Boxing decisions by Artifical Intelligence

Discussion in 'World Boxing Forum' started by CleneloAnavarez, Apr 20, 2023.


  1. CleneloAnavarez

    CleneloAnavarez Well-Known Member Full Member

    1,795
    1,471
    Nov 18, 2021
    A new software startup named Jabbr.ai has created an application called Deep Strike which uses deep learning to count punches and recognise their impact, aldo quantifies attributes like aggression and pressure.

    I have seen analysis of several well known fights and have been impressed by the technical breakdown. You can check it for yourself on Instagram.

    https://www.instagram.com/jabbr.ai/
     
    McGrain and Keleneki like this.
  2. 007 373 5963

    007 373 5963 Active Member Full Member

    944
    1,718
    May 30, 2020
    I can already see the wiseguys bribing the AI with more RAM.
     
  3. heerko koois

    heerko koois Obsessed with Boxing

    24,294
    18,047
    Apr 26, 2006
    dos 4.2 would be smart enough to pick GGG in the first 2 fights.....
     
  4. AngryBirds

    AngryBirds Well-Known Member banned Full Member

    1,840
    2,024
    Sep 3, 2022
    They would also probably threaten to take RAM away from the AI if didn’t give their guy the win.
     
  5. Sandman_

    Sandman_ Undisputed Full Member

    5,192
    6,783
    Feb 20, 2009
  6. Dangerwood84

    Dangerwood84 Boxing Junkie Full Member

    9,484
    14,104
    Sep 21, 2017
    So we'd still get the same decisions!
     
  7. WhataRock

    WhataRock Loyal Member Full Member

    35,327
    18,871
    Jul 29, 2004
    Imagine asking AI what they think the term "floored" means. It would implode.
     
  8. FrankinDallas

    FrankinDallas FRANKINAUSTIN

    30,905
    37,886
    Jul 24, 2004
    An Atari 2600 console would have scored both fights for GGG.
     
  9. WhataRock

    WhataRock Loyal Member Full Member

    35,327
    18,871
    Jul 29, 2004
    A Babbage Difference Engine would have scored both fights for GGG.
     
  10. Saintpat

    Saintpat Obsessed with Boxing Full Member

    23,747
    27,399
    Jun 26, 2009
    It’s called GIGO.

    Garbage In, Garbage Out.

    You program something wrong and it gives you wrong results.

    Boxing in every jurisdiction is judged on four points. Per your post, not one of those four points is part of the programming.

    The four points are:

    1) Clean and effective/impactful punches landed.

    2) EFFECTIVE aggression. Coming forward means nothing. Coming forward effectively (i.e. cutting off the ring, closing the distance, etc.) is what this means.

    3) Ring generalship. Who is dictating the terms of the fight. If you’re following Ali around as he floats like a butterfly and stings you like a bee, he’s the ring general. If you’re Joe Frazier bullying him to the ropes and banging away at his head and ribs, you’re the ring general.

    4) Defense. Making the other man miss, taking the sting off punches so they don’t land cleanly or effectively so only glance, parrying or blocking, slipping and ducking under them.

    It’s a worthless exercise if they don’t even know how fights are scored. But then again, a lot of people on this forum have no idea how fights are scored.
     
  11. catchwtboxing

    catchwtboxing Obsessed with Boxing

    28,345
    38,094
    Jul 4, 2014
    How will Canelo continue to be marketable?
     
  12. 007 373 5963

    007 373 5963 Active Member Full Member

    944
    1,718
    May 30, 2020
    An abacus would have scored both fights for GGG
     
  13. CleneloAnavarez

    CleneloAnavarez Well-Known Member Full Member

    1,795
    1,471
    Nov 18, 2021
    We are not talking about deterministic algos. Deep learning methods are trained with huge datasets, they evolve with use.
    They know how to calculate ring generalship, aggression and effective punching. Collected stats show it round by round.

    Deep Strike has done a better job scoring Loma Lopez, Canelo-Golovkin and Kovalev-Ward than the judges.
     
  14. UmarIFLUmar

    UmarIFLUmar Boxing Addict Full Member

    5,151
    8,076
    Jan 8, 2021
    The AI results are in and the winner by unanimous decision is.... CAMPBELL HATTOOOOOOON
     
  15. Saintpat

    Saintpat Obsessed with Boxing Full Member

    23,747
    27,399
    Jun 26, 2009
    You’re OP says ‘quantifies attributes like aggression and pressure,’ which are not part of scoring fights.

    You get results that you like better than the actual scoring so you say it’s better. Which doesn’t mean it’s better.

    GIGO.