Aintree 2021: Grand National Chase Handicapping

Throwing caution to the wind, we will attempt to find live-longshots (20/1+) to finish in the money in the 40 runner Grand National Chase (G3) at Aintree 2021.

Taking our lead once again from Shannon-Fano Crowd Handicapping, we can see that the implied probabilities of the win-market prices allow us to divide the race entrants into five separate groups or sub-races (ABCDE). For example, we could decide to handicap each sub-race separately and leave ourselves five selections for punting.

Constraint Satisfaction Handicapping

Our approach is to use the recently posted Trader Probabilities Derivation And K-L Divergence (Part 3) method, which is basically solving a constraint satisfaction problem (CSP).

Model 20210410_17-15_Aintree_Grand-National
   
     Variables
   
       ! --- Public market information ---
   
       ! Trader win probabilities.
       ! Initial values set to market price implied probabilities.
       ! Lower-bound set to 1.25% - half the natural-odds implied 
       ! probability (1/48).
       ! Upper-bound set to 100%.
   
       acapella_bourgeois     = 1/36.00, >=0.00, <=1.00
       alpha_des_obeaux       = 1/95.00, >=0.00, <=1.00
       ami_desbois            = 1/140.00, >=0.00, <=1.00
       anibale_fly            = 1/32.00, >=0.00, <=1.00
       any_second_now         = 1/14.50, >=0.00, <=1.00
       balko_des_flos         = 1/130.00, >=0.00, <=1.00
       ballyoptic             = 1/110.00, >=0.00, <=1.00
       blaklion               = 1/100.00, >=0.00, <=1.00
       bristol_de_mai         = 1/38.00, >=0.00, <=1.00
       burrows_saint          = 1/13.00, >=0.00, <=1.00
       cabaret_queen          = 1/120.00, >=0.00, <=1.00
       canelo                 = 1/60.00, >=0.00, <=1.00
       chriss_dream           = 1/80.00, >=0.00, <=1.00
       class_conti            = 1/65.00, >=0.00, <=1.00
       cloth_cap              = 1/6.40, >=0.00, <=1.00
       definitly_red          = 1/90.00, >=0.00, <=1.00
       discorama              = 1/22.00, >=0.00, <=1.00
       double_shuffle         = 1/150.00, >=0.00, <=1.00
       fagan                  = 1/130.00, >=0.00, <=1.00
       farclas                = 1/23.00, >=0.00, <=1.00
       give_me_a_copper       = 1/95.00, >=0.00, <=1.00
       hogans_height          = 1/95.00, >=0.00, <=1.00
       jett                   = 1/110.00, >=0.00, <=1.00
       kauto_riko             = 1/150.00, >=0.00, <=1.00
       kimberlite_candy       = 1/18.50, >=0.00, <=1.00
       lake_view_lad          = 1/75.00, >=0.00, <=1.00
       lord_du_mesnil         = 1/40.00, >=0.00, <=1.00
       magic_of_light          = 1/25.00, >=0.00, <=1.00
       milan_native           = 1/36.00, >=0.00, <=1.00
       minellacelebration     = 1/12.00, >=0.00, <=1.00
       minella_times          = 1/120.00, >=0.00, <=1.00
       mister_malarky         = 1/38.00, >=0.00, <=1.00
       ok_corral              = 1/100.00, >=0.00, <=1.00
       potters_corner         = 1/30.00, >=0.00, <=1.00
       secret_reprieve        = 1/20.00, >=0.00, <=1.00
       shattered_love         = 1/75.00, >=0.00, <=1.00
       some_neck              = 1/48.00, >=0.00, <=1.00
       sub_lieutenant         = 1/100.00, >=0.00, <=1.00
       takingrisks            = 1/42.00, >=0.00, <=1.00
       talkischeap            = 1/75.00, >=0.00, <=1.00
       the_long_mile          = 1/100.00, >=0.00, <=1.00
       tout_est_permis        = 1/140.00, >=0.00, <=1.00
       vieux_lion_rouge       = 1/100.00, >=0.00, <=1.00
       yala_enki              = 1/50.00, >=0.00, <=1.00
   
       kld
   
     End Variables
   
     Equations
     
       ! Total trader win probabilities must sum to one.
   
