In [1]:
from pathlib import Path
import altair as alt
import pandas as pd
import seaborn as sns
sns.set_style('whitegrid')
In [2]:
DATA_DIR = Path.cwd() / 'data'
# http://mkweb.bcgsc.ca/colorblind/palettes.mhtml
DOMAIN = ['Fully Maccinated', 'Federal Way Yu', 'Draftin Herbert First Pick', 'Spenny Willy',
'Too Many Cooks', "Rippin' Swigs", 'Herb Your Enthusiasm', 'Elite starts with Eli',
'Seattle Fantasy Team', 'Team barrett', 'Scooty Lewis and the News', 'Raiders Are Good']
RANGE = ['#9F0162', '#009F81', '#FF5AAF', '#00FCCF', '#8400CD', '#008DF9',
'#00C2F9', '#FFB2FD', '#A40122', '#E20134', '#FF6E3A', '#FFC33B']
In [3]:
drafts = pd.read_csv(DATA_DIR / 'draft_data_2021.csv').sort_values('bid_amount', ascending=False)
espn = pd.read_csv(DATA_DIR / 'espn_draft_trends_2021.csv') # draft trends from ESPN
df = pd.merge(drafts, espn, on='player_id')
assert len(df) == len(drafts)
leftover draft money¶
In [4]:
bids = df.groupby('team_name')['bid_amount'].sum().reset_index()
bids['remaining'] = 200 - bids['bid_amount']
bids['pct_remaining'] = bids['remaining'] / 200
bids = bids.sort_values('remaining', ascending=False).drop('bid_amount', axis=1)
bids.style.format({'remaining': '${:.0f}', 'pct_remaining': '{:.0%}'})
Out[4]:
| team_name | remaining | pct_remaining | |
|---|---|---|---|
| 11 | Too Many Cooks | $14 | 7% |
| 5 | Raiders Are Good | $10 | 5% |
| 6 | Rippin' Swigs | $8 | 4% |
| 2 | Federal Way Yu | $3 | 2% |
| 4 | Herb Your Enthusiasm | $3 | 2% |
| 0 | Draftin Herbert First Pick | $2 | 1% |
| 3 | Fully Maccinated | $1 | 0% |
| 7 | Scooty Lewis and the News | $1 | 0% |
| 1 | Elite starts with Eli | $0 | 0% |
| 8 | Seattle Fantasy Team | $0 | 0% |
| 9 | Spenny Willy | $0 | 0% |
| 10 | Team barrett | $0 | 0% |
In [5]:
df['team_name'].unique()
Out[5]:
array(['Fully Maccinated', 'Federal Way Yu',
'Draftin Herbert First Pick', 'Spenny Willy', 'Too Many Cooks',
"Rippin' Swigs", 'Herb Your Enthusiasm', 'Elite starts with Eli',
'Seattle Fantasy Team', 'Team barrett',
'Scooty Lewis and the News', 'Raiders Are Good'], dtype=object)
breakdown of nominating teams winning their bid?¶
In [6]:
(pd.DataFrame(((df['team_name'] == df['nominating_team_name']).value_counts() / len(drafts)))
.rename({0: 'nominating_team_won_bid'}, axis=1)
.style.format('{:.0%}'))
Out[6]:
| nominating_team_won_bid | |
|---|---|
| False | 77% |
| True | 23% |
League spending versus ESPN data¶
ppr_auc_value - ESPN recommended value for PPR league
In [7]:
cols = ['team_name', 'nominating_team_name', 'player_name', 'pos', 'ppr_auc_value', 'avg_auc_value',
'bid_amount', 'avg_draft_pos', 'overall_pick']
prices = df[cols].copy()
# filter out $1 and $2 bids
# prices = prices[prices['bid_amount'] > 2]
# actual bid price minus ESPN recommended price
prices['actual_vs_rec_diff'] = prices['bid_amount'] - prices['ppr_auc_value']
# actual bid price minus average draft prices
prices['actual_vs_avg_diff'] = prices['bid_amount'] - prices['avg_auc_value']
Summed differences by team and position¶
actual_vs_rec_diff - Actual price minus ESPN recommendation
In [8]:
rec_pivot = pd.pivot_table(prices, index='team_name', columns='pos', values='actual_vs_rec_diff', aggfunc='sum').fillna(0).astype(int)
sns.heatmap(rec_pivot, annot=True);
actual_vs_avg_diff - actual bid minus avg price
In [9]:
avg_pivot = pd.pivot_table(prices, index='team_name', columns='pos', values='actual_vs_avg_diff', aggfunc='sum').fillna(0).astype(int)
sns.heatmap(avg_pivot, annot=True);
Per player breakdown¶
The line is a linear regression across all drafted players
In [10]:
chart = (alt.Chart(prices)
.mark_point()
.encode(x='ppr_auc_value', y='bid_amount', color=alt.Color('pos:N'), tooltip=['pos', 'player_name', 'team_name', 'nominating_team_name',
'bid_amount', 'ppr_auc_value', 'actual_vs_rec_diff'])
.properties(width=400, height=400, title='Bid amount vs ESPN recommended PPR value'))
(chart + chart.transform_regression('ppr_auc_value', 'bid_amount').mark_line()).interactive()
Out[10]:
here we trended closer to the average
In [11]:
chart = (alt.Chart(prices)
.mark_point()
.encode(x='avg_auc_value', y='bid_amount', color='pos:N', tooltip=['pos', 'player_name', 'team_name', 'nominating_team_name',
'bid_amount', 'avg_auc_value', 'actual_vs_avg_diff'])
.properties(width=400, height=400, title='Bid amount vs ESPN average PPR bid'))
(chart + chart.transform_regression('avg_auc_value', 'bid_amount').mark_line()).interactive()
Out[11]:
Diff between bid amount and ESPN recommendation by team and player¶
Only showing diffs more than $2
In [12]:
(alt.Chart(prices)
.mark_bar()
.encode(x='actual_vs_rec_diff:Q', y=alt.Y('player_name:N', sort='-x'),
color=alt.Color('team_name', scale=alt.Scale(domain=DOMAIN, range=RANGE)),
tooltip=['team_name', 'pos', 'player_name', 'bid_amount', 'ppr_auc_value', 'actual_vs_rec_diff'])
.properties(height=600, title='Price diff with ESPN PPR projections')
.transform_filter(
(abs(alt.datum.actual_vs_rec_diff) > 2)
))
Out[12]:
Diff between bid amount and ESPN averages by team and player¶
In [13]:
(alt.Chart(prices)
.mark_bar()
.encode(x='actual_vs_avg_diff:Q', y=alt.Y('player_name:N', sort='-x'),
color=alt.Color('team_name', scale=alt.Scale(domain=DOMAIN, range=RANGE)),
tooltip=['team_name', 'pos', 'player_name', 'bid_amount', 'avg_auc_value', 'actual_vs_avg_diff'])
.properties(height=600, title='Price diff with avg ESPN PPR bid amounts')
.transform_filter(
(abs(alt.datum.actual_vs_rec_diff) > 2)
))
Out[13]: