Upgrade to Pro — share decks privately, control downloads, hide ads and more …

坂本勇人選手はいつ通算3,000安打を達成するか? AIに聞いてみました / Hayato Sakamoto Performance Prediction Using Feature Engineering with Machine Learning and Python

坂本勇人選手はいつ通算3,000安打を達成するか? AIに聞いてみました / Hayato Sakamoto Performance Prediction Using Feature Engineering with Machine Learning and Python

Sports Analytics Meetup #9 2020/12/13 LT

#Baseball #SABRmetrics #ML #Python

2c0947c6a28e7f771ebd9859ecf54e5c?s=128

Shinichi Nakagawa

December 13, 2020
Tweet

Transcript

  1. ӫޫͷഎ൪߸6⃣ ࡔຊ༐ਓ3,000ຊ҆ଧه೦LT Shinichi Nakagawa(@shinyorke) Sports Analyst Meetup #9 2020/12/13

  2. ʁʁʁʮ༐ਓ͸·ͩ2,000ຊ҆ଧ΍Ζʯ

  3. ͦͷͱ͓ΓͰ͍͟͝·͢, ࣦྱ͠·ͨ͠

  4. ࡔຊ༐ਓ͸͍ͭ௨ࢉ3,000ຊ҆ଧΛ ୡ੒͢Δ͔AIʹฉ͍ͯΈ·ͨ͠ Shinichi Nakagawa(@shinyorke) Sports Analyst Meetup #9 2020/12/13

  5. ຊ೔ͷςʔϚ • ࡔຊ༐ਓ͕͍ͭ͝Ζ௨ࢉ3,000ຊ҆ଧΛୡ੒͢Δ͔༧૝͢Δ • ਅ໘໨ͳ࿩, ੒੷༧ଌ͸Ͳ͜·ͰͰ͖Δ͔ࢼͯ͠ΈΔ • ʮӫޫͷഎ൪߸6⃣ࡔຊ༐ਓ3,000ຊ҆ଧ΁ͷಓʯ͕ 
 Կ೥ޙʹ์ө͞ΕΔ͔Θ͔Δ΄͏͕͍͍ΑͶʢద౰ʣ

  6. Who am I ?ʢ͓લ୭Αʣ • Shinichi Nakagawaʢத઒ ৳Ұʣ • େ఍ͷSNSͰʮshinyorkeʢ͠ΜΑʔ͘ʣʯͱ໊৐͍ͬͯ·͢

    • JX Press Corporation Senior Engineer 
 ʢJX௨৴ࣾ γχΞɾΤϯδχΞʣ • Baseball Engineer, Data Scientist 
 ʢ໺ੜͷ໺ٿΤϯδχΞɾσʔλαΠΤϯςΟετʣ • Ҏલ͸͓࢓ࣄͰ໺ٿΤϯδχΞʮͩͬͨʯਓ
  7. ʲCMʳαʔόʔαΠυΠϯλʔϯืूͯ͠·͢ https://www.wantedly.com/projects/543767 ※ֶੜ͞ΜݶఆͰ͢&ผʹεϙʔπͷ࢓ࣄͬͯ༁͡Όͳ͍Ͱ͢

  8. 26.4ඵͰৼΓฦΔ2020೥ͷϓϩ໺ٿ • ιϑτόϯΫϗʔΫε೔ຊҰʢ4࿈೼ʣ • όϯςϦϯυʔϜφΰϠ&౦ژυʔϜͷ਌ձࣾ(ry • ࡔຊ༐ਓʢڊਓʣ, ӈଧऀͱͯ͠࠷೥গͰ2,000ຊ҆ଧୡ੒ ͦͷଞʹ΋͍ͬͺ͍͋Δ͚ͲׂѪʢద౰ʣ

  9. ࡔຊ༐ਓબखͳΒ3,000ຊ҆ଧ༨༟Ͱ͸ • 31ࡀ10ϲ݄Ͱͷୡ੒͸ӈଧऀ࠷଎ • গͳ͘ͱ΋͋ͱ4, 5೥͸ݱ໾͢ΔͰ͠ΐ 
 γϣʔτͰݩؾʹΠέͯ·͢͠. • ͡Ό͍͋ͭࠒ3,000ຊ҆ଧ΍Δͷ͞?

    ͜Εͬͯաڈͷσʔλ͔Β͏·͍۩߹ʹ΍Ε͹༧ଌՄೳͰ͸? https://www.nikkansports.com/baseball/news/202011080000831.html
  10. ͱ͍͏Θ͚Ͱ༧ଌϞσϧΛ࡞Γ·ͨ͠. ࠓճ͸PyCon JP 2020Ͱ΍ͬͨ΍ͭΛݩʹͪΐͬͱΞϨϯδͯ͠࡞Γ·ͨ͠. https://shinyorke.hatenablog.com/entry/baseball-and-ml-with-python

  11. ࠓճͷΞϓϩʔνʢΊͬͪΌཁ໿ʣ • ϝδϟʔϦʔάͷ੒੷σʔλΛ࢖ͬͯ 1.࠷ۙ๣୳ࡧܥͷΞϧΰϦζϜͰ͍ۙબख୳͠ 2.֬཰෼෍ʢͬΆ͍ʣํ๏Ͱ༧ଌ੒੷Λ࡞Δ • ↑ͷ݁ՌΛStreamlitͰՄࢹԽ