       acapella_bourgeois + &
       alpha_des_obeaux + &
       ami_desbois + &
       anibale_fly + &
       any_second_now + &
       balko_des_flos + &
       ballyoptic + &
       blaklion + &
       bristol_de_mai + &
       burrows_saint + &
       cabaret_queen + &
       canelo + &
       chriss_dream + &
       class_conti + &
       cloth_cap + &
       definitly_red + &
       discorama + &
       double_shuffle + &
       fagan + &
       farclas + &
       give_me_a_copper + &
       hogans_height + &
       jett + &
       kauto_riko + &
       kimberlite_candy + &
       lake_view_lad + &
       lord_du_mesnil + &
       magic_of_light + &
       milan_native + &
       minellacelebration + &
       minella_times + &
       mister_malarky + &
       ok_corral + &
       potters_corner + &
       secret_reprieve + &
       shattered_love + &
       some_neck + &
       sub_lieutenant + &
       takingrisks + &
       talkischeap + &
       the_long_mile + &
       tout_est_permis + &
       vieux_lion_rouge + &
       yala_enki = 1.00
       
   
       ! --- Private trader opinions ---
   
       ! Stable preference groups. 
   
       (acapella_bourgeois + burrows_saint + cabaret_queen + class_conti + &
        cloth_cap + secret_reprieve + vieux_lion_rouge) > &
       (alpha_des_obeaux + ami_desbois + anibale_fly + any_second_now + &
        balko_des_flos + ballyoptic + blaklion + bristol_de_mai + canelo + &
        chriss_dream + definitly_red + discorama + double_shuffle + &
        fagan + farclas + give_me_a_copper + hogans_height + jett + &
        kauto_riko + kimberlite_candy + lake_view_lad + lord_du_mesnil + &
        magic_of_light + milan_native + minellacelebration + &
        minella_times + mister_malarky + ok_corral + potters_corner + &
        shattered_love + some_neck + sub_lieutenant + takingrisks + &
        talkischeap + the_long_mile + tout_est_permis + yala_enki)
   
       ! Class preference groups. 
   
       (acapella_bourgeois + any_second_now + bristol_de_mai + &
        burrows_saint + chriss_dream + cloth_cap + definitly_red + &
        discorama + farclas + milan_native + minella_times + &
        mister_malarky + secret_reprieve + shattered_love) > &
       (alpha_des_obeaux + ami_desbois + anibale_fly + balko_des_flos + &
        ballyoptic + blaklion + cabaret_queen + canelo + class_conti + &
        double_shuffle + fagan + give_me_a_copper + hogans_height + &
        jett + kauto_riko + kimberlite_candy + lake_view_lad + &
        lord_du_mesnil + magic_of_light + minellacelebration + &
        ok_corral + potters_corner + some_neck + sub_lieutenant + &
        takingrisks + talkischeap + the_long_mile + tout_est_permis + &
        vieux_lion_rouge + yala_enki)
   
       ! Age preference groups. 
   
       (any_second_now + cabaret_queen + chriss_dream + class_conti + &
        cloth_cap + kimberlite_candy + talkischeap) > &
       (acapella_bourgeois + alpha_des_obeaux + ami_desbois + &
        anibale_fly + balko_des_flos + ballyoptic + blaklion + &
        bristol_de_mai + burrows_saint + canelo + definitly_red + &
        discorama + double_shuffle + fagan + farclas + give_me_a_copper + &
        hogans_height + jett + kauto_riko + lake_view_lad + &
        lord_du_mesnil + magic_of_light + milan_native + &
        minellacelebration + minella_times + mister_malarky + ok_corral + &
        potters_corner + secret_reprieve + shattered_love + some_neck + &
        sub_lieutenant + takingrisks + the_long_mile + tout_est_permis + &
        vieux_lion_rouge + yala_enki)
   
       ! Saddle-weight preference groups.
   
