# NEEDS TO BE RUN ON Python3

import matplotlib.pylab as P

import json

import hickle

from curvfife import CurvFiFE

from analyze_golf_poly import G, pars, Golf_GPCM_IRF, strokes

import torch as T

from tqdm import tqdm # for progress bar

P.rc('text', usetex=True)

P.rc('font', family='serif')

N = 501

# Access data as

# G["Player Name"][year][round, hole]

# Round and hole indexed from 0. 4 rounds, 18 holes.

Theta_bundles, params_hat = \

hickle.load("April_15_2019_at_1052PM/round_7.hkl")

# hickle.load("April_15_2019_at_0425PM/round_4.hkl")

# hickle.load("April_14_2019_at_1042PM/round_14.hkl")

best_names = [name for name in G.keys() if len(G[name])>1]

track = [

]

names = sorted(G.keys())

I = P.searchsorted(names, best_names)

Theta_C = { name: CurvFiFE() for name in best_names }

for i in tqdm(I, ascii=True, desc="Restoring Ability distribution functions"):

name = names[i]

bundle = Theta_bundles[i]

Theta_C[name].restore_from_bundle(bundle)

items = [

Golf_GPCM_IRF(params_hat[i], strokes[i], pars[i])

for i in range(18)

]

winners = {}

for year in range(1937, 2019, 1):

best = None

for name in G.keys():

if year in G[name]:

score = G[name][year].sum()

if (best is None) or (score < best):

best = score

winners[year] = name

P.figure(figsize=(5.5, 4))

theta = P.linspace(-6, 6, N)

t = P.linspace(1937-.5, 2018+.5, N)

i1 = 0

i2 = 0

temp = 0

for name, col in [

("Arnold Palmer", 'red'),

("Jack Nicklaus", 'orange'),

("Tiger Woods", 'green')

]:

temp += 1

years = sorted(G[name].keys())

y1, y2 = years[0], years[-1]

t = P.linspace(y1-.5, y2+.5, N)

C = Theta_C[name]

theta_hat = P.array(C.get_y_hat(t))

P.plot(t, theta_hat, linewidth=2, color=col,

label=name)

label = "Attended Masters" if temp==1 else None

P.scatter(years, P.array(C.get_y_hat(years)), color='blue',

label=label)

label = "70\% conf interval" if temp==1 else None

c_lo, c_hi = map(P.array, C.get_confidence_interval(t, 70))

P.fill_between(t, c_lo, c_hi,

facecolor='purple', alpha=0.5, label=label)

P.legend()

# P.title("Ability Trajectories of 3 legendary players")

P.xlabel("Year")

P.ylabel("Ability")

P.ylim([-3, 3])

# P.savefig('../writeup/figs/legend_trajectories.pdf')

P.show()

abils = { y : [] for y in range(1937, 2019) }

for name in tqdm(best_names, desc="Plotting trajectories", ascii=True):

years = sorted(G[name].keys())

y1, y2 = years[0], years[-1]

t = P.linspace(y1-.5, y2+.5, N)

C = Theta_C[name]

theta_hat = P.array(C.get_y_hat(t))

theta_years = P.array(C.get_y_hat(years))

for i, year in enumerate(years):

abils[year].append(theta_years[i])

if name in track:

i1 += 1

label = None if i1!=1 else "Top 15 players"

P.plot(t, theta_hat, linewidth=2, color='black', label=label,

zorder=10)

if name in winners.values():

i2 +=1

label = None if i2!=1 else "Tournament win (top 15 only)"

yy = [y for y in years if winners[y]==name]

P.scatter(yy, P.array(C.get_y_hat(yy)), marker="*",

color="green", label=label, zorder=15)

else:

P.plot(t, theta_hat, linewidth=1, color='grey',

alpha=0.5, zorder=1)

years, med_theta = P.array([

(year, P.median(abils[year]))

for year in range(1937, 2019) if len(abils[year])>0

]).T

P.plot(years, med_theta, linewidth=2,

color='red', zorder=20, label="Median ability")

P.legend()

# P.title("Trajectories for abilities of Masters Tournament attendees")

P.xlabel("Year")

P.ylabel("Ability")

P.ylim([-3, 3])

# P.savefig('../writeup/figs/overall_golf_trajectories.pdf')

P.show()

name = "Arnold Palmer"

C = Theta_C[name]

ldists = []

X = []

for y, S in G[name].items():

for r in range(4):

for h in range(18):

X.append(y)

ldists.append(items[h].get_log_func(S[r, h]))

curr_h = float(C._h)

_, LL = C.feed_data_CV(X, ldists, runs=5, k=2,

monte_samps=20000, bar=True, hh=[curr_h], force_CV=True)

C._h = curr_h

with open("new_golf_data.json") as f:

Gnew = json.load(f)

# Convert keys back to integers

for player in list(Gnew.values()):

for ys, yv in list(player.items()):

player[int(ys)] = P.array([ yv[str(i)] for i in range(1, 5) ])

del player[ys]

static_bundles, static_params = \

hickle.load("April_26_2019_at_1019PM/round_8.hkl")

C = CurvFiFE()

i = P.searchsorted(sorted(Gnew.keys()), name)

C.restore_from_bundle(static_bundles[i])

static_items = [

Golf_GPCM_IRF(static_params[i], strokes[i], pars[i])

for i in range(18)

]

ldists = []

X = []

for y, S in Gnew[name].items():

for r in range(4):

for h in range(18):

X.append(y)

ldists.append(static_items[h].get_log_func(S[r, h]))

curr_h = float(C._h)

_, LL_static = C.feed_data_CV(X, ldists, runs=5, k=2,

monte_samps=20000, bar=True, hh=[curr_h], force_CV=True)

C._h = curr_h

for _ in range(3): print()

print("Comparison of bandwidths for Palmer")

for h, ll in sorted(LL.items()):

print("h: {0:.2f},\tll: {1:.8f}".format(h, ll))

for h, ll in sorted(LL_static.items()):

print("h: {0:.2f},\tll: {1:.8f}".format(h, ll))

for h, item in enumerate(items):

if not (h+1) in [6, 13]:

continue

params = P.array(item.get_params())

s1 = len(params)//2

a, b = map(list, P.around([params[:s1], params[s1:]], 2))

i = P.searchsorted(strokes[h], pars[h])

a.insert(i, "N/A")

b.insert(i, "N/A")

p = P.array(item(theta.reshape(-1, 1), [strokes[h]]))

P.plot(theta, p, linewidth=2)

P.ylabel("Probability of getting stroke count")

P.xlabel(r"$\theta$ (ability)")

P.legend([

"{s} strokes, a={aa}, b={bb}.".format(s=strokes[h][j], aa=a[j], bb=b[j])

for j in range(len(strokes[h]))

])

P.title(

"Hole \#{h}, Par {p}"\

.format(h=h+1, p=pars[h])

)

# P.savefig('../writeup/figs/hole_{0}_irf.eps'.format(h+1))

P.show()