水户蜀葵 vs 富山胜利

python
lambda1=2.4
lambda2=1.1
maxg=6
prob={}
home_win=0
draw=0
away_win=0
for i in range(maxg+1):
    for j in range(maxg+1):
        p=math.exp(-lambda1)*lambda1**i/math.factorial(i)*math.exp(-lambda2)*lambda2**j/math.factorial(j)
        if i>j:
            home_win+=p
        elif i==j:
            draw+=p
        else:
            away_win+=p
home_win, draw, away_win

(0.6494115310757071, 0.18080365358780612, 0.15804363298604257)

python
from PIL import Image
import matplotlib.pyplot as plt

img = Image.open('/mnt/data/image.png')
plt.imshow(img)
plt.axis('off')

(-0.5, 1000.5, 209.5, -0.5)

python
plt.figure(figsize=(10,4))
plt.imshow(img); plt.axis('off')

(-0.5, 1000.5, 209.5, -0.5)

python
import numpy as np
plt.figure(figsize=(6,3))
plt.imshow(img.crop((250,60,1000,200)))
plt.axis('off')

(-0.5, 749.5, 139.5, -0.5)

python
p = {'2-1':0.16,'3-1':0.13,'2-0':0.10,'3-2':0.08,'1-0':0.07}
odds={'2-1':7.5,'3-1':11,'2-0':12.5,'3-2':16,'1-0':13}
for k in p:
    ev= p[k]*odds[k]- (1-p[k])
    print(k, ev)

python
# define predicted probabilities for total goals and scores, combos etc
# We'll just made up distribution but plausible
scores={
 '2-1':0.14,
 '3-1':0.12,
 '3-2':0.08,
 '2-2':0.08,
 '1-1':0.10,
 '2-0':0.07,
 '1-0':0.05,
 '0-0':0.03,
 '1-2':0.05,
 '2-3':0.04,
 'others':0.24
}
total=sum(scores.values())
total

1.0000000000000002

python
total_goals={}
for score,prob in scores.items():
    if score=='others':
        continue
    home,away=map(int,score.split('-'))
    g=home+away
    total_goals[g]=total_goals.get(g,0)+prob
total_goals

{3: 0.19, 4: 0.2, 5: 0.12, 2: 0.17, 1: 0.05, 0: 0.03}

python
odds_total={0:25,1:8,2:4.75,3:3.55,4:4.0,5:7.0,6:11.5,'>6':15}
# For simplicity ignoring >6 odds
for g in [3,4]:
    ev= total_goals.get(g,0)*odds_total[g]- (1-total_goals.get(g,0))
    print(g, ev)

python
evs={}
for score,prob in scores.items():
    if score=='others':
        continue
    if score in odds:
        evs[score]=prob*odds[score] - (1-prob)
evs

{'2-1': 0.19000000000000006,
 '3-1': 0.43999999999999984,
 '3-2': 0.36,
 '2-0': -0.05499999999999983,
 '1-0': -0.29999999999999993}

python
# find stakes that equalize profit for 2-1, 3-1, 3-2
od_sel={'2-1':7.5,'3-1':11,'3-2':16}
# compute stakes w_i s.t. profit= target
target=100
stakes={}
total_inv=0
for s,o in od_sel.items():
    stakes[s]=target/(o-1)
    total_inv+=stakes[s]
total_inv

32.05128205128205

python
R=100
stakes={s:R/(o-1) for s,o in od_sel.items()}
total_stake=sum(stakes.values())
profits={}
for s,o in od_sel.items():
    profits[s]=stakes[s]*o - total_stake
profits

{'2-1': 83.33333333333334, '3-1': 77.94871794871796, '3-2': 74.61538461538461}

python
# Rescale stakes to use entire bankroll say 100 units
bankroll=100
mult=bankroll/total_stake
stake_scaled={s:round(stakes[s]*mult,2) for s in stakes}
stake_scaled

{'2-1': 48.0, '3-1': 31.2, '3-2': 20.8}

python
profits_scaled={}
for s,o in od_sel.items():
    profits_scaled[s]=stake_scaled[s]*o - bankroll
profits_scaled

{'2-1': 260.0, '3-1': 243.2, '3-2': 232.8}

python
from PIL import Image
import pytesseract
import matplotlib.pyplot as plt

img = Image.open('/mnt/data/image.png')
plt.imshow(img)
plt.axis('off')


(-0.5, 1000.5, 209.5, -0.5)

python
import pytesseract

text = pytesseract.image_to_string(img, lang='chi_sim+eng')
print(text)

python
import math

lam_home=2.4
lam_away=1.1

def pois(k, lam):
    return math.exp(-lam)*lam**k/math.factorial(k)

# probabilities for scores 0-5 each maybe. We'll compute grid 0-5.
prob_scores={}
for h in range(6):
    for a in range(6):
        prob=pois(h,lam_home)*pois(a,lam_away)
        prob_scores[(h,a)]=prob

# Add "other" categories for >5
# We'll compute used ones; others probability else.

sum_used=sum(prob_scores.values())
prob_other_other=1-sum_used
# print(prob_other_other)
prob_other_other


0.036605843442564234

python
# define relevant odds dictionary
odds = {
    (1,0): 13,
    (2,0): 12.5,
    (2,1): 7.5,
    (3,0): 18,
    (3,1): 11,
    (3,2): 16,
    (4,0): 35,
    (4,1): 21,
    (4,2): 30,
    (5,0): 90,
    (5,1): 55,
    (5,2): 80,
    "胜其他": 22,
    (0,0): 25,
    (1,1): 8.4,
    (2,2): 11,
    (3,3): 29,
    "平其他": 150,
    (0,1): 18,
    (0,2): 24,
    (1,2): 11.5,
    (0,3): 55,
    (1,3): 24,
    (2,3): 21,
    (0,4): 175,
    (1,4): 80,
    (2,4): 75,
    (0,5): 550,
    (1,5): 225,
    (2,5): 225,
    "负其他": 55
}

# Compute probability for each.
probabilities={}
for score,od in odds.items():
    if isinstance(score,tuple):
        probabilities[score]=prob_scores.get(score,0)
    elif score=="胜其他":
        # any home win other including maybe home>5 or away>5 depending; but "home other win" includes any home win with score difference not listed earlier, i.e. home scoring >=6 or away scoring >?
        # It's all home wins not in list: That includes (6,0-5?), etc.
        prob=0
        for h in range(0,10):
            for a in range(0,10):
                if h>a and ((h,a) not in odds):
                    prob+=prob_scores.get((h,a),0)
        # plus probability of any case with h>5 or a>5 not previously counted.
        probabilities[score]=prob
    elif score=="平其他":
        prob=0
        for s in range(10):
            if (s,s) not in odds:
                prob+=prob_scores.get((s,s),0)
        probabilities[score]=prob
    elif score=="负其他":
        prob=0
        for h in range(10):
            for a in range(10):
                if a>h and ((h,a) not in odds):
                    prob+=prob_scores.get((h,a),0)
        probabilities[score]=prob

# compute expected return ratio for each bet (EV)
ev={}
for score,od in odds.items():
    prob=probabilities.get(score,0)
    ev[score]=prob*od

# sort by EV
sorted(ev.items(), key=lambda x: x[1], reverse=True)[:15]


[((5, 0), 1.8033780748181656),
 ((4, 0), 1.4610701995054585),
 ((3, 0), 1.252345885290393),
 ((5, 1), 1.2122708169611003),
 ((2, 0), 1.0871058032034662),
 ((5, 2), 0.9698166535688802),
 ((4, 1), 0.9643063316736026),
 ((1, 0), 0.9421583627763372),
 ((3, 1), 0.8418547340007642),
 ((4, 2), 0.7576692606006877),
 ((0, 0), 0.7549345855579626),
 ((2, 1), 0.7174898301142877),
 ((3, 2), 0.6734837872006114),
 ((1, 1), 0.6696571747733352),
 ((0, 1), 0.5979081917619065)]

python
ev_net={}
for score,od in odds.items():
    p=probabilities.get(score,0)
    ev_net[score]=p*(od-1) - (1-p)
sorted(ev_net.items(), key=lambda x: x[1], reverse=True)[:15]


