Simulating today’s hockey games

tutorials

Run monte carlo simulations using the expected goals model to predict the winners of today’s games

Author

chicken

Published

January 27, 2026

Intro

Use the chickenstats library to scrape play-by-play data, aggregate to the team level, then run monte carlo simulations to predict the winners of today’s games.

Parts of this tutorial are optional and will be clearly marked as such. For help, or any questions, please don’t hesitate to reach out to chicken@chickenandstats.com or @chickenandstats.com on Blue Sky.

Housekeeping

Import dependencies

Import the dependencies we’ll need for the guide

import matplotlib.patheffects as mpe
import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
from matplotlib.lines import Line2D
import seaborn as sns

import polars as pl
import numpy as np
from scipy.stats import poisson

from pathlib import Path

import concurrent
from concurrent.futures import ThreadPoolExecutor

from chickenstats.chicken_nhl import Scraper, Season
from chickenstats.chicken_nhl._helpers import charts_directory
from chickenstats.chicken_nhl._validation import TeamSchemaPolars
from chickenstats.chicken_nhl.team import NHL_COLORS, Team
from chickenstats.utilities import ChickenProgress

Polars options

Set different polars options. This cell is optional

pl.Config.set_tbl_cols(-1)

polars.config.Config

Folder structure

charts_directory()

Chickenstats matplotlib style

chickenstats.utilities includes a custom style package - this activates it. This cell is also optional

plt.style.use("chickenstats")

Scrape data

Schedule

Scrape the schedule using the Season object

season = Season(2025, backend="polars")

schedule = season.schedule()

Standings and team names

Scrape the standings and create team name dictionaries to use later

standings = season.standings

standings.head(5)

shape: (5, 57)

season	date	team	team_name	conference	division	games_played	points	points_pct	wins	regulation_wins	shootout_wins	losses	ot_losses	shootout_losses	ties	win_pct	regulation_win_pct	streak_code	streak_count	goals_for	goals_against	goals_for_pct	goal_differential	goal_differential_pct	home_games_played	home_points	home_goals_for	home_goals_against	home_goal_differential	home_wins	home_losses	home_ot_losses	home_ties	home_regulation_wins	road_games_played	road_points	road_goals_for	road_goals_against	road_goal_differential	road_wins	road_losses	road_ot_losses	road_ties	road_regulation_wins	l10_points	l10_goals_for	l10_goals_against	l10_goal_differential	l10_wins	l10_losses	l10_ot_losses	l10_ties	l10_regulation_wins	team_logo	wildcard_sequence	waivers_sequence
i64	str	str	str	str	str	i64	i64	f64	i64	i64	i64	i64	i64	i64	i64	f64	f64	str	i64	i64	i64	f64	i64	f64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	str	i64	i64
20252026	"2026-01-25"	"COL"	"Colorado Avalanche"	"Western"	"Central"	50	79	0.79	35	32	1	6	9	null	0	0.7	0.64	"W"	1	198	122	3.96	76	1.52	26	44	113	62	51	20	2	4	0	20	24	35	85	60	25	15	4	5	0	12	10	34	31	3	4	4	2	0	4	"https://assets.nhle.com/logos/…	0	32
20252026	"2026-01-25"	"CAR"	"Carolina Hurricanes"	"Eastern"	"Metropolitan"	52	69	0.663462	32	23	4	15	5	null	0	0.615385	0.442308	"W"	1	180	150	3.461538	30	0.576923	28	38	106	85	21	18	8	2	0	13	24	31	74	65	9	14	7	3	0	10	16	39	22	17	7	1	2	0	7	"https://assets.nhle.com/logos/…	0	30
20252026	"2026-01-25"	"DET"	"Detroit Red Wings"	"Eastern"	"Atlantic"	53	69	0.650943	32	22	2	16	5	null	0	0.603774	0.415094	"W"	1	167	159	3.150943	8	0.150943	27	37	88	86	2	18	8	1	0	11	26	32	79	73	6	14	8	4	0	11	17	36	21	15	8	1	1	0	5	"https://assets.nhle.com/logos/…	0	29
20252026	"2026-01-25"	"TBL"	"Tampa Bay Lightning"	"Eastern"	"Atlantic"	50	68	0.68	32	25	3	14	4	null	0	0.64	0.5	"L"	1	177	130	3.54	47	0.94	22	26	69	61	8	13	9	0	0	9	28	42	108	69	39	19	5	4	0	16	17	42	23	19	8	1	1	0	6	"https://assets.nhle.com/logos/…	0	31
20252026	"2026-01-25"	"MIN"	"Minnesota Wild"	"Western"	"Central"	53	68	0.641509	29	19	3	14	10	null	0	0.54717	0.358491	"OT"	1	171	154	3.226415	17	0.320755	27	35	85	84	1	14	6	7	0	8	26	33	86	70	16	15	8	3	0	11	10	33	39	-6	4	4	2	0	1	"https://assets.nhle.com/logos/…	0	27

team_names = standings.sort(by="team_name")["team_name"].str.to_uppercase().to_list()
team_codes = standings.sort(by="team_name")["team"].str.to_uppercase().to_list()
team_names_dict = dict(zip(team_codes, team_names, strict=False))

home_teams = dict(zip(schedule["game_id"].to_list(), schedule["home_team"].to_list(), strict=False))

Game IDs

Create a list of game IDs to scrape

conds = pl.col("game_state") == "OFF"

game_ids = schedule.filter(conds)["game_id"].unique().to_list()

Latest date

Important! If you don’t set this, it will just pull from the last completed game

conds = pl.col("game_state") == "OFF"
latest_date = schedule.filter(conds)["game_date"].max()  # In YYYY-MM-DD format
latest_date = "2026-01-24"

Checking to see if you’ve already scraped the data

Check to see if you’ve already scraped, so you’re only getting the game_ids that you need

data_directory = Path.cwd() / "data"
stats_file = data_directory / "team_stats.csv"

if not data_directory:
    data_directory.mkdir()

if stats_file:
    team_stats = pl.read_csv(source=stats_file, infer_schema_length=2000)

    saved_game_ids = team_stats["game_id"].to_list()
    game_ids = [x for x in game_ids if x not in saved_game_ids]

Scrape play-by-play and team stats

Scrape play-by-play data and aggregate team stats using the Scraper object, but only if there are game_ids to scrape

if game_ids:
    scraper = Scraper(game_ids, backend="polars")

    pbp = scraper.play_by_play
    scraped_team_stats = scraper.team_stats

Combine scraped and saved data

Combine the scraped data and the saved data, then save the new file, but only if there are game_ids

if game_ids:
    team_stats = pl.concat(
        [team_stats.with_columns(pl.col("bsf_adj_percent").cast(pl.Float64)), scraped_team_stats], strict=False
    )  # Quick, don't ask
    team_stats.write_csv(stats_file)

team_stats.head(5)

shape: (5, 141)