       (acapella_bourgeois + alpha_des_obeaux + anibale_fly + &
        any_second_now + balko_des_flos + burrows_saint + &
        kimberlite_candy + lake_view_lad + magic_of_light + &
        mister_malarky + ok_corral + talkischeap + &
        tout_est_permis) > &
       (ami_desbois + ballyoptic + blaklion + bristol_de_mai + &
       cabaret_queen + canelo + chriss_dream + class_conti + &
       cloth_cap + definitly_red + discorama + double_shuffle + fagan + &
       farclas + give_me_a_copper + hogans_height + jett + kauto_riko + &
       lord_du_mesnil + milan_native + minellacelebration + &
       minella_times + potters_corner + secret_reprieve + shattered_love + &
       some_neck + sub_lieutenant + takingrisks + the_long_mile + &
       vieux_lion_rouge + yala_enki) 


       ! --- Combination of public and private information ---
   
       ! Minimize K-LD.
   
       kld=(1/36.00)*log(((1/36.00)/acapella_bourgeois)) + &
           (1/95.00)*log(((1/95.00)/alpha_des_obeaux)) + &
           (1/140.00)*log(((1/140.00)/ami_desbois)) + &
           (1/32.00)*log(((1/32.00)/anibale_fly)) + &
           (1/14.50)*log(((1/14.50)/any_second_now)) + &
           (1/130.00)*log(((1/130.00)/balko_des_flos)) + &
           (1/110.00)*log(((1/110.00)/ballyoptic)) + &
           (1/100.00)*log(((1/100.00)/blaklion)) + &
           (1/38.00)*log(((1/38.00)/bristol_de_mai)) + &
           (1/13.00)*log(((1/13.00)/burrows_saint)) + &
           (1/120.00)*log(((1/120.00)/cabaret_queen)) + &
           (1/60.00)*log(((1/60.00)/canelo)) + &
           (1/80.00)*log(((1/80.00)/chriss_dream)) + &
           (1/65.00)*log(((1/65.00)/class_conti)) + &
           (1/6.40)*log(((1/6.40)/cloth_cap)) + &
           (1/90.00)*log(((1/90.00)/definitly_red)) + &
           (1/22.00)*log(((1/22.00)/discorama)) + &
           (1/150.00)*log(((1/150.00)/double_shuffle)) + &
           (1/130.00)*log(((1/130.00)/fagan)) + &
           (1/23.00)*log(((1/23.00)/farclas)) + &
           (1/95.00)*log(((1/95.00)/give_me_a_copper)) + &
           (1/95.00)*log(((1/95.00)/hogans_height)) + &
           (1/110.00)*log(((1/110.00)/jett)) + &
           (1/150.00)*log(((1/150.00)/kauto_riko)) + &
           (1/18.50)*log(((1/18.50)/kimberlite_candy)) + &
           (1/75.00)*log(((1/75.00)/lake_view_lad)) + &
           (1/40.00)*log(((1/40.00)/lord_du_mesnil)) + &
           (1/25.00)*log(((1/25.00)/magic_of_light)) + &
           (1/36.00)*log(((1/36.00)/milan_native)) + &
           (1/12.00)*log(((1/12.00)/minellacelebration)) + &
           (1/120.00)*log(((1/120.00)/minella_times)) + &
           (1/38.00)*log(((1/38.00)/mister_malarky)) + &
           (1/100.00)*log(((1/100.00)/ok_corral)) + &
           (1/30.00)*log(((1/30.00)/potters_corner)) + &
           (1/20.00)*log(((1/20.00)/secret_reprieve)) + &
           (1/75.00)*log(((1/75.00)/shattered_love)) + &
           (1/48.00)*log(((1/48.00)/some_neck)) + &
           (1/100.00)*log(((1/100.00)/sub_lieutenant)) + &
           (1/42.00)*log(((1/42.00)/takingrisks)) + &
           (1/75.00)*log(((1/75.00)/talkischeap)) + &
           (1/100.00)*log(((1/100.00)/the_long_mile)) + &
           (1/140.00)*log(((1/140.00)/tout_est_permis)) + &
           (1/100.00)*log(((1/100.00)/vieux_lion_rouge)) + &
           (1/50.00)*log(((1/50.00)/yala_enki))
   