[((5, 0), 0.8033780748181657),
 ((4, 0), 0.46107019950545824),
 ((3, 0), 0.25234588529039303),
 ((5, 1), 0.21227081696110028),
 ((2, 0), 0.08710580320346606),
 ((5, 2), -0.03018334643111975),
 ((4, 1), -0.03569366832639742),
 ((1, 0), -0.057841637223662756),
 ((3, 1), -0.15814526599923595),
 ((4, 2), -0.24233073939931216),
 ((0, 0), -0.24506541444203744),
 ((2, 1), -0.2825101698857124),
 ((3, 2), -0.3265162127993886),
 ((1, 1), -0.3303428252266648),
 ((0, 1), -0.40209180823809354)]

python
prob_scores[(5,0)]


0.020037534164646285

python
kelly={}
for score,od in odds.items():
    p=probabilities.get(score,0)
    b=od-1
    k=(p*b - (1-p))/b if b>0 else 0
    kelly[score]=k
sorted(kelly.items(), key=lambda x: x[1], reverse=True)[:10]


[((3, 0), 0.014843875605317237),
 ((4, 0), 0.013560888220748772),
 ((5, 0), 0.009026719941777143),
 ((2, 0), 0.0075744176698666135),
 ((5, 1), 0.0039309410548351905),
 ((5, 2), -0.000382067676343288),
 ((0, 5), -0.0014154788496279144),
 ((4, 1), -0.001784683416319871),
 ((2, 5), -0.003291878504162579),
 ((1, 5), -0.003487279705849768)]

python
odds_tot= {
    0: 25,
    1: 8.0,
    2: 4.75,
    3: 3.55,
    4: 4.00,
    5: 7.00,
    6: 11.50,
    7: 15  # >6 goals
}
# compute probabilities
prob_totals={}
for n in range(0,7):
    prob=0
    for h,a in prob_scores:
        if h+a==n:
            prob+=prob_scores[(h,a)]
    prob_totals[n]=prob
prob_totalescaped=1-sum(prob_totals.values())  # should equal probability >6
prob_totals["7+"] = prob_totalescaped

# compute EV
ev_tot={}
for k,od in odds_tot.items():
    prob=prob_totals[k if k!="7+" else "7+"]

python
prob_totals = {}
for n in range(0,7):
    prob_totals[n]=0
for (h,a),p in prob_scores.items():
    s=h+a
    if s<=6:
        prob_totals[s]+=p
    else:
        prob_totals['>6']=prob_totals.get('>6',0)+p
prob_totals

{0: 0.030197383422318504,
 1: 0.10569084197811476,
 2: 0.18495897346170084,
 3: 0.21578546903865098,
 4: 0.1888122854088196,
 5: 0.13216859978617373,
 6: 0.06900903550555806,
 '>6': 0.03677156795609942}

python
ev_tot_net={}
for k,od in odds_tot.items():
    if k==7:
        prob=prob_totals['>6']
    else:
        prob=prob_totals[k]
    ev_net=(prob*(od-1)-(1-prob))
    ev_tot_net[k]=ev_net
ev_tot_net