season	session	game_id	game_date	team	opp_team	strength_state	toi	gf	ga	gf_adj	ga_adj	hdgf	hdga	xgf	xga	xgf_adj	xga_adj	sf	sa	sf_adj	sa_adj	hdsf	hdsa	ff	fa	ff_adj	fa_adj	hdff	hdfa	cf	ca	cf_adj	ca_adj	bsf	bsa	bsf_adj	bsa_adj	msf	msa	msf_adj	msa_adj	hdmsf	hdmsa	teammate_block	hf	ht	give	take	ozf	nzf	dzf	fow	fol	ozfw	ozfl	nzfw	nzfl	dzfw	dzfl	pent0	pent2	pent4	pent5	pent10	pend0	pend2	pend4	pend5	pend10	gf_p60	ga_p60	gf_adj_p60	ga_adj_p60	hdgf_p60	hdga_p60	xgf_p60	xga_p60	xgf_adj_p60	xga_adj_p60	sf_p60	sa_p60	sf_adj_p60	sa_adj_p60	hdsf_p60	hdsa_p60	ff_p60	fa_p60	ff_adj_p60	fa_adj_p60	hdff_p60	hdfa_p60	cf_p60	ca_p60	cf_adj_p60	ca_adj_p60	bsf_p60	bsa_p60	bsf_adj_p60	bsa_adj_p60	msf_p60	msa_p60	msf_adj_p60	msa_adj_p60	hdmsf_p60	hdmsa_p60	teammate_block_p60	hf_p60	ht_p60	give_p60	take_p60	pent0_p60	pent2_p60	pent4_p60	pent5_p60	pent10_p60	pend0_p60	pend2_p60	pend4_p60	pend5_p60	pend10_p60	gf_percent	gf_adj_percent	hdgf_percent	xgf_percent	xgf_adj_percent	sf_percent	sf_adj_percent	hdsf_percent	ff_percent	ff_adj_percent	hdff_percent	cf_percent	cf_adj_percent	bsf_percent	bsf_adj_percent	msf_percent	msf_adj_percent	hdmsf_percent	hf_percent	take_percent
i64	str	i64	str	str	str	str	f64	i64	i64	f64	f64	i64	i64	f64	f64	f64	f64	i64	i64	f64	f64	i64	i64	i64	i64	f64	f64	i64	i64	i64	i64	f64	f64	i64	i64	f64	f64	i64	i64	f64	f64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	str	f64	f64	f64	f64	f64
20252026	"R"	2025020099	"2025-10-21"	"EDM"	"OTT"	"4v3"	1.983333	1	0	1.114907	0.0	0	0	0.444541	0.0	0.511756	0.0	3	0	3.427895	0.0	0	0	3	0	3.377222	0.0	0	0	3	0	3.377222	0.0	0	0	0.0	0.0	0	0	0.0	0.0	0	0	0	0	0	0	0	1	0	1	1	1	1	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	30.252101	0.0	33.728274	0.0	0.0	0.0	13.448286	0.0	15.481704	0.0	90.756303	0.0	103.701017	0.0	0.0	0.0	90.756303	0.0	102.168056	0.0	0.0	0.0	90.756303	0.0	102.168056	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	1.0	1.0	null	1.0	1.0	1.0	1.0	null	1.0	1.0	null	1.0	1.0	null	null	null	null	null	null	null
20252026	"R"	2025020518	"2025-12-15"	"LAK"	"DAL"	"5v4"	5.2	0	0	0.0	0.0	0	0	1.068176	0.002684	1.189075	0.002455	5	1	5.40411	1.080822	1	0	10	1	11.108795	1.087149	5	0	11	1	11.108795	1.087149	1	0	0.0	0.0	5	0	5.673051	0.0	4	0	0	2	1	2	0	3	2	1	3	3	2	1	1	1	0	1	0	0	0	0	0	0	1	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0	12.325111	0.030966	13.720099	0.028331	57.692308	11.538462	62.355111	12.471022	11.538462	0.0	115.384615	11.538462	128.178406	12.544025	57.692308	0.0	126.923077	11.538462	128.178406	12.544025	11.538462	0.0	0.0	0.0	57.692308	0.0	65.458283	0.0	46.153846	0.0	0.0	23.076923	11.538462	23.076923	0.0	0.0	0.0	0.0	0.0	0.0	0.0	11.538462	0.0	0.0	0.0	null	null	null	0.997494	0.997939	0.833333	0.833333	1.0	0.909091	0.91086	1.0	0.916667	0.91086	1.0	null	1.0	1.0	1.0	0.666667	0.0
20252026	"R"	2025020006	"2025-10-08"	"EDM"	"CGY"	"5v5"	48.75	1	2	0.98142	2.038595	0	1	1.57485	0.92412	1.542409	0.938942	22	13	22.683231	12.7549	3	2	30	24	30.718999	23.469424	4	3	46	38	30.718999	23.469424	14	14	0.0	0.0	8	11	8.09824	10.667752	1	1	2	22	34	15	3	11	18	10	22	17	6	5	10	8	6	4	0	2	0	0	0	0	3	0	0	0	1.230769	2.461538	1.207901	2.50904	0.0	1.230769	1.938277	1.137379	1.898349	1.155621	27.076923	16.0	27.917822	15.698339	3.692308	2.461538	36.923077	29.538462	37.807998	28.885445	4.923077	3.692308	56.615385	46.769231	37.807998	28.885445	17.230769	17.230769	0.0	0.0	9.846154	13.538462	9.967065	13.129541	1.230769	1.230769	2.461538	27.076923	41.846154	18.461538	3.692308	0.0	2.461538	0.0	0.0	0.0	0.0	3.692308	0.0	0.0	0.0	0.333333	0.324972	0.0	0.6302	0.6216	0.628571	0.64008	0.6	0.555556	0.566892	0.571429	0.547619	0.566892	0.5	null	0.421053	0.431538	0.5	0.392857	0.166667
20252026	"R"	2025020090	"2025-10-19"	"EDM"	"DET"	"5v4"	2.0	0	0	0.0	0.0	0	0	0.220914	0.0	0.220276	0.0	2	0	1.899528	0.0	0	0	3	0	2.877334	0.0	0	0	5	0	2.877334	0.0	1	0	0.0	0.0	1	0	0.959111	0.0	0	0	1	0	1	0	0	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0	6.627434	0.0	6.608293	0.0	60.0	0.0	56.985827	0.0	0.0	0.0	90.0	0.0	86.320023	0.0	0.0	0.0	150.0	0.0	86.320023	0.0	30.0	0.0	0.0	0.0	30.0	0.0	28.773341	0.0	0.0	0.0	30.0	0.0	30.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	null	null	null	1.0	1.0	1.0	1.0	null	1.0	1.0	null	1.0	1.0	1.0	null	1.0	1.0	null	0.0	null
20252026	"R"	2025020473	"2025-12-09"	"DAL"	"WPG"	"4vE"	0.933333	0	0	0.0	0.0	0	0	0.0	0.146976	0.0	0.146976	0	1	0.0	1.0	0	0	0	1	0.0	1.0	0	0	1	1	0.0	1.0	0	0	0.0	0.0	0	0	0.0	0.0	0	0	1	0	0	0	0	0	0	1	1	0	0	0	0	0	1	0	0	0	0	0	0	0	0	0	0	0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	9.44848	0.0	9.44848	0.0	64.285714	0.0	64.285714	0.0	0.0	0.0	64.285714	0.0	64.285714	0.0	0.0	64.285714	64.285714	0.0	64.285714	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	64.285714	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	null	null	null	0.0	0.0	0.0	0.0	null	0.0	0.0	null	0.5	0.0	null	null	null	null	null	null	null

Munge the data

Cleaning the data and prepping for analysis

Add Home games to the data set

Check to see if team is the home team, adding a dummy column

home_map = dict(zip(schedule["game_id"], schedule["home_team"], strict=False))

team_stats = team_stats.with_columns(
    is_home=pl.when(pl.col("game_id").replace_strict(home_map, return_dtype=str) == pl.col("team"))
    .then(pl.lit(1))
    .otherwise(pl.lit(0))
)

Aggregate strength state

Add a column to aggregate strength states (e.g., powerplay for 5v4, 5v3, and 4v3)

def aggregate_strength_states(team_stats: pl.DataFrame) -> pl.DataFrame:
    """Add a secondary strength state column to team stats data.

    Parameters:
        team_stats (pl.DataFrame):
            Polars dataframe of team statistics aggregated from the `chickenstats` library

    """
    # Create a new strength state column to combine powerplay and shorthanded
    even_strength = ["4v4", "3v3"]
    powerplay_list = ["5v4", "5v3", "4v3"]
    shorthanded_list = ["4v5", "3v4", "3v5"]

    team_stats = team_stats.with_columns(
        strength_state2=pl.when(pl.col("strength_state") == "5v5")
        .then(pl.lit("5v5"))
        .otherwise(
            pl.when(pl.col("strength_state").is_in(even_strength))
            .then(pl.lit("even_strength"))
            .otherwise(
                pl.when(pl.col("strength_state").is_in(powerplay_list))
                .then(pl.lit("powerplay"))
                .otherwise(
                    pl.when(pl.col("strength_state").is_in(shorthanded_list))
                    .then(pl.lit("shorthanded"))
                    .otherwise(pl.col("strength_state"))
                )
            )
        )
    )

    return team_stats

team_stats = aggregate_strength_states(team_stats)

Prep overall NHL stats

Function to calculate the goals scored and allowed above the expected goals model, at the overall NHL season level, adjusted for score and venue

def prep_nhl_stats(team_stats: pl.DataFrame) -> pl.DataFrame:
    """Function to calculate the goals scored and allowed above the expected goals model, adjusted for score and venue.

    Parameters:
        team_stats (pl.DataFrame):
            Polars dataframe of team statistics aggregated from the `chickenstats` library


    """
    if "strength_state2" not in team_stats.columns:
        team_stats = aggregate_strength_states(team_stats)

    group_columns = ["season", "session", "is_home", "strength_state2"]
    non_agg_columns = ["team", "strength_state", "game_id", "game_date", "opp_team"]

    agg_stats = tuple(
        pl.col(x).sum()
        for x in team_stats.columns
        if x not in group_columns and x not in non_agg_columns and "p60" not in x and "percent" not in x
    )

    agg_stats = agg_stats + (pl.col("game_id").n_unique(),)

    nhl_stats = team_stats.group_by(group_columns).agg(agg_stats)

    nhl_stats = nhl_stats.with_columns(
        g_score_ax=pl.col("gf_adj") - pl.col("xgf_adj"), g_save_ax=pl.col("xga_adj") - pl.col("ga_adj")
    )

    nhl_stats = nhl_stats.with_columns(
        toi_gp=pl.col("toi") / pl.col("game_id"),
        gf_p60=pl.col("gf") / pl.col("toi") * 60,
        ga_p60=pl.col("ga") / pl.col("toi") * 60,
        gf_adj_p60=pl.col("gf_adj") / pl.col("toi") * 60,
        ga_adj_p60=pl.col("ga_adj") / pl.col("toi") * 60,
        xgf_p60=pl.col("xgf") / pl.col("toi") * 60,
        xga_p60=pl.col("xga") / pl.col("toi") * 60,
        xgf_adj_p60=pl.col("xgf_adj") / pl.col("toi") * 60,
        xga_adj_p60=pl.col("xga_adj") / pl.col("toi") * 60,
        g_score_ax_p60=pl.col("g_score_ax") / pl.col("toi") * 60,
        g_save_ax_p60=pl.col("g_save_ax") / pl.col("toi") * 60,
    )

    return nhl_stats

nhl_stats = prep_nhl_stats(team_stats)

nhl_stats.head(5)

shape: (5, 81)

season	session	is_home	strength_state2	toi	gf	ga	gf_adj	ga_adj	hdgf	hdga	xgf	xga	xgf_adj	xga_adj	sf	sa	sf_adj	sa_adj	hdsf	hdsa	ff	fa	ff_adj	fa_adj	hdff	hdfa	cf	ca	cf_adj	ca_adj	bsf	bsa	bsf_adj	bsa_adj	msf	msa	msf_adj	msa_adj	hdmsf	hdmsa	teammate_block	hf	ht	give	take	ozf	nzf	dzf	fow	fol	ozfw	ozfl	nzfw	nzfl	dzfw	dzfl	pent0	pent2	pent4	pent5	pent10	pend0	pend2	pend4	pend5	pend10	game_id	g_score_ax	g_save_ax	toi_gp	gf_p60	ga_p60	gf_adj_p60	ga_adj_p60	xgf_p60	xga_p60	xgf_adj_p60	xga_adj_p60	g_score_ax_p60	g_save_ax_p60
i64	str	i32	str	f64	i64	i64	f64	f64	i64	i64	f64	f64	f64	f64	i64	i64	f64	f64	i64	i64	i64	i64	f64	f64	i64	i64	i64	i64	f64	f64	i64	i64	f64	f64	i64	i64	f64	f64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	u32	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64
20252026	"R"	0	"Ev3"	0.95	0	0	0.0	0.0	0	0	0.570574	0.0	0.570574	0.0	1	0	1.0	0.0	0	0	2	0	2.0	0.0	0	0	2	0	2.0	0.0	0	0	0.0	0.0	1	0	1.0	0.0	0	0	0	0	0	0	0	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0	0	4	0	0	0	5	-0.570574	0.0	0.19	0.0	0.0	0.0	0.0	36.036246	0.0	36.036246	0.0	-36.036246	0.0
20252026	"R"	0	"shorthanded"	4349.6	68	538	63.750348	497.900242	29	232	58.837307	663.579193	64.380619	614.13167	737	3776	696.292742	3549.176824	136	1134	943	5620	887.272599	5244.646296	176	1612	1089	7216	887.272599	5244.646296	141	1596	0.0	0.0	206	1844	221.448081	1717.691838	40	478	5	588	269	551	357	2122	675	1863	2137	2523	85	2037	311	364	1741	122	1	107	2	1	7	1	167	2	2	8	809	-0.630271	116.231428	5.376514	0.938017	7.421372	0.879396	6.868221	0.811624	9.153658	0.88809	8.471561	-0.008694	1.60334
20252026	"R"	0	"4vE"	57.316667	7	6	7.0	6.0	2	4	11.032402	10.128565	11.032402	10.128565	7	68	7.0	68.0	2	21	21	103	21.0	103.0	2	28	28	153	21.0	103.0	6	50	0.0	0.0	14	35	14.0	35.0	0	7	1	7	7	8	2	39	3	20	22	40	0	39	2	1	20	0	0	12	0	0	0	0	2	0	0	0	64	-4.032402	4.128565	0.895573	7.327712	6.280896	7.327712	6.280896	11.548895	10.602743	11.548895	10.602743	-4.221183	4.321847
20252026	"R"	0	"5v5"	39408.2	1626	1649	1681.560855	1584.330452	752	719	1794.319746	1927.468912	1875.327219	1845.673238	17175	17807	17438.993767	17324.570508	3819	3938	26435	27489	26935.165024	26636.611338	5673	5878	37054	37213	26935.165024	26636.611338	9669	9724	0.0	0.0	9260	9682	9427.907134	9376.066002	1854	1940	950	15354	15900	11009	3079	11103	12426	11840	17624	17745	5397	5706	6239	6187	5988	5852	11	2202	32	183	68	6	1987	21	181	56	820	-193.766363	261.342785	48.05878	2.475627	2.510645	2.56022	2.412184	2.731898	2.934621	2.855234	2.810085	-0.295014	0.397901
20252026	"R"	1	"4vE"	38.566667	9	8	9.0	8.0	2	2	7.132744	7.381997	7.132744	7.381997	9	41	9.0	41.0	2	5	14	72	14.0	72.0	2	10	21	100	14.0	72.0	7	28	0.0	0.0	5	31	5.0	31.0	0	5	0	4	1	7	3	31	1	11	12	31	0	31	1	0	11	0	0	8	0	0	0	0	2	0	0	0	51	1.867256	-0.618003	0.756209	14.001729	12.445981	14.001729	12.445981	11.096749	11.484524	11.096749	11.484524	2.904979	-0.961457

Prep team stats

Function to prep and aggregate team stats to match up with the schedule for predicting the game’s values

def prep_team_stats(team_stats: pl.DataFrame, nhl_stats: pl.DataFrame, latest_date: str = latest_date) -> pl.DataFrame:
    """Prepare team stats dataframe for later analysis. Nested within the prep today's function.

    Parameters:
        team_stats (pl.DataFrame):
            Polars dataframe of team statistics aggregated from the `chickenstats` library
        nhl_stats (pl.DataFrame):
            Polars dataframe of aggregated NHL-level statistics from the team stats dataframe at an earlier stage

    """
    if "strength_state2" not in team_stats.columns:
        team_stats = aggregate_strength_states(team_stats)

    team_stats = team_stats.filter(pl.col("game_date") < latest_date)

    # Aggregate statistics with the new columns
    group_columns = ["season", "session", "team", "is_home", "strength_state2"]
    non_agg_columns = ["game_id", "game_date", "opp_team", "strength_state"]

    agg_stats = tuple(
        pl.col(x).sum()
        for x in team_stats.columns
        if x not in group_columns and x not in non_agg_columns and "p60" not in x and "percent" not in x
    )

    agg_stats = agg_stats + (pl.col("game_id").n_unique(),)

    team_stats = team_stats.group_by(group_columns).agg(agg_stats)

    # Adding columns for goals scored and allowed above the expected goals model, adjusted for score and venue

    team_stats = team_stats.with_columns(
        g_score_ax=pl.col("gf_adj") - pl.col("xgf_adj"), g_save_ax=pl.col("xga_adj") - pl.col("ga_adj")
    )

    team_stats = team_stats.with_columns(
        toi_gp=pl.col("toi") / pl.col("game_id"),
        gf_p60=pl.col("gf") / pl.col("toi") * 60,
        ga_p60=pl.col("ga") / pl.col("toi") * 60,
        gf_adj_p60=pl.col("gf_adj") / pl.col("toi") * 60,
        ga_adj_p60=pl.col("ga_adj") / pl.col("toi") * 60,
        xgf_p60=pl.col("xgf") / pl.col("toi") * 60,
        xga_p60=pl.col("xga") / pl.col("toi") * 60,
        xgf_adj_p60=pl.col("xgf_adj") / pl.col("toi") * 60,
        xga_adj_p60=pl.col("xga_adj") / pl.col("toi") * 60,
        g_score_ax_p60=pl.col("g_score_ax") / pl.col("toi") * 60,
        g_save_ax_p60=pl.col("g_save_ax") / pl.col("toi") * 60,
    )

    # Adding mean NHL columns for the columns we'll use to predict the games
    predict_columns = [
        "xgf_p60",
        "xga_p60",
        "xgf_adj_p60",
        "xga_adj_p60",
        "gf_p60",
        "ga_p60",
        "gf_adj_p60",
        "ga_adj_p60",
        "g_score_ax_p60",
        "g_save_ax_p60",
        "toi_gp",
    ]

    for column in predict_columns:
        nhl_mean_map = dict(
            zip((nhl_stats["strength_state2"] + nhl_stats["is_home"].cast(pl.String)), nhl_stats[column], strict=False)
        )

        team_stats = team_stats.with_columns(
            (pl.col("strength_state2") + pl.col("is_home").cast(pl.String))
            .replace_strict(nhl_mean_map, return_dtype=pl.Float64, default=None)
            .alias(f"mean_{column}")
        )

    # Calculating the strength scores

    team_stats = team_stats.with_columns(
        off_strength=pl.col("xgf_adj_p60") / pl.col("mean_xgf_adj_p60"),
        def_strength=pl.col("xga_adj_p60") / pl.col("mean_xga_adj_p60"),
        toi_comparison=pl.col("toi_gp") / pl.col("mean_toi_gp"),
        scoring_strength=pl.col("g_score_ax_p60") / pl.col("mean_g_score_ax_p60"),
        goalie_strength=pl.col("g_save_ax_p60") / pl.col("mean_g_save_ax_p60"),
    )

    return team_stats

team_stats_agg = prep_team_stats(team_stats=team_stats, nhl_stats=nhl_stats, latest_date=latest_date)

team_stats_agg.head(5)

shape: (5, 98)

season	session	team	is_home	strength_state2	toi	gf	ga	gf_adj	ga_adj	hdgf	hdga	xgf	xga	xgf_adj	xga_adj	sf	sa	sf_adj	sa_adj	hdsf	hdsa	ff	fa	ff_adj	fa_adj	hdff	hdfa	cf	ca	cf_adj	ca_adj	bsf	bsa	bsf_adj	bsa_adj	msf	msa	msf_adj	msa_adj	hdmsf	hdmsa	teammate_block	hf	ht	give	take	ozf	nzf	dzf	fow	fol	ozfw	ozfl	nzfw	nzfl	dzfw	dzfl	pent0	pent2	pent4	pent5	pent10	pend0	pend2	pend4	pend5	pend10	game_id	g_score_ax	g_save_ax	toi_gp	gf_p60	ga_p60	gf_adj_p60	ga_adj_p60	xgf_p60	xga_p60	xgf_adj_p60	xga_adj_p60	g_score_ax_p60	g_save_ax_p60	mean_xgf_p60	mean_xga_p60	mean_xgf_adj_p60	mean_xga_adj_p60	mean_gf_p60	mean_ga_p60	mean_gf_adj_p60	mean_ga_adj_p60	mean_g_score_ax_p60	mean_g_save_ax_p60	mean_toi_gp	off_strength	def_strength	toi_comparison	scoring_strength	goalie_strength
i64	str	str	i32	str	f64	i64	i64	f64	f64	i64	i64	f64	f64	f64	f64	i64	i64	f64	f64	i64	i64	i64	i64	f64	f64	i64	i64	i64	i64	f64	f64	i64	i64	f64	f64	i64	i64	f64	f64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	i64	u32	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64
20252026	"R"	"DET"	0	"5vE"	23.083333	5	3	5.0	3.0	1	2	3.702068	3.046261	3.702068	3.046261	5	23	5.0	23.0	1	8	7	33	7.0	33.0	1	12	14	51	7.0	33.0	7	18	0.0	0.0	2	10	2.0	10.0	0	4	0	2	3	4	2	16	0	9	9	16	0	16	0	0	9	0	0	15	0	0	1	0	1	0	0	1	17	1.297932	0.046261	1.357843	12.99639	7.797834	12.99639	7.797834	9.622703	7.91808	9.622703	7.91808	3.373687	0.120246	20.533321	8.290257	20.533321	8.290257	18.884813	7.26339	18.752751	7.26339	-1.78057	1.026867	1.071541	0.468638	0.955107	1.267187	-1.894723	0.117099
20252026	"R"	"DAL"	0	"ILLEGAL"	0.383333	0	0	0.0	0.0	0	0	0.0	0.0	0.0	0.0	0	0	0.0	0.0	0	0	0	0	0.0	0.0	0	0	0	0	0.0	0.0	0	0	0.0	0.0	0	0	0.0	0.0	0	0	0	0	1	0	1	0	0	0	0	0	0	0	0	0	0	0	0	1	0	0	0	0	1	0	0	0	3	0.0	0.0	0.127778	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.0	0.09902	NaN	NaN	1.290429	NaN	NaN
20252026	"R"	"WSH"	0	"shorthanded"	143.116667	1	22	0.868644	20.284237	0	10	1.941511	26.230574	2.121831	24.042556	29	142	27.490349	132.334614	3	39	35	223	32.789011	208.023041	3	58	38	284	32.789011	208.023041	3	61	0.0	0.0	6	81	6.641546	76.493509	0	19	0	10	5	19	13	73	16	63	67	85	2	71	6	10	59	4	0	7	0	0	0	0	3	0	0	0	25	-1.253187	3.758318	5.724667	0.419238	9.223244	0.364169	8.503931	0.813956	10.996864	0.889553	10.079562	-0.525384	1.575631	0.811624	9.153658	0.88809	8.471561	0.938017	7.421372	0.879396	6.868221	-0.008694	1.60334	5.376514	1.001647	1.189812	1.064754	60.429292	0.982718
20252026	"R"	"NYR"	1	"4vE"	2.033333	0	1	0.0	1.0	0	1	0.0	1.069307	0.0	1.069307	0	4	0.0	4.0	0	1	0	7	0.0	7.0	0	2	0	7	0.0	7.0	0	0	0.0	0.0	0	3	0.0	3.0	0	1	0	0	0	1	0	1	0	0	0	1	0	1	0	0	0	0	0	1	0	0	0	0	0	0	0	0	3	0.0	0.069307	0.677778	0.0	29.508197	0.0	29.508197	0.0	31.55331	0.0	31.55331	0.0	2.045114	11.096749	11.484524	11.096749	11.484524	14.001729	12.445981	14.001729	12.445981	2.904979	-0.961457	0.756209	0.0	2.747463	0.896283	0.0	-2.127099
20252026	"R"	"SEA"	0	"even_strength"	46.733333	3	2	1.992679	2.007375	0	1	3.262164	2.997766	3.357584	2.922361	28	14	27.596736	13.568321	4	2	35	25	33.614441	24.478333	6	3	41	30	33.614441	24.478333	6	5	0.0	0.0	7	11	6.138113	10.873773	2	1	0	3	5	3	4	11	12	15	21	17	7	4	6	6	8	7	0	2	0	0	0	0	1	0	0	0	15	-1.364905	0.914987	3.115556	3.851641	2.56776	2.558362	2.577228	4.188228	3.848773	4.310736	3.751962	-1.752375	1.174733	4.209139	4.748321	4.333296	4.613874	3.866962	3.91412	3.873704	3.844325	-0.459592	0.769549	2.695586	0.994794	0.813191	1.155799	3.812893	1.526522

Prep today’s games

Function to prepare today’s games

def prep_todays_games(
    schedule: pl.DataFrame, team_stats: pl.DataFrame, nhl_stats: pl.DataFrame, latest_date: str = latest_date
) -> pl.DataFrame:
    """Function to prep today's games."""
    team_stats = prep_team_stats(team_stats=team_stats, nhl_stats=nhl_stats, latest_date=latest_date)

    todays_games = schedule.filter(pl.col("game_date") == latest_date)  # Filter the schedule for today's games

    venues = ["home", "away"]  # We need to account for venue effects on scorekeeping and performance
    strength_states = ["5v5", "powerplay", "shorthanded"]  # Segmenting by strength state

    team_value_dicts = {}  # Dictionary to store values to map to team stats columns

    for venue in venues:  # Looping through home and away for team stats
        for strength_state in strength_states:  # Looping through the strength states
            values = [  # These are the columns we need to append to the dataframe
                "off_strength",
                "def_strength",
                "scoring_strength",
                "goalie_strength",
                "toi_comparison",
            ]

            if strength_state == "powerplay":  # We don't care about defensive comparisons for powerplay
                remove_list = ["def_strength", "goalie_strength"]
                values = [x for x in values if x not in remove_list]

            elif strength_state == "shorthanded":  # We don't care about defensive comparisons for shorthanded
                remove_list = ["off_strength", "scoring_strength"]
                values = [x for x in values if x not in remove_list]

            for value in values:  # Need to loop through the new column values
                # Disaggregating the next couple of steps for readability

                if venue == "home":  # Getting a dummy value for filtering the dataframe
                    venue_dummy = 1

                else:
                    venue_dummy = 0

                dummy_replacements = {1: "home", 0: "away"}  # Dictionary to replace dummy variables later

                filter_conditions = (pl.col("strength_state2") == strength_state, pl.col("is_home") == venue_dummy)
                filter_df = team_stats.filter(filter_conditions)  # Filtering the dataframe

                # Getting a mapping column value for the nested dictionary, e.g., NAShome

                team_column = filter_df["team"]  # Getting the team names portion of the mapping key
                venue_column = filter_df["is_home"].replace_strict(
                    dummy_replacements, return_dtype=pl.String
                )  # Getting venue portion, but need to replace the dummy values

                mapping_column = team_column + venue_column  # Mapping column for nested dictionary, e.g., NAShome
                value_column = filter_df[value]  # Value column for nested dictionary
                nested_dictionary = dict(
                    zip(mapping_column, value_column, strict=False)
                )  # Nested dictionary being added to team_value_dicts

                dictionary_key = f"{venue}_{strength_state}_{value}"  # Key for the dictionary being added to team_value_dicts, e.g., home_5v5_off_strength

                team_value_dicts.update(
                    {dictionary_key: nested_dictionary}
                )  # Adding the nested dictionary to team_value_dicts

    predict_columns = [  # Columns to predict as part of the simulation
        "xgf_p60",
        "xga_p60",
        "xgf_adj_p60",
        "xga_adj_p60",
        "gf_p60",
        "ga_p60",
        "gf_adj_p60",
        "ga_adj_p60",
        "g_score_ax_p60",
        "g_save_ax_p60",
        "toi_gp",
    ]

    nhl_value_dicts = {}

    for venue in venues:  # Looping through home and away for mean NHL stats
        for strength_state in strength_states:  # Looping through the strength states
            for column in predict_columns:  # Need to loop through the new column values
                if venue == "home":  # Setting the integer for the dummy column
                    venue_dummy = 1

                else:
                    venue_dummy = 0

                # Getting the column to match on for the NHL stats
                filter_conditions = (
                    pl.col("strength_state2") == strength_state,
                    pl.col("is_home") == venue_dummy,
                )  # Splitting out for readability
                field_value = nhl_stats.filter(filter_conditions)[column][0]  # We need the value

                nhl_value_dicts.update(
                    {f"{venue}_{strength_state}_{column}": field_value}
                )  # Becomes part of the replace function later on

    add_columns = []  # Columns being added to the team stats dataframe e.g., off_strength_score, based on strength state and venue

    for (
        key,
        value,
    ) in team_value_dicts.items():  # Iterating through the dictionary created earlier to add them to the dataframe
        venue = f"{key[:4]}"  # This pulls in either "home" or "away" from the key, which is something like home_5v5_off_strength
        venue_team = f"{venue}_team"

        # Adds a new column with the values based on the team and strength state, with a new column name like home_5v5_off_strength
        new_column = (pl.col(venue_team) + venue).replace_strict(value, return_dtype=pl.Float64).alias(key)

        add_columns.append(new_column)  # Appending to the new columns list

    for key, value in nhl_value_dicts.items():  # Adding the NHL mean columns to the dataframe
        add_columns.append(
            pl.lit(value).alias(f"mean_{key}")
        )  # Adding the NHL mean column to the list to be added to the dataframe

    todays_games = todays_games.with_columns(add_columns)  # Adding the columns to the dataframe

    # Predicting home and away time on ice, goals, and expected goals based on historical performance and NHL means

    todays_games = todays_games.with_columns(
        predicted_home_5v5_toi=pl.col("home_5v5_toi_comparison")  # predicted home 5v5 TOI
        * pl.col("away_5v5_toi_comparison")
        * pl.col("mean_home_5v5_toi_gp"),
        predicted_home_powerplay_toi=pl.col("home_powerplay_toi_comparison")  # predicted home powerplay TOI
        * pl.col("away_shorthanded_toi_comparison")
        * pl.col("mean_home_powerplay_toi_gp"),
        predicted_home_shorthanded_toi=pl.col("home_shorthanded_toi_comparison")  # predicted home shorthanded TOI
        * pl.col("away_powerplay_toi_comparison")
        * pl.col("mean_home_shorthanded_toi_gp"),
        predicted_home_5v5_xgf_p60=pl.col("home_5v5_off_strength")
        * pl.col("away_5v5_def_strength")
        * pl.col("mean_home_5v5_xgf_adj_p60"),
        predicted_home_5v5_xga_p60=pl.col("home_5v5_def_strength")
        * pl.col("away_5v5_off_strength")
        * pl.col("mean_home_5v5_xga_adj_p60"),
        predicted_home_5v5_gf_p60=pl.col("home_5v5_off_strength")
        * pl.col("away_5v5_def_strength")
        * pl.col("home_5v5_scoring_strength")
        * pl.col("away_5v5_goalie_strength")
        * pl.col("mean_home_5v5_gf_adj_p60"),
        predicted_home_5v5_ga_p60=pl.col("home_5v5_def_strength")
        * pl.col("away_5v5_off_strength")
        * pl.col("home_5v5_goalie_strength")
        * pl.col("away_5v5_scoring_strength")
        * pl.col("mean_home_5v5_ga_adj_p60"),
        predicted_home_powerplay_xgf_p60=pl.col("home_powerplay_off_strength")
        * pl.col("away_shorthanded_def_strength")
        * pl.col("mean_home_powerplay_xgf_adj_p60"),
        predicted_home_powerplay_gf_p60=pl.col("home_powerplay_off_strength")
        * pl.col("away_shorthanded_def_strength")
        * pl.col("home_powerplay_scoring_strength")
        * pl.col("away_shorthanded_goalie_strength")
        * pl.col("mean_home_powerplay_gf_adj_p60"),
        predicted_home_shorthanded_xga_p60=pl.col("home_shorthanded_def_strength")
        * pl.col("away_powerplay_off_strength")
        * pl.col("mean_home_shorthanded_xga_adj_p60"),
        predicted_home_shorthanded_ga_p60=pl.col("home_shorthanded_def_strength")
        * pl.col("away_powerplay_off_strength")
        * pl.col("home_shorthanded_goalie_strength")
        * pl.col("away_powerplay_scoring_strength")
        * pl.col("mean_home_shorthanded_ga_adj_p60"),
        predicted_away_5v5_xgf_p60=pl.col("away_5v5_off_strength")
        * pl.col("home_5v5_def_strength")
        * pl.col("mean_away_5v5_xgf_adj_p60"),
        predicted_away_5v5_xga_p60=pl.col("away_5v5_def_strength")
        * pl.col("home_5v5_off_strength")
        * pl.col("mean_away_5v5_xga_adj_p60"),
        predicted_away_5v5_gf_p60=pl.col("away_5v5_off_strength")
        * pl.col("home_5v5_def_strength")
        * pl.col("away_5v5_scoring_strength")
        * pl.col("home_5v5_goalie_strength")
        * pl.col("mean_away_5v5_gf_adj_p60"),
        predicted_away_5v5_ga_p60=pl.col("away_5v5_def_strength")
        * pl.col("home_5v5_off_strength")
        * pl.col("away_5v5_goalie_strength")
        * pl.col("home_5v5_scoring_strength")
        * pl.col("mean_away_5v5_ga_adj_p60"),
        predicted_away_powerplay_xgf_p60=pl.col("away_powerplay_off_strength")
        * pl.col("home_shorthanded_def_strength")
        * pl.col("mean_away_powerplay_xgf_adj_p60"),
        predicted_away_powerplay_gf_p60=pl.col("away_powerplay_off_strength")
        * pl.col("home_shorthanded_def_strength")
        * pl.col("away_powerplay_scoring_strength")
        * pl.col("home_shorthanded_goalie_strength")
        * pl.col("mean_away_powerplay_gf_adj_p60"),
        predicted_away_shorthanded_xga_p60=pl.col("away_shorthanded_def_strength")
        * pl.col("home_powerplay_off_strength")
        * pl.col("mean_away_shorthanded_xga_adj_p60"),
        predicted_away_shorthanded_ga_p60=pl.col("away_shorthanded_def_strength")
        * pl.col("home_powerplay_off_strength")
        * pl.col("away_shorthanded_goalie_strength")
        * pl.col("home_powerplay_scoring_strength")
        * pl.col("mean_away_shorthanded_ga_adj_p60"),
    )