     End Equations
   
   End Model
   

{
     "acapella_bourgeois" : [1.0344483718E-01],
     "alpha_des_obeaux"   : [1.1092490127E-02],
     "ami_desbois"        : [6.4013255253E-03],
     "anibale_fly"        : [1.1092489557E-02],
     "any_second_now"     : [1.1905042800E-01],
     "balko_des_flos"     : [1.1092490142E-02],
     "ballyoptic"         : [6.4013255253E-03],
     "blaklion"           : [6.4013255253E-03],
     "bristol_de_mai"     : [6.9307674653E-03],
     "burrows_saint"      : [1.0344530933E-01],
     "cabaret_queen"      : [5.6574957561E-02],
     "canelo"             : [6.4013255292E-03],
     "chriss_dream"       : [1.3449984823E-02],
     "class_conti"        : [5.6574614025E-02],
     "cloth_cap"          : [1.6988521027E-01],
     "definitly_red"      : [6.9307668564E-03],
     "discorama"          : [6.9307681009E-03],
     "double_shuffle"     : [6.4013255253E-03],
     "fagan"              : [6.4013255253E-03],
     "farclas"            : [6.9307680439E-03],
     "give_me_a_copper"   : [6.4013255256E-03],
     "hogans_height"      : [6.4013255256E-03],
     "jett"               : [6.4013255253E-03],
     "kauto_riko"         : [6.4013255253E-03],
     "kimberlite_candy"   : [4.9381265450E-02],
     "lake_view_lad"      : [1.1092490049E-02],
     "lord_du_mesnil"     : [6.4013255341E-03],
     "magic_of_light"     : [1.1092489303E-02],
     "milan_native"       : [6.9307675202E-03],
     "minellacelebration" : [6.4013255548E-03],
     "minella_times"      : [6.9307668115E-03],
     "mister_malarky"     : [1.2784649318E-02],
     "ok_corral"          : [1.1092490142E-02],
     "potters_corner"     : [6.4013255385E-03],
     "secret_reprieve"    : [1.3221176794E-02],
     "shattered_love"     : [6.9307669456E-03],
     "some_neck"          : [6.4013255317E-03],
     "sub_lieutenant"     : [6.4013255253E-03],
     "takingrisks"        : [6.4013255334E-03],
     "talkischeap"        : [4.9381264855E-02],
     "the_long_mile"      : [6.4013255253E-03],
     "tout_est_permis"    : [1.1092490143E-02],
     "vieux_lion_rouge"   : [1.1419641645E-02],
     "yala_enki"          : [6.4013255312E-03],
   }
   

On this occasion, there are four standout horses (apart from the top three in the betting), which have an edge on the market and are worth each-way consideration:

      a. acapella_bourgeois:  36.00
      b. cabaret_queen:      120.00
      c. class_conti:         65.00
      d. talkischeap:         75.00
   

We have completed our task without having to generate 40 win probabilities but simply by identifying preference groups within categories that we ourselves have chosen!

One notable exception to our process is secret_reprieve: 20.00 who is the least-exposed runner in the race but with both high class and fpr ratings - great potential but will need a lot of luck in running to be competitivve!

Note: We are not saying that those horses we have eliminated are not going to win - simply that they did not meet our criteria for live longshots to run in the money. Also, whether we succeed or fail in a single race is not the measure of the two processes we present here but their long-term peformance against the market!

Cheltenham 2021: Supreme-Novices Hurdle Handicapping

Suffice to say that this year's Supreme Novices Championship Hurdle has been gutted with 10 defections at the last stage of declaration. That said, the following analysis was completed before this happened...Remember that handicapping is all about getting the process right. After that, the results will naturally follow. Enjoy!

'Once more unto the breach, dear friends...'

It is time once again for our annual attempt to find live-longshots (10/1+) to finish in the money in the Supreme Novices Hurdle (G1) at Cheltenham 2021.

Taking our lead from Shannon-Fano Crowd Handicapping, we can see that the implied probabilities of the win-market prices allow us to divide the race entrants into five separate groups (ABCDE). For example, the first group (Appreciate It and Metier) accounts for approximately 50% of the market. The question we then ask ourselves is whether the winner will come from this group or the remainder of the field? Similarly with the second group - comparing it to the remainder of the field below it. As a result, we have effectively reduced the race from one comprising 18 runners initially to four sub-races (AvBCDE, BvCDE. CvDE, and DvE), which is a much simpler cognitive task to undertake. However, for this particular exercise, we will effectively be ignoring groups A and B.