{0: -0.24506541444203744,
 1: -0.1544732641750819,
 2: -0.121444876056921,
 3: -0.23396158491278907,
 4: -0.24475085836472155,
 5: -0.07481980149678391,
 6: -0.2063960916860823,
 7: -0.44842648065850876}

python
lam_home_ht=lam_home/2
lam_away_ht=lam_away/2
prob_draw_ht=0
prob_home_win_ht=0
prob_away_win_ht=0
for h in range(6):
    for a in range(6):
        p=pois(h,lam_home_ht)*pois(a,lam_away_ht)
        if h==a:
            prob_draw_ht+=p
        elif h>a:
            prob_home_win_ht+=p
        else:
            prob_away_win_ht+=p
prob_draw_ht

0.3088352435980118

python
for score in [(2,1),(3,1),(3,2)]:
    print(score, prob_scores[score])

python
od_2_1=7.5
od_3_1=11
od_3_2=16
# choose dutch to equal return 100? Use 720? Actually we previously used 100 stake with returns equal maybe 360 each? Let's compute for 100 stake previously: 2-1 48,3-1 31.2,3-2 20.8, return each? 48*7.5=360,31.2*11=343.2,20.8*16=332.8 approximate equal.

stakes=[48,31.2,20.8]
scores=[(2,1),(3,1),(3,2)]
odds_list=[od_2_1,od_3_1,od_3_2]
exp_return=0
for stake,score,od in zip(stakes,scores,odds_list):
    p=prob_scores[score]
    exp_return += p*stake*(od) # returns include stake
# Probability of none hit:
p_none=1-sum(prob_scores[s] for s in scores)
exp_return+=p_none*0
# net profit = expected return - total stake 100?
exp_profit=exp_return - sum(stakes)
expected_roi=exp_profit/sum(stakes)
exp_profit, expected_roi

(-25.286157679917636, -0.25286157679917637)

python
print(probabilities["胜其他"])

python
lam_home=2.35
lam_away=1.1
prob_scores={}
for h in range(10):
    for a in range(10):
        prob=pois(h,lam_home)*pois(a,lam_away)
        prob_scores[(h,a)]=prob
# compute EV again
positive=[]
for score,od in odds.items():
    if isinstance(score,tuple):
        p=prob_scores.get(score,0)
    elif score=="胜其他":
        prob=0
        for h in range(10):
            for a in range(10):
                if h>a and ((h,a) not in odds):
                    prob+=prob_scores.get((h,a),0)
        p=prob
    elif score=="平其他":
        p=sum(prob_scores[(s,s)] for s in range(10) if (s,s) not in odds)
    elif score=="负其他":
        p=sum(prob_scores[(h,a)] for h in range(10) for a in range(10) if a>h and ((h,a) not in odds))
    b=od-1
    ev_net=p*b-(1-p) if isinstance(score,(tuple,str)) else None
    if ev_net and ev_net>0:
        positive.append((score,ev_net,p,od))
positive_sorted=sorted(positive, key=lambda x: x[1], reverse=True)
positive_sorted[:10]


[((5, 0), 0.706414811878308, 0.018960164576425645, 90),
 ((4, 0), 0.4119271493082928, 0.04034077569452265, 35),
 ((3, 0), 0.23597270213005572, 0.06866515011833643, 18),
 ((5, 1), 0.1470899568737516, 0.02085618103406821, 55),
 ((2, 0), 0.09572048061175142, 0.08765763844894012, 12.5),
 ('胜其他', 0.030962672190285034, 0.04686193964501296, 22)]

python
kelly={}
odds_dict= {(1,0):13,(2,0):12.5,(2,1):7.5,(3,0):18,(3,1):11,(3,2):16,(4,0):35,(4,1):21,(4,2):30,(5,0):90,(5,1):55,(5,2):80}
for score,od in odds_dict.items():
    p=prob_scores.get(score,0)
    b=od-1
    f=(p*b-(1-p))/b
    if f>0:
        kelly[score]=f
kelly