    return todays_games

todays_games = prep_todays_games(schedule=schedule, team_stats=team_stats, nhl_stats=nhl_stats, latest_date=latest_date)

todays_games.head(5)

shape: (5, 128)

season	session	game_id	game_date	start_time	game_state	home_team	home_team_id	home_score	away_team	away_team_id	away_score	venue	venue_timezone	neutral_site	game_date_dt_utc	tv_broadcasts	home_logo	home_logo_dark	away_logo	away_logo_dark	home_5v5_off_strength	home_5v5_def_strength	home_5v5_scoring_strength	home_5v5_goalie_strength	home_5v5_toi_comparison	home_powerplay_off_strength	home_powerplay_scoring_strength	home_powerplay_toi_comparison	home_shorthanded_def_strength	home_shorthanded_goalie_strength	home_shorthanded_toi_comparison	away_5v5_off_strength	away_5v5_def_strength	away_5v5_scoring_strength	away_5v5_goalie_strength	away_5v5_toi_comparison	away_powerplay_off_strength	away_powerplay_scoring_strength	away_powerplay_toi_comparison	away_shorthanded_def_strength	away_shorthanded_goalie_strength	away_shorthanded_toi_comparison	mean_home_5v5_xgf_p60	mean_home_5v5_xga_p60	mean_home_5v5_xgf_adj_p60	mean_home_5v5_xga_adj_p60	mean_home_5v5_gf_p60	mean_home_5v5_ga_p60	mean_home_5v5_gf_adj_p60	mean_home_5v5_ga_adj_p60	mean_home_5v5_g_score_ax_p60	mean_home_5v5_g_save_ax_p60	mean_home_5v5_toi_gp	mean_home_powerplay_xgf_p60	mean_home_powerplay_xga_p60	mean_home_powerplay_xgf_adj_p60	mean_home_powerplay_xga_adj_p60	mean_home_powerplay_gf_p60	mean_home_powerplay_ga_p60	mean_home_powerplay_gf_adj_p60	mean_home_powerplay_ga_adj_p60	mean_home_powerplay_g_score_ax_p60	mean_home_powerplay_g_save_ax_p60	mean_home_powerplay_toi_gp	mean_home_shorthanded_xgf_p60	mean_home_shorthanded_xga_p60	mean_home_shorthanded_xgf_adj_p60	mean_home_shorthanded_xga_adj_p60	mean_home_shorthanded_gf_p60	mean_home_shorthanded_ga_p60	mean_home_shorthanded_gf_adj_p60	mean_home_shorthanded_ga_adj_p60	mean_home_shorthanded_g_score_ax_p60	mean_home_shorthanded_g_save_ax_p60	mean_home_shorthanded_toi_gp	mean_away_5v5_xgf_p60	mean_away_5v5_xga_p60	mean_away_5v5_xgf_adj_p60	mean_away_5v5_xga_adj_p60	mean_away_5v5_gf_p60	mean_away_5v5_ga_p60	mean_away_5v5_gf_adj_p60	mean_away_5v5_ga_adj_p60	mean_away_5v5_g_score_ax_p60	mean_away_5v5_g_save_ax_p60	mean_away_5v5_toi_gp	mean_away_powerplay_xgf_p60	mean_away_powerplay_xga_p60	mean_away_powerplay_xgf_adj_p60	mean_away_powerplay_xga_adj_p60	mean_away_powerplay_gf_p60	mean_away_powerplay_ga_p60	mean_away_powerplay_gf_adj_p60	mean_away_powerplay_ga_adj_p60	mean_away_powerplay_g_score_ax_p60	mean_away_powerplay_g_save_ax_p60	mean_away_powerplay_toi_gp	mean_away_shorthanded_xgf_p60	mean_away_shorthanded_xga_p60	mean_away_shorthanded_xgf_adj_p60	mean_away_shorthanded_xga_adj_p60	mean_away_shorthanded_gf_p60	mean_away_shorthanded_ga_p60	mean_away_shorthanded_gf_adj_p60	mean_away_shorthanded_ga_adj_p60	mean_away_shorthanded_g_score_ax_p60	mean_away_shorthanded_g_save_ax_p60	mean_away_shorthanded_toi_gp	predicted_home_5v5_toi	predicted_home_powerplay_toi	predicted_home_shorthanded_toi	predicted_home_5v5_xgf_p60	predicted_home_5v5_xga_p60	predicted_home_5v5_gf_p60	predicted_home_5v5_ga_p60	predicted_home_powerplay_xgf_p60	predicted_home_powerplay_gf_p60	predicted_home_shorthanded_xga_p60	predicted_home_shorthanded_ga_p60	predicted_away_5v5_xgf_p60	predicted_away_5v5_xga_p60	predicted_away_5v5_gf_p60	predicted_away_5v5_ga_p60	predicted_away_powerplay_xgf_p60	predicted_away_powerplay_gf_p60	predicted_away_shorthanded_xga_p60	predicted_away_shorthanded_ga_p60
i64	i64	i64	str	str	str	str	i64	i64	str	i64	i64	str	str	i64	datetime[μs, UTC]	list[struct[5]]	str	str	str	str	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64
20252026	2	2025020810	"2026-01-24"	"13:00"	"OFF"	"NYI"	2	0	"BUF"	7	5	"UBS ARENA"	"America/New_York"	0	2026-01-24 18:00:00 UTC	[{282,"N","CA","SN",101}, {28,"A","US","MSG-B",392}, {409,"H","US","MSGSN",411}]	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	1.034561	1.096581	1.158714	2.416368	1.007736	1.044055	1.521772	1.015781	1.188409	2.726963	0.925069	1.012501	0.961882	1.151541	0.215725	1.011636	1.035095	2.307369	1.007818	0.986214	1.665816	0.798287	2.934621	2.731898	2.810085	2.855234	2.5076	2.477149	2.412184	2.56022	-0.397901	0.295014	48.05878	9.153658	0.811624	8.471561	0.88809	7.421372	0.938017	6.868221	0.879396	-1.60334	0.008694	5.376514	0.8438	9.05498	0.790489	9.832002	0.776952	7.297256	0.855383	7.877595	0.064894	1.954407	4.886435	2.731898	2.934621	2.855234	2.810085	2.475627	2.510645	2.56022	2.412184	-0.295014	0.397901	48.05878	9.05498	0.8438	9.832002	0.790489	7.297256	0.807421	7.877595	0.855383	-1.954407	-0.064894	4.886435	0.811624	9.153658	0.88809	8.471561	0.938017	7.421372	0.879396	6.868221	-0.008694	1.60334	5.376514	48.994101	4.359734	4.55563	2.796388	3.170135	0.60002	7.909622	8.722842	17.927331	12.09451	60.972804	3.170135	2.796388	7.909622	0.60002	12.09451	60.972804	8.722842	17.927331
20252026	2	2025020811	"2026-01-24"	"12:30"	"OFF"	"NSH"	18	2	"UTA"	68	5	"BRIDGESTONE ARENA"	"US/Central"	0	2026-01-24 18:30:00 UTC	[{282,"N","CA","SN",101}, {284,"N","CA","SN1",102}, … {558,"A","US","Utah16",381}]	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	1.037462	0.988341	2.150117	-0.352738	0.971059	0.765395	0.373227	1.075436	0.887251	1.967482	1.087414	1.078247	0.916217	0.99246	0.332	1.004526	0.755183	0.959497	1.048428	0.935059	1.076783	0.946185	2.934621	2.731898	2.810085	2.855234	2.5076	2.477149	2.412184	2.56022	-0.397901	0.295014	48.05878	9.153658	0.811624	8.471561	0.88809	7.421372	0.938017	6.868221	0.879396	-1.60334	0.008694	5.376514	0.8438	9.05498	0.790489	9.832002	0.776952	7.297256	0.855383	7.877595	0.064894	1.954407	4.886435	2.731898	2.934621	2.855234	2.810085	2.475627	2.510645	2.56022	2.412184	-0.295014	0.397901	48.05878	9.05498	0.8438	9.832002	0.790489	7.297256	0.807421	7.877595	0.855383	-1.954407	-0.064894	4.886435	0.811624	9.153658	0.88809	8.471561	0.938017	7.421372	0.879396	6.868221	-0.008694	1.60334	5.376514	46.879108	5.470938	5.570908	2.671097	3.042752	1.636746	-0.955142	6.063007	1.975465	6.587806	9.964306	3.042752	2.671097	-0.955142	1.636746	6.587806	9.964306	6.063007	1.975465
20252026	2	2025020812	"2026-01-24"	"19:00"	"OFF"	"BOS"	6	4	"MTL"	8	3	"TD GARDEN"	"US/Eastern"	0	2026-01-25 00:00:00 UTC	[{4,"N","CA","CBC",25}, {287,"N","CA","SNE",30}, … {31,"H","US","NESN",405}]	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	0.987704	1.062432	0.142162	2.533451	0.986729	0.967806	-0.130838	0.950063	1.109736	1.666703	1.108385	0.891365	0.940475	-0.56647	1.081176	0.975597	1.018541	-0.282751	1.18028	1.050737	0.70316	1.044419	2.934621	2.731898	2.810085	2.855234	2.5076	2.477149	2.412184	2.56022	-0.397901	0.295014	48.05878	9.153658	0.811624	8.471561	0.88809	7.421372	0.938017	6.868221	0.879396	-1.60334	0.008694	5.376514	0.8438	9.05498	0.790489	9.832002	0.776952	7.297256	0.855383	7.877595	0.064894	1.954407	4.886435	2.731898	2.934621	2.855234	2.810085	2.475627	2.510645	2.56022	2.412184	-0.295014	0.397901	48.05878	9.05498	0.8438	9.832002	0.790489	7.297256	0.807421	7.877595	0.855383	-1.954407	-0.064894	4.886435	0.811624	9.153658	0.88809	8.471561	0.938017	7.421372	0.879396	6.868221	-0.008694	1.60334	5.376514	46.263771	5.334918	6.392454	2.610318	2.703949	0.344401	-3.479556	8.614815	-0.642564	11.113224	-4.196178	2.703949	2.610318	-3.479556	0.344401	11.113224	-4.196178	8.614815	-0.642564
20252026	2	2025020813	"2026-01-24"	"19:00"	"OFF"	"OTT"	9	1	"CAR"	12	4	"CANADIAN TIRE CENTRE"	"US/Eastern"	0	2026-01-25 00:00:00 UTC	[{286,"N","CA","CITY",26}, {284,"N","CA","SN1",102}, {561,"A","US","FDSNSO",202}]	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	0.99401	0.817523	1.472839	-0.876273	0.97729	0.917459	-0.504441	1.084449	0.738982	-0.222868	1.006613	1.197254	0.969167	2.77916	1.349061	1.002667	0.987445	1.758074	1.084635	0.730912	0.435137	1.012118	2.934621	2.731898	2.810085	2.855234	2.5076	2.477149	2.412184	2.56022	-0.397901	0.295014	48.05878	9.153658	0.811624	8.471561	0.88809	7.421372	0.938017	6.868221	0.879396	-1.60334	0.008694	5.376514	0.8438	9.05498	0.790489	9.832002	0.776952	7.297256	0.855383	7.877595	0.064894	1.954407	4.886435	2.731898	2.934621	2.855234	2.810085	2.475627	2.510645	2.56022	2.412184	-0.295014	0.397901	48.05878	9.05498	0.8438	9.832002	0.790489	7.297256	0.807421	7.877595	0.855383	-1.954407	-0.064894	4.886435	0.811624	9.153658	0.88809	8.471561	0.938017	7.421372	0.879396	6.868221	-0.008694	1.60334	5.376514	47.092626	5.90121	5.33505	2.707128	2.794655	4.617284	-6.102626	5.680874	-1.010956	7.174449	-2.252291	2.794655	2.707128	-6.102626	4.617284	7.174449	-2.252291	5.680874	-1.010956
20252026	2	2025020814	"2026-01-24"	"19:00"	"OFF"	"CBJ"	29	8	"TBL"	14	5	"NATIONWIDE ARENA"	"US/Eastern"	0	2026-01-25 00:00:00 UTC	[{581,"A","US","The Spot",211}, {565,"H","US","FDSNOH",286}]	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	"https://assets.nhle.com/logos/…	1.012536	1.05335	0.720217	2.488828	1.040104	1.094538	2.480706	0.774975	1.056585	-0.306012	0.909909	1.183861	0.855816	0.194922	1.82045	0.974809	0.966583	0.140879	1.041937	0.983017	2.338719	1.034219	2.934621	2.731898	2.810085	2.855234	2.5076	2.477149	2.412184	2.56022	-0.397901	0.295014	48.05878	9.153658	0.811624	8.471561	0.88809	7.421372	0.938017	6.868221	0.879396	-1.60334	0.008694	5.376514	0.8438	9.05498	0.790489	9.832002	0.776952	7.297256	0.855383	7.877595	0.064894	1.954407	4.886435	2.731898	2.934621	2.855234	2.810085	2.475627	2.510645	2.56022	2.412184	-0.295014	0.397901	48.05878	9.05498	0.8438	9.832002	0.790489	7.297256	0.807421	7.877595	0.855383	-1.954407	-0.064894	4.886435	0.811624	9.153658	0.88809	8.471561	0.938017	7.421372	0.879396	6.868221	-0.008694	1.60334	5.376514	48.726928	4.309246	4.632673	2.435063	3.560536	2.740585	1.548844	9.114975	42.873575	10.041199	-0.346835	3.560536	2.435063	1.548844	2.740585	10.041199	-0.346835	9.114975	42.873575

Simulating the actual games

Functions to simulate the games

Random float

Generate a random floating number between 0 and 1

def random_float() -> float:
    """Generate a random floating number between 0 and 1."""
    random_generator = np.random.default_rng()

    # return random_generator.triangular(left=0.0, mode=0.5, right=1.0)
    return random_generator.random()

Simulating an individual game

Predicts various game values, based on a poissoin distribution

def simulate_game(game: dict) -> dict:
    """Docstring."""
    prediction = {}

    home_5v5_toi = poisson.ppf(random_float(), game["predicted_home_5v5_toi"])
    home_pp_toi = poisson.ppf(random_float(), game["predicted_home_powerplay_toi"])
    home_sh_toi = poisson.ppf(random_float(), game["predicted_home_shorthanded_toi"])

    total_toi = home_5v5_toi + home_pp_toi + home_sh_toi

    if total_toi > 60:
        home_5v5_toi = home_5v5_toi - ((home_5v5_toi / total_toi) * (total_toi - 60))
        home_pp_toi = home_pp_toi - ((home_pp_toi / total_toi) * (total_toi - 60))
        home_sh_toi = home_sh_toi - ((home_sh_toi / total_toi) * (total_toi - 60))

    home_5v5_xgf_p60 = poisson.ppf(random_float(), game["predicted_home_5v5_xgf_p60"])
    home_5v5_gf_p60 = poisson.ppf(random_float(), game["predicted_home_5v5_gf_p60"])
    home_pp_xgf_p60 = poisson.ppf(random_float(), game["predicted_home_powerplay_xgf_p60"])
    home_pp_gf_p60 = poisson.ppf(random_float(), game["predicted_home_powerplay_gf_p60"])

    away_5v5_xgf_p60 = poisson.ppf(random_float(), game["predicted_away_5v5_xgf_p60"])
    away_5v5_gf_p60 = poisson.ppf(random_float(), game["predicted_away_5v5_gf_p60"])
    away_pp_xgf_p60 = poisson.ppf(random_float(), game["predicted_away_powerplay_xgf_p60"])
    away_pp_gf_p60 = poisson.ppf(random_float(), game["predicted_away_powerplay_gf_p60"])

    home_5v5_goals = home_5v5_xgf_p60 * (home_5v5_toi / 60)
    home_pp_goals = home_pp_xgf_p60 * (home_pp_toi / 60)
    home_total_goals = home_5v5_goals + home_pp_goals

    away_5v5_goals = away_5v5_xgf_p60 * (home_5v5_toi / 60)
    away_pp_goals = away_pp_xgf_p60 * (home_sh_toi / 60)
    away_total_goals = away_5v5_goals + away_pp_goals

    if home_total_goals > away_total_goals:
        home_win = 1
        away_win = 0
        draw = 0

    elif away_total_goals > home_total_goals:
        home_win = 0
        away_win = 1
        draw = 0

    else:
        home_win = 0
        away_win = 0
        draw = 1

    prediction.update(
        {
            "game_id": game["game_id"],
            "home_team": game["home_team"],
            "away_team": game["away_team"],
            "predicted_home_5v5_toi": home_5v5_toi,
            "predicted_home_powerplay_toi": home_pp_toi,
            "predicted_home_shorthanded_toi": home_sh_toi,
            "predicted_away_5v5_toi": home_5v5_toi,
            "predicted_away_powerplay_toi": home_sh_toi,
            "predicted_away_shorthanded_toi": home_pp_toi,
            "predicted_home_5v5_gf_p60": home_5v5_gf_p60,
            "predicted_home_5v5_xgf_p60": home_5v5_xgf_p60,
            "predicted_home_powerplay_gf_p60": home_pp_gf_p60,
            "predicted_home_powerplay_xgf_p60": home_pp_xgf_p60,
            "predicted_home_5v5_goals": home_5v5_goals,
            "predicted_home_powerplay_goals": home_pp_goals,
            "predicted_home_total_goals": home_total_goals,
            "predicted_away_5v5_gf_p60": away_5v5_gf_p60,
            "predicted_away_5v5_xgf_p60": away_5v5_xgf_p60,
            "predicted_away_powerplay_gf_p60": away_pp_gf_p60,
            "predicted_away_powerplay_xgf_p60": away_pp_xgf_p60,
            "predicted_away_5v5_goals": away_5v5_goals,
            "predicted_away_powerplay_goals": away_pp_goals,
            "predicted_away_total_goals": away_total_goals,
            "home_win": home_win,
            "away_win": away_win,
            "draw": draw,
        }
    )

    return prediction

game = todays_games.filter(pl.col("home_team") == "NSH").to_dicts()[0]

pl.DataFrame(simulate_game(game))  # Showing it as a dataframe just to make it easy

shape: (1, 26)

game_id	home_team	away_team	predicted_home_5v5_toi	predicted_home_powerplay_toi	predicted_home_shorthanded_toi	predicted_away_5v5_toi	predicted_away_powerplay_toi	predicted_away_shorthanded_toi	predicted_home_5v5_gf_p60	predicted_home_5v5_xgf_p60	predicted_home_powerplay_gf_p60	predicted_home_powerplay_xgf_p60	predicted_home_5v5_goals	predicted_home_powerplay_goals	predicted_home_total_goals	predicted_away_5v5_gf_p60	predicted_away_5v5_xgf_p60	predicted_away_powerplay_gf_p60	predicted_away_powerplay_xgf_p60	predicted_away_5v5_goals	predicted_away_powerplay_goals	predicted_away_total_goals	home_win	away_win	draw
i64	str	str	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	i64	i64	i64
2025020811	"NSH"	"UTA"	45.0	8.823529	6.176471	45.0	6.176471	8.823529	0.0	2.0	2.0	9.0	1.5	1.323529	2.823529	NaN	1.0	6.0	8.0	0.75	0.823529	1.573529	1	0	0

Predict an entire game

Predict an entire game based on n number of simulations

def predict_game(
    game: dict,
    total_simulations: int = 10_000,
    disable_progress_bar: bool = False,
    transient_progress_bar: bool = False,
    save: bool = False,
    overwrite: bool = False,
) -> pl.DataFrame:
    """Predict game based on n number of simulations."""
    predictions = []

    with ChickenProgress(disable=disable_progress_bar, transient=transient_progress_bar) as progress:
        pbar_message = f"Simulating {game['game_id']}..."
        simulation_task = progress.add_task(pbar_message, total=total_simulations)

        for sim_number in range(0, total_simulations):
            prediction = simulate_game(game=game)
            predictions.append(prediction)

            if sim_number == total_simulations - 1:
                pbar_message = f"Finished simulating {game['game_id']}"

            progress.update(simulation_task, description=pbar_message, advance=1, refresh=True)

    predictions = pl.DataFrame(predictions)

    if save:
        predictions_path = Path("./results/predictions.csv")

        if predictions_path.exists() and overwrite:
            saved_predictions = pl.read_csv(predictions_path, infer_schema_length=2000)

            predictions = pl.concat([saved_predictions, predictions], strict=False)

        predictions.write_csv(predictions_path)

    return predictions

predicted_game = predict_game(game, 100)

predicted_game.head(5)

shape: (5, 26)

game_id	home_team	away_team	predicted_home_5v5_toi	predicted_home_powerplay_toi	predicted_home_shorthanded_toi	predicted_away_5v5_toi	predicted_away_powerplay_toi	predicted_away_shorthanded_toi	predicted_home_5v5_gf_p60	predicted_home_5v5_xgf_p60	predicted_home_powerplay_gf_p60	predicted_home_powerplay_xgf_p60	predicted_home_5v5_goals	predicted_home_powerplay_goals	predicted_home_total_goals	predicted_away_5v5_gf_p60	predicted_away_5v5_xgf_p60	predicted_away_powerplay_gf_p60	predicted_away_powerplay_xgf_p60	predicted_away_5v5_goals	predicted_away_powerplay_goals	predicted_away_total_goals	home_win	away_win	draw
i64	str	str	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	i64	i64	i64
2025020811	"NSH"	"UTA"	50.0	1.0	6.0	50.0	6.0	1.0	0.0	6.0	1.0	5.0	5.0	0.083333	5.083333	NaN	2.0	15.0	8.0	1.666667	0.8	2.466667	1	0	0
2025020811	"NSH"	"UTA"	53.0	3.0	4.0	53.0	4.0	3.0	1.0	2.0	1.0	1.0	1.766667	0.05	1.816667	NaN	4.0	9.0	8.0	3.533333	0.533333	4.066667	0	1	0
2025020811	"NSH"	"UTA"	32.0	5.0	10.0	32.0	10.0	5.0	0.0	1.0	1.0	4.0	0.533333	0.333333	0.866667	NaN	2.0	5.0	4.0	1.066667	0.666667	1.733333	0	1	0
2025020811	"NSH"	"UTA"	45.0	2.0	4.0	45.0	4.0	2.0	0.0	6.0	3.0	4.0	4.5	0.133333	4.633333	NaN	4.0	11.0	6.0	3.0	0.4	3.4	1	0	0
2025020811	"NSH"	"UTA"	50.769231	5.538462	3.692308	50.769231	3.692308	5.538462	2.0	4.0	2.0	6.0	3.384615	0.553846	3.938462	NaN	2.0	11.0	4.0	1.692308	0.246154	1.938462	1	0	0

Simulating multiple games

Simulating multiple games at once

def predict_games(
    predict_game,
    todays_games: pl.DataFrame,
    n_workers: int = 6,
    total_simulations: int = 10_000,
    disable_progress_bar: bool = False,
    transient_progress_bar: bool = False,
    save: bool = False,
    overwrite: bool = False,
) -> pl.DataFrame:
    """Simulate today's games."""
    predictions_list = []

    with ThreadPoolExecutor(max_workers=n_workers) as executor:
        futures = [
            executor.submit(
                predict_game, game, total_simulations, disable_progress_bar, transient_progress_bar, save, overwrite
            )
            for game in todays_games.to_dicts()
        ]

        for future in concurrent.futures.as_completed(futures):
            predictions_list.append(future.result())

    return pl.concat(predictions_list)

predictions = predict_games(
    predict_game, todays_games, total_simulations=10_000, n_workers=6, disable_progress_bar=True
)

predictions.head(5)

shape: (5, 26)

game_id	home_team	away_team	predicted_home_5v5_toi	predicted_home_powerplay_toi	predicted_home_shorthanded_toi	predicted_away_5v5_toi	predicted_away_powerplay_toi	predicted_away_shorthanded_toi	predicted_home_5v5_gf_p60	predicted_home_5v5_xgf_p60	predicted_home_powerplay_gf_p60	predicted_home_powerplay_xgf_p60	predicted_home_5v5_goals	predicted_home_powerplay_goals	predicted_home_total_goals	predicted_away_5v5_gf_p60	predicted_away_5v5_xgf_p60	predicted_away_powerplay_gf_p60	predicted_away_powerplay_xgf_p60	predicted_away_5v5_goals	predicted_away_powerplay_goals	predicted_away_total_goals	home_win	away_win	draw
i64	str	str	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	i64	i64	i64
2025020815	"WPG"	"DET"	46.875	5.625	7.5	46.875	7.5	5.625	0.0	2.0	20.0	8.0	1.5625	0.75	2.3125	16.0	1.0	7.0	9.0	0.78125	1.125	1.90625	1	0	0
2025020815	"WPG"	"DET"	48.0	1.0	6.0	48.0	6.0	1.0	0.0	5.0	20.0	7.0	4.0	0.116667	4.116667	23.0	1.0	3.0	9.0	0.8	0.9	1.7	1	0	0
2025020815	"WPG"	"DET"	44.0	3.0	2.0	44.0	2.0	3.0	1.0	1.0	19.0	5.0	0.733333	0.25	0.983333	17.0	2.0	11.0	10.0	1.466667	0.333333	1.8	0	1	0
2025020815	"WPG"	"DET"	52.5	2.5	5.0	52.5	5.0	2.5	0.0	0.0	17.0	10.0	0.0	0.416667	0.416667	22.0	6.0	1.0	7.0	5.25	0.583333	5.833333	0	1	0
2025020815	"WPG"	"DET"	40.0	6.0	5.0	40.0	5.0	6.0	0.0	3.0	15.0	7.0	2.0	0.7	2.7	14.0	5.0	8.0	8.0	3.333333	0.666667	4.0	0	1	0

Assessing results

Check to see if the simulation is accurate

Process winners

Aggregate the predictions to “predict” a winner

def process_winners(predictions: pl.DataFrame, save: bool = False, overwrite: bool = True) -> pl.DataFrame:
    """Aggregate the predictions to "predict" a winner of the game."""
    group_list = ["game_id", "home_team", "away_team"]

    agg_stats_sum = ["home_win", "away_win", "draw"]
    agg_stats_mean = [x for x in predictions.columns if x not in group_list and x not in agg_stats_sum]

    agg_stats = [pl.col(x).sum().alias(f"predicted_{x}") for x in agg_stats_sum] + [
        pl.col(x).mean().alias(f"{x}_mean") for x in agg_stats_mean
    ]

    predicted_winners = predictions.group_by(group_list).agg(agg_stats)

    sum_stats = [f"predicted_{x}" for x in agg_stats_sum]
    add_stats = []
    for stat in agg_stats_sum:
        add_stat = (pl.col(f"predicted_{stat}") / pl.sum_horizontal(sum_stats)).alias(f"predicted_{stat}_percent")
        add_stats.append(add_stat)

    predicted_winners = predicted_winners.with_columns(add_stats)

    {x: f"predicted_{x}" for x in agg_stats_sum} | {x: f"{x}_mean" for x in agg_stats_mean}

    # predicted_winners = predicted_winners.rename(mapping=rename_columns)

    predicted_winners = predicted_winners.with_columns(
        predicted_winner=pl.when(pl.col("predicted_home_win_percent") > pl.col("predicted_away_win_percent"))
        .then(pl.col("home_team"))
        .otherwise(pl.col("away_team"))
    )

    columns = [
        "game_id",
        "home_team",
        "away_team",
        "predicted_winner",
        "predicted_home_win",
        "predicted_away_win",
        "predicted_draw",
        "predicted_home_win_percent",
        "predicted_away_win_percent",
        "predicted_draw_percent",
        "predicted_home_5v5_goals_mean",
        "predicted_home_powerplay_goals_mean",
        "predicted_home_total_goals_mean",
        "predicted_home_5v5_xgf_p60_mean",
        "predicted_home_powerplay_xgf_p60_mean",
        "predicted_home_5v5_toi_mean",
        "predicted_home_powerplay_toi_mean",
        "predicted_home_shorthanded_toi_mean",
        "predicted_away_5v5_goals_mean",
        "predicted_away_powerplay_goals_mean",
        "predicted_away_total_goals_mean",
        "predicted_away_5v5_xgf_p60_mean",
        "predicted_away_powerplay_xgf_p60_mean",
        "predicted_away_5v5_toi_mean",
        "predicted_away_powerplay_toi_mean",
        "predicted_away_shorthanded_toi_mean",
    ]

    predicted_winners = predicted_winners.select(columns)

    if save:
        predicted_winners_path = Path("./results/predicted_winners.csv")

