct taped
    your mouths..? If it is what I think it is, a bug
    like this could be the suspense in a film noir.
    Why are you foregoing the fun of solving it? ;)

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)