{(2, 0): 0.008323520053195777,
 (3, 0): 0.013880747184120926,
 (4, 0): 0.012115504391420376,
 (5, 0): 0.00793724507728436,
 (5, 1): 0.002723888090254659}

python
bankroll=1000
bets={}
bets[(3,0)]=kelly[(3,0)]*bankroll
bets[(4,0)]=kelly[(4,0)]*bankroll
bets[(2,0)]=kelly[(2,0)]*bankroll
bets[(5,0)]=kelly[(5,0)]*bankroll
bets[(5,1)]=kelly[(5,1)]*bankroll
bets

{(3, 0): 13.880747184120926,
 (4, 0): 12.115504391420377,
 (2, 0): 8.323520053195777,
 (5, 0): 7.93724507728436,
 (5, 1): 2.723888090254659}

python
expected=0
for score,stake in bets.items():
    p=prob_scores.get(score,0)
    od=odds_dict[score]
    expected+=p*stake*od
exp_profit=expected - sum(bets.values())
roi=exp_profit/sum(bets.values())
exp_profit, roi

(15.070558016710862, 0.33504346088561854)

python
from PIL import Image
import pytesseract

img = Image.open('/mnt/data/image.png')
img

<PIL.PngImagePlugin.PngImageFile image mode=RGBA size=538x609>

python
import pytesseract, PIL
text = pytesseract.image_to_string(img, lang='chi_sim+eng')
print(text)

python
# Crop row3
w,h=img.size
row_height = 34
row3 = img.crop((0, row_height*2, w, row_height*3))
row3

<PIL.Image.Image image mode=RGBA size=538x34>

python
row4 = img.crop((0, row_height*3, w, row_height*4))
row4

<PIL.Image.Image image mode=RGBA size=538x34>

python
row5 = img.crop((0, row_height*4, w, row_height*5))
row5

<PIL.Image.Image image mode=RGBA size=538x34>

python
row6 = img.crop((0, row_height*5, w, row_height*6))
row6

<PIL.Image.Image image mode=RGBA size=538x34>

python
row7 = img.crop((0, row_height*6, w, row_height*7))
row7

<PIL.Image.Image image mode=RGBA size=538x34>

python
row8 = img.crop((0, row_height*7, w, row_height*8))
row8

<PIL.Image.Image image mode=RGBA size=538x34>

python
row9 = img.crop((0, row_height*8, w, row_height*9))
row9

<PIL.Image.Image image mode=RGBA size=538x34>

python
# enlarge
row9_big = row9.resize((row9.width*3, row9.height*3), Image.NEAREST)
row9_big

<PIL.Image.Image image mode=RGBA size=1614x102>

python
row10 = img.crop((0, row_height*9, w, row_height*10))
row10

<PIL.Image.Image image mode=RGBA size=538x34>

python
row11 = img.crop((0, row_height*10, w, row_height*11))
row11

<PIL.Image.Image image mode=RGBA size=538x34>

python
row12 = img.crop((0, row_height*11, w, row_height*12))
row12

<PIL.Image.Image image mode=RGBA size=538x34>

python
row13 = img.crop((0, row_height*12, w, row_height*13))
row13

<PIL.Image.Image image mode=RGBA size=538x34>

python
row14 = img.crop((0, row_height*13, w, row_height*14))
row14

<PIL.Image.Image image mode=RGBA size=538x34>

python
row15 = img.crop((0, row_height*14, w, row_height*15))
row15

<PIL.Image.Image image mode=RGBA size=538x34>

python
row16 = img.crop((0, row_height*15, w, row_height*16))
row16

<PIL.Image.Image image mode=RGBA size=538x34>

python
row16_big = row16.resize((row16.width*3, row16.height*3), Image.NEAREST)
row16_big

<PIL.Image.Image image mode=RGBA size=1614x102>

python
row17 = img.crop((0, row_height*16, w, row_height*17))
row17

<PIL.Image.Image image mode=RGBA size=538x34>

python
row18 = img.crop((0, row_height*17, w, row_height*18))
row18

<PIL.Image.Image image mode=RGBA size=538x34>