        if predicted_winners_path.exists() and overwrite:
            saved_predicted_winners = pl.read_csv(predicted_winners_path, infer_schema_length=2000)

            predicted_winners = pl.concat([saved_predicted_winners, predicted_winners], strict=False)

        predicted_winners.write_csv(predicted_winners_path)

    return predicted_winners

predicted_winners = process_winners(predictions)

predicted_winners

shape: (9, 26)

game_id	home_team	away_team	predicted_winner	predicted_home_win	predicted_away_win	predicted_draw	predicted_home_win_percent	predicted_away_win_percent	predicted_draw_percent	predicted_home_5v5_goals_mean	predicted_home_powerplay_goals_mean	predicted_home_total_goals_mean	predicted_home_5v5_xgf_p60_mean	predicted_home_powerplay_xgf_p60_mean	predicted_home_5v5_toi_mean	predicted_home_powerplay_toi_mean	predicted_home_shorthanded_toi_mean	predicted_away_5v5_goals_mean	predicted_away_powerplay_goals_mean	predicted_away_total_goals_mean	predicted_away_5v5_xgf_p60_mean	predicted_away_powerplay_xgf_p60_mean	predicted_away_5v5_toi_mean	predicted_away_powerplay_toi_mean	predicted_away_shorthanded_toi_mean
i64	str	str	str	i64	i64	i64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64	f64
2025020817	"MIN"	"FLA"	"FLA"	4730	5210	60	0.473	0.521	0.006	2.043366	0.750291	2.793657	2.7232	6.7833	44.985984	6.615618	4.151098	2.081702	0.837106	2.918809	2.7769	12.1098	44.985984	4.151098	6.615618
2025020811	"NSH"	"UTA"	"UTA"	4263	5674	63	0.4263	0.5674	0.0063	1.987513	0.533736	2.521248	2.6498	6.0804	45.023345	5.261969	5.339087	2.2737	0.58656	2.860259	3.0309	6.578	45.023345	5.339087	5.261969
2025020815	"WPG"	"DET"	"DET"	3549	6416	35	0.3549	0.6416	0.0035	2.181233	0.574016	2.755249	2.796	9.0512	46.82656	3.800218	4.907622	2.585273	0.906052	3.491325	3.3141	11.0862	46.82656	4.907622	3.800218
2025020810	"NYI"	"BUF"	"BUF"	3923	6036	41	0.3923	0.6036	0.0041	2.218041	0.607624	2.825666	2.8193	8.6695	47.253475	4.206953	4.414072	2.498976	0.889559	3.388535	3.1732	12.0668	47.253475	4.414072	4.206953
2025020814	"CBJ"	"TBL"	"TBL"	3080	6874	46	0.308	0.6874	0.0046	1.904084	0.630926	2.53501	2.4384	9.0795	46.91877	4.17841	4.46062	2.788274	0.750063	3.538337	3.565	10.064	46.91877	4.46062	4.17841
2025020816	"STL"	"LAK"	"LAK"	4162	5773	65	0.4162	0.5773	0.0065	1.83382	0.409691	2.243511	2.3119	5.2265	47.610556	4.704324	4.02682	1.976607	0.615554	2.592161	2.4909	9.1846	47.610556	4.02682	4.704324
2025020813	"OTT"	"CAR"	"CAR"	4612	5337	51	0.4612	0.5337	0.0051	2.01902	0.5345	2.55352	2.6788	5.6709	45.282782	5.654618	5.145	2.099652	0.611163	2.710815	2.7795	7.1477	45.282782	5.145	5.654618
2025020818	"EDM"	"WSH"	"WSH"	4840	5124	36	0.484	0.5124	0.0036	2.179586	1.081675	3.261261	2.8582	12.5859	45.784646	5.162671	4.760983	2.393757	0.965454	3.359211	3.1407	12.1579	45.784646	4.760983	5.162671
2025020812	"BOS"	"MTL"	"MTL"	3998	5969	33	0.3998	0.5969	0.0033	1.94943	0.736576	2.686006	2.6237	8.6032	44.584212	5.145947	6.169241	2.004061	1.142854	3.146915	2.6999	11.1295	44.584212	6.169241	5.145947

Assessing predictions

Checking to see if the predictions were correct

def assess_predictions(
    predicted_winners: pl.DataFrame, schedule: pl.DataFrame, save: bool = False, overwrite: bool = True
) -> pl.DataFrame:
    """Takes the scheduled and checkes to see if the predicted results are corrected."""
    schedule = schedule.filter(pl.col("game_state") == "OFF")  # Getting only the finished games
    game_ids = schedule["game_id"].to_list()  # Taking the game IDs as a list

    home_win = pl.col("home_score") > pl.col("away_score")  # Condition to check if the home team won
    winners = schedule.select(
        pl.when(home_win).then(pl.col("home_team")).otherwise(pl.col("away_team")).alias("actual_winner")
    )["actual_winner"].to_list()  # Getting the winners as a list
    winners_dict = dict(zip(game_ids, winners, strict=False))  # Combining game IDs and winning teams as a dictionary

    predicted_winners = predicted_winners.with_columns(
        actual_winner=pl.col("game_id").replace_strict(winners_dict, return_dtype=pl.String)
    )

    predicted_winners = predicted_winners.with_columns(
        prediction_correct=pl.when(pl.col("actual_winner") == pl.col("predicted_winner"))
        .then(pl.lit(1))
        .otherwise(pl.lit(0))
    )

    columns = [
        "game_id",
        "home_team",
        "away_team",
        "predicted_winner",
        "actual_winner",
        "prediction_correct",
        "predicted_home_win",
        "predicted_away_win",
        "predicted_draw",
        "predicted_home_win_percent",
        "predicted_away_win_percent",
        "predicted_draw_percent",
        "predicted_home_5v5_goals_mean",
        "predicted_home_powerplay_goals_mean",
        "predicted_home_total_goals_mean",
        "predicted_home_5v5_xgf_p60_mean",
        "predicted_home_powerplay_xgf_p60_mean",
        "predicted_home_5v5_toi_mean",
        "predicted_home_powerplay_toi_mean",
        "predicted_home_shorthanded_toi_mean",
        "predicted_away_5v5_goals_mean",
        "predicted_away_powerplay_goals_mean",
        "predicted_away_total_goals_mean",
        "predicted_away_5v5_xgf_p60_mean",
        "predicted_away_powerplay_xgf_p60_mean",
        "predicted_away_5v5_toi_mean",
        "predicted_away_powerplay_toi_mean",
        "predicted_away_shorthanded_toi_mean",
    ]

    assessed_predicted_winners = predicted_winners.select(columns)

    if save:
        assessed_predicted_winners_path = Path("./results/predicted_winners_assessed.csv")

        if assessed_predicted_winners_path.exists() and overwrite:
            saved_assessed_predicted_winners = pl.read_csv(assessed_predicted_winners_path, infer_schema_length=2000)

            assessed_predicted_winners = pl.concat(
                [saved_assessed_predicted_winners, assessed_predicted_winners], strict=False
            )

        assessed_predicted_winners.write_csv(assessed_predicted_winners_path)

    return assessed_predicted_winners

assessed_predictions = assess_predictions(predicted_winners, schedule)

Charts

Plotting the results of your simulations

Color dictionary

Taking the NHL colors and making a dictionary out of them

color_dict = {
    key: (value["GOAL"] if value["GOAL"] != "#FFFFFF" else value["SHOT"]) for key, value in NHL_COLORS.items()
}

Plotting a single game

Plotting the results for a single game

dpi = 650
figsize = (8, 5)
nrows = 1  # len(assessed_predictions["home_team"].to_list())
ncols = 1

fig, ax = plt.subplots(nrows=nrows, ncols=ncols, figsize=figsize, dpi=dpi, sharex=True, sharey=True)

for idx, game in enumerate(
    assessed_predictions.filter(pl.col("home_team") == "NSH").sort("predicted_home_total_goals_mean").to_dicts()
):
    # ax = axs[idx]

    plot_predictions = predictions.filter(pl.col("game_id") == game["game_id"])

    venues = ["away", "home"]

    for venue in venues:
        sns.kdeplot(
            data=plot_predictions,
            x=plot_predictions[f"predicted_{venue}_total_goals"],
            bw_adjust=0.5,
            clip_on=False,
            fill=True,
            alpha=0.1,
            linewidth=1.5,
            hue=plot_predictions[f"{venue}_team"],
            palette=color_dict,
            legend=False,
            ax=ax,
        )

    sns.despine(left=True)

    ax.set(yticks=[], ylabel="", xlabel="predicted total goals")

    if idx != (nrows - 1):
        ax.set(xticks=[])

    # else:
    #     ax.set(xlabel="predicted total goals")

    ax.set_title(f"{game['home_team']} vs. {game['away_team']}")

fig.savefig(f"./charts/{latest_date}_NSH.png", dpi=650, bbox_inches="tight", facecolor="white")

Plotting today’s games

Plotting the predicted goals for all of today’s games

dpi = 500
figsize = (4, 12)
nrows = len(assessed_predictions["home_team"].to_list())
ncols = 1

fig, axs = plt.subplots(nrows=nrows, ncols=ncols, figsize=figsize, dpi=dpi, sharex=True, sharey=True)

for idx, game in enumerate(assessed_predictions.sort("predicted_home_total_goals_mean").to_dicts()):
    ax = axs[idx]

    plot_predictions = predictions.filter(pl.col("game_id") == game["game_id"])

    venues = ["away", "home"]

    for venue in venues:
        sns.kdeplot(
            data=plot_predictions,
            x=plot_predictions[f"predicted_{venue}_total_goals"],
            bw_adjust=0.5,
            clip_on=False,
            fill=True,
            alpha=0.1,
            linewidth=1.5,
            hue=plot_predictions[f"{venue}_team"],
            palette=color_dict,
            legend=False,
            ax=ax,
        )

    sns.despine(left=True)

    ax.set(yticks=[], ylabel="")

    if idx != (nrows - 1):
        ax.set(xticks=[])

    # else:
    #     ax.set(xlabel="predicted total goals")

    ax.set_title(f"{game['home_team']} vs. {game['away_team']}", y=0.8)

fig.savefig(f"./charts/{latest_date}.png", dpi=650, bbox_inches="tight", facecolor="white")