This year we will apply two distinct approaches to our analysis:

a. Pencil-and-Paper (Qualitative) and
b. Constraint Satisfaction (Quantitative).

Pencil-And-Paper Handicapping

This is our traditional approach for the last few years, where we use a process of elimination for identifying contenders.

a. Short Starting Price [Appreciate It, Ballyadam, Blue Lord, Metier, Soaring Glory].
b. Connections lack confidence [Galopin Des Champs, Ganapathi, Guard Your Dreams, M C Muldoon, Shakem Up'Arry].
c. Poor FPR [For Pleasure, Galopin Des Champs, Grumpy Charley, Shakem Up'Arry].
d. Pedigree mismatch to former winners [Blue Lord, Ganapathi, Grumpy Charley].
e. Poor "Late-Speed" [For Pleasure, M C Muldoon, Shakem Up'Arry].
f. Poor Cheltenham form [Shakem Up'Arry].
g. Never run on L-H track [Ganapathi].
h. Over-exposed form [For Pleasure].
i. Weak "Strength-Of-Schedule" [Guard Your Dreams].
j. Convincingly beaten by current favorite [Irascible]

We are left with the following five horses, which are worth each-way consideration:

a. Bob Olinger         17.00
b. Keskonrisk          32.00
c. Gowel Road          44.00
d. Fifty Ball          70.00
e. Third Time Lucki    85.00

Constraint Satisfaction Handicapping

Our alternate approach is to use the recently posted Trader Probabilities Derivation And K-L Divergence (Part 3) method, which is basically solving a constraint satisfaction problem (CSP).

Model Cheltenham_Supreme-Novices-Hurdle

  Variables

    ! --- Public market information ---

    ! Trader win probabilities.
    ! Initial values set to market price implied probabilities.
    ! Lower-bound set to 0%.
    ! Upper-bound set to 100%.
    appreciate_it      = 1/2.60, >=0.00, <=1.00
    ballyadam          = 1/8.00, >=0.00, <=1.00
    blue_lord          = 1/12.00, >=0.00, <=1.00
    bob_olinger        = 1/17.00, >=0.00, <=1.00
    fifty_ball         = 1/70.00, >=0.00, <=1.00
    for_pleasure       = 1/65.00, >=0.00, <=1.00
    galopin_des_champs = 1/200.00, >=0.00, <=1.00
    ganapathi          = 1/42.00, >=0.00, <=1.00
    gowel_road         = 1/44.00, >=0.00, <=1.00
    grumpy_charley     = 1/75.00, >=0.00, <=1.00
    guard_your_dreams  = 1/130.00, >=0.00, <=1.00
    irascible          = 1/40.00, >=0.00, <=1.00
    keskonrisk         = 1/32.00, >=0.00, <=1.00
    m_c_muldoon        = 1/90.00, >=0.00, <=1.00
    metier             = 1/6.00, >=0.00, <=1.00
    shakem_uparry      = 1/410.00, >=0.00, <=1.00
    soaring_glory      = 1/10.00, >=0.00, <=1.00
    third_time_lucki   = 1/85.00, >=0.00, <=1.00

    kld

  End Variables

  Equations
  
    ! Total trader win probabilities must sum to one.
    appreciate_it + &
    ballyadam + &
    blue_lord + &
    bob_olinger + &
    fifty_ball + &
    for_pleasure + &
    galopin_des_champs + &
    ganapathi + &
    gowel_road + &
    grumpy_charley + &
    guard_your_dreams + &
    irascible + &
    keskonrisk + &
    m_c_muldoon + &
    metier + &
    shakem_uparry + &
    soaring_glory + &
    third_time_lucki = 1.00
    

    ! --- Private trader opinions ---

    ! Ratings preference groups.
    (appreciate_it + ballyadam + soaring_glory + irascible + blue_lord + &
     keskonrisk + fifty_ball) > &
    (metier + guard_your_dreams + galopin_des_champs + ganapathi + &
     gowel_road + shakem_uparry + bob_olinger + third_time_lucki + &
     grumpy_charley + m_c_muldoon + for_pleasure)

    ! Runs preference groups. 
    (m_c_muldoon + galopin_des_champs + irascible + keskonrisk + metier + &
     blue_lord + ganapathi + bob_olinger + gowel_road + fifty_ball) > &
    (ballyadam + guard_your_dreams + appreciate_it + soaring_glory + &
     grumpy_charley + shakem_uparry + third_time_lucki + for_pleasure)

    ! Pedigrees preference groups. 
    (bob_olinger + gowel_road + shakem_uparry + appreciate_it) > &
    (third_time_lucki + galopin_des_champs + irascible + fifty_ball + &
     ballyadam + guard_your_dreams + soaring_glory + keskonrisk + &
     metier + m_c_muldoon + for_pleasure + blue_lord + grumpy_charley + &
     ganapathi)


    ! --- Combination of public and private information ---

    ! Minimize K-LD.
    kld=(1/2.60)*log(((1/2.60)/appreciate_it)) + &
        (1/8.00)*log(((1/8.00)/ballyadam)) + &
        (1/12.00)*log(((1/12.00)/blue_lord)) + &
        (1/17.00)*log(((1/17.00)/bob_olinger)) + &
        (1/70.00)*log(((1/70.00)/fifty_ball)) + &
        (1/65.00)*log(((1/65.00)/for_pleasure)) + &
        (1/200.00)*log(((1/200.00)/galopin_des_champs)) + &
        (1/42.00)*log(((1/42.00)/ganapathi)) + &
        (1/44.00)*log(((1/44.00)/gowel_road)) + &
        (1/75.00)*log(((1/75.00)/grumpy_charley)) + &
        (1/130.00)*log(((1/130.00)/guard_your_dreams)) + &
        (1/40.00)*log(((1/40.00)/irascible)) + &
        (1/32.00)*log(((1/32.00)/keskonrisk)) + &
        (1/90.00)*log(((1/90.00)/m_c_muldoon)) + &
        (1/6.00)*log(((1/6.00)/metier)) + &
        (1/410.00)*log(((1/410.00)/shakem_uparry)) + &
        (1/10.00)*log(((1/10.00)/soaring_glory)) + &
        (1/85.00)*log(((1/85.00)/third_time_lucki))

  End Equations

End Model

{
  "appreciate_it"     :[2.7185684754E-01],
  "ballyadam"         :[3.0378824642E-02],
  "blue_lord"         :[6.4188212946E-02],
  "bob_olinger"       :[1.1189333156E-01],
  "fifty_ball"        :[3.7051562905E-02],
  "for_pleasure"      :[1.7925815422E-02],
  "galopin_des_champs":[2.6029458023E-02],
  "ganapathi"         :[2.6029458036E-02],
  "gowel_road"        :[1.1192387537E-01],
  "grumpy_charley"    :[1.7925811477E-02],
  "guard_your_dreams" :[1.7925805273E-02],
  "irascible"         :[6.4188228012E-02],
  "keskonrisk"        :[6.4188226406E-02],
  "m_c_muldoon"       :[2.6029458048E-02],
  "metier"            :[2.6029460057E-02],
  "shakem_uparry"     :[3.8130987929E-02],
  "soaring_glory"     :[3.0378827903E-02],
  "third_time_lucki"  :[1.7925808455E-02],
}

This time, there are two standout horses (apart from the favorite), which are worth each-way consideration:

a. Bob Olinger   17.00
c. Gowel Road    44.00

As a sanity check, we should expect some degree of overlap in terms of the selections generated by the two methods. Otherwise, we are being incoherent in our overall process.

One notable exception to our process is For Pleasure - (we got lucky -paid 9.50 to show!) who is a course and distance winner albeit at a lower grade and with a projected starting price of 50/1+ is worth an each-way saver!

Note: Given the limited exposure of all the runners, we are not saying that those horses we have eliminated are not going to win - simply that they did not meet our criteria for live longshots to run in the money. Also, whether we succeed or fail in a single race is not the measure of the two processes we present here but their long-term peformance against the market!