  12. ͳͥϝδϟʔͷσʔλͳͷ͔ • 3,000ຊ҆ଧୡ੒ऀ, ೔ຊϓϩ໺ٿ͸Ұਓ͔͍͠ͳ͍ʢ࡯͠ʣ 
 ※ʮ୭Ͱ͔͢ʁʯ࣭ͬͯ໰͸׃΍ͧ • ϝδϟʔ͸େਖ਼ٛΠνϩʔ༷ଞ, 3,000ຊ҆ଧୡ੒ऀ͕ଟ͍. •

    σʔλͷϥΠηϯε໰୊&εΫϨΠϐϯάͱ͔͸େมͰ͠ΐ.
  13. ࡔຊ༐ਓʹ͍ۙϝδϟʔϦʔΨʔ ࢲʢshinyorkeʣ࡞੒, ʮzobristʯϞσϧͰग़ͨ݁͠Ռʢ΄΅ANNͰ͢ʣ ϝδϟʔϦʔάΛ஌ͬͯΔਓ͔ΒΈΔͱೲಘͷ݁ՌͩͱࢥΘΕ ໊͓લνʔϜ ੒੷ʢ௨ࢉʣ ଧ཰ຊྥଧ௨ࢉ҆ଧ ಛ௃ͱ͔ 9BOEFS#PHBFSUT 


    ʢ3FE4PYʣ  ଧ੮͸ӈଧ ௨ࢉ014 ݱ໾۶ࢦͷ௒߈ܸܕγϣʔτ %FSFL+FUFS ʢ:BOLFFTʣ  ଧ੮͸ӈଧ આ໌ෆཁͷελʔ ಺໺खʹݶΔͱ௨ࢉ҆ଧҐ 5SPZ5VMPXJU[LJ ʢ3PDLJFT FUDʜʣ  ଧ੮͸ӈଧ ௨ࢉ014 ௒߈ܸܕγϣʔτ ͳ͓ຊڌ஍ +JNNZ3PMMJOT ʢ1IJMMJFT FUDʜʣ  ଧ੮͸྆ଧ कඋܕͳγϣʔτ ࣮ಇ೥ͷແࣄ೭໊അ
  14. σϞ͠·͢

  15. ࡔຊ༐ਓͷࠓޙ - ҆ଧɾຊྥଧɾଧ఺ ࣅ͍ͯΔϝδϟʔϦʔΨʔXਓͷ੒੷Λ75%λΠϧͰࢉग़

  16. ࡔຊ༐ਓͷࠓޙ - ଧ཰ ࣅ͍ͯΔϝδϟʔϦʔΨʔXਓͷ੒੷Λ75%λΠϧͰࢉग़

  17. ࡔຊ༐ਓͷࠓޙΛ·ͱΊΔͱ 2027೥ʢ38ࡀʣ·Ͱنఆଧ਺֬อͰ͖ΔͬΆ͍. 
 ※نఆଧ਺͸443ଧ੮ʢ2019೥ͷࢼ߹਺143×3.1Ͱܭࢉ, ࢛ࣺޒೖʣ ೥౓ ೥ྸ ଧ਺ ҆ଧ ຊྥଧ

    ଧ఺ ଧ཰                                                        
  18. ࡔຊ༐ਓબख, ௨ࢉ੒੷ʢ༧ଌ஋ʣ ͜ΕͰγϣʔτͩͬͨΒڧ͗͢Ͱ͸ʢ਒͑ʣ ظؒ ଧ਺ ҆ଧ ຊྥଧ ଧ఺ ଧ཰ ೥·Ͱ

    ˞ݱ࣮       ˞༧ଌ੒੷      ௨ࢉʢ༧ଌʣ     
  19. ߟ࡯ • 39ࡀ͝Ζʹ3,000ຊ҆ଧୡ੒…ͷϖʔε͸·͋·͋͋Γͦ͏. 
 ͨͩ͠ྼԽ͸ආ͚ΒΕͳ͍. • 36ࡀ͔ΒͷٸܹྼԽ͸कඋҐஔมߋͱ͔ͰઌԆ͹͠Ͱ͖ͦ͏. 
 ʲࢀߟʳѨ෦৻೭ॿ͸36ࡀ͔Βัख->ϑΝʔετʹίϯόʔτ •

    ௨ࢉຊྥଧʢ༧ଌʣ321ຊ…334ຊ͸ߦͬͯཉ͍͚͠ͲͲ͏͔
  20. ͜ͷ࿩͸͓͠·͍Ͱ͢…͕ʂʁ • ༧ଌϞσϧ࡞Γ͸ηΠόʔϝτϦΫεແ͠Ͱ͸Ͱ͖ͳ͔ͬͨ • ʮRʹΑΔηΠόʔϝτϦΫεೖ໳ʯ΋ग़ͨ͠, 
 ͜ͷลΛಛ௃ྔΤϯδχΞϦϯάతʹৼΓฦΓ͍ͨ • ͍ͬͯ͏ϩϯάτʔΫ͕Ͱ͖ͨΒ͍͍ͳ⚾ 


    ʢҙ༁ɿࠓճ୺ંͬͨϞσϧͷ࿩Λ͍ͨ͠ʣ ӡӦͷօ༷, ͝ݕ౼ΑΖ͓͘͠ئ͍͠·͢
  21. ήʔϜηοτ⚾ ͝ਗ਼ௌ͋Γ͕ͱ͏͍͟͝·ͨ͠. Shinichi Nakagawa(Twitter/Facebook/etc… @shinyorke)