Formula 1 - Predicting the winner of a race¶

NOTE: Due to featuretools' memory requirement, this notebook will not run on MyBinder.

In this notebook we will benchmark getML against featuretools to predict the winner of a Formula 1 race.

Summary:

Prediction type: Classification model
Domain: Sports
Prediction target: Win
Population size: 31578

Author: Dr. Patrick Urbanke

Background¶

We would like to develop a prediction model for Formula 1 races, that would allow us to predict the winner of a race before the race has started.

We use dataset of all Formula 1 races from 1950 to 2017. The dataset includes information such as the time taken in each lap, the time taken for pit stops, the performance in the qualifying rounds etc.

The dataset has been downloaded from the CTU Prague relational learning repository (Motl and Schulte, 2015).

We will benchmark getML 's feature learning algorithms against featuretools, an open-source implementation of the propositionalization algorithm, similar to getML's FastProp.

A web frontend for getML¶

The getML monitor is a frontend built to support your work with getML. The getML monitor displays information such as the imported data frames, trained pipelines and allows easy data and feature exploration. You can launch the getML monitor here.

Where is this running?¶

Your getML live session is running inside a docker container on mybinder.org, a service built by the Jupyter community and funded by Google Cloud, OVH, GESIS Notebooks and the Turing Institute. As it is a free service, this session will shut down after 10 minutes of inactivity.

Analysis¶

Let's get started with the analysis and set up your session:

import copy
import os
from urllib import request

import numpy as np
import pandas as pd
from IPython.display import Image
import matplotlib.pyplot as plt
plt.style.use('seaborn')
%matplotlib inline  

import featuretools
import getml

getml.engine.set_project('formula1')



Connected to project 'formula1'

1. Loading data¶

1.1 Download from source¶

We begin by downloading the data:

conn = getml.database.connect_mariadb(
    host="relational.fit.cvut.cz",
    dbname="ErgastF1",
    port=3306,
    user="guest",
    password="relational"
)

conn

Connection(conn_id='default',
           dbname='ErgastF1',
           dialect='mysql',
           host='relational.fit.cvut.cz',
           port=3306)

def load_if_needed(name):
    """
    Loads the data from the relational learning
    repository, if the data frame has not already
    been loaded.
    """
    if not getml.data.exists(name):
        data_frame = getml.data.DataFrame.from_db(
            name=name,
            table_name=name,
            conn=conn
        )
        data_frame.save()
    else:
        data_frame = getml.data.load_data_frame(name)
    return data_frame

driverStandings = load_if_needed("driverStandings")
drivers = load_if_needed("drivers")
lapTimes = load_if_needed("lapTimes")
pitStops = load_if_needed("pitStops")
races = load_if_needed("races")
qualifying = load_if_needed("qualifying")

driverStandings

drivers

lapTimes

pitStops

races

qualifying

1.2 Prepare data for getML¶

racesPd = races.to_pandas()
racesPd

We actually need some set-up, because the target variable is not readily available. The wins column in driverStandings is actually the accumulated number of wins over a year, but what we want is a boolean variable indicated whether someone has one a particular race or not.

driverStandingsPd = driverStandings.to_pandas()

driverStandingsPd = driverStandingsPd.merge(
    racesPd[["raceId", "year", "date", "round"]],
    on="raceId"
)

previousStanding = driverStandingsPd.merge(
    driverStandingsPd[["driverId", "year", "wins", "round"]],
    on=["driverId", "year"],
)

isPreviousRound = (previousStanding["round_x"] - previousStanding["round_y"] == 1.0)

previousStanding = previousStanding[isPreviousRound]

previousStanding["win"] = previousStanding["wins_x"] - previousStanding["wins_y"]

driverStandingsPd = driverStandingsPd.merge(
    previousStanding[["raceId", "driverId", "win"]],
    on=["raceId", "driverId"],
    how="left",
)

driverStandingsPd["win"] = [win if win == win else wins for win, wins in zip(driverStandingsPd["win"], driverStandingsPd["wins"])]

driver_standings = getml.data.DataFrame.from_pandas(driverStandingsPd, "driver_standings")

driver_standings

We also need to include the date of the race to lapTimes and pitStops, because we cannot use this data for the race we would like to predict. We can only take lap times and pit stops from previous races.

lapTimesPd = lapTimes.to_pandas()

lapTimesPd = lapTimesPd.merge(
    racesPd[["raceId", "date", "year"]],
    on="raceId"
)

lap_times = getml.data.DataFrame.from_pandas(lapTimesPd, "lap_times")

lap_times

pitStopsPd = pitStops.to_pandas()

pitStopsPd = pitStopsPd.merge(
    racesPd[["raceId", "date", "year"]],
    on="raceId"
)

pit_stops = getml.data.DataFrame.from_pandas(pitStopsPd, "pit_stops")

pit_stops

getML requires that we define roles for each of the columns.

driver_standings.set_role("win", getml.data.roles.target)
driver_standings.set_role(["raceId", "driverId", "year"], getml.data.roles.join_key)
driver_standings.set_role("position", getml.data.roles.numerical)
driver_standings.set_role("date", getml.data.roles.time_stamp)

driver_standings

drivers.set_role("driverId", getml.data.roles.join_key)
drivers.set_role(["nationality", "driverRef"], getml.data.roles.categorical)

drivers

lap_times.set_role(["raceId", "driverId", "year"], getml.data.roles.join_key)
lap_times.set_role(["lap", "milliseconds", "position"], getml.data.roles.numerical)
lap_times.set_role("date", getml.data.roles.time_stamp)

lap_times

pit_stops.set_role(["raceId", "driverId", "year"], getml.data.roles.join_key)
pit_stops.set_role(["lap", "milliseconds", "stop"], getml.data.roles.numerical)
pit_stops.set_role("date", getml.data.roles.time_stamp)

pit_stops

qualifying.set_role(["raceId", "driverId", "qualifyId"], getml.data.roles.join_key)
qualifying.set_role(["position", "number"], getml.data.roles.numerical)

qualifying

2. Predictive modeling¶

We loaded the data and defined the roles and units. Next, we create a getML pipeline for relational learning.

split = getml.data.split.random(train=0.8, test=0.2)
split

2.1 Define relational model¶

star_schema = getml.data.StarSchema(population=driver_standings.drop(["position"]), alias="population", split=split)

star_schema.join(
    driver_standings,
    on=["driverId"],
    time_stamps="date",
    horizon=getml.data.time.days(1),
    lagged_targets=True,
)

# We cannot use lap times for the race
# we would like to predict, so we set
# a non-zero horizon.
star_schema.join(
    lap_times,
    on=["driverId"],
    time_stamps="date",
    horizon=getml.data.time.days(1),
)

# We cannot use pit stops for the race
# we would like to predict, so we set
# a non-zero horizon.
star_schema.join(
    pit_stops,
    on=["driverId"],
    time_stamps="date",
    horizon=getml.data.time.days(1),
)

star_schema.join(
    qualifying,
    on=["driverId", "raceId"],
    relationship=getml.data.relationship.many_to_one,
)

star_schema.join(
    drivers,
    on=["driverId"],
    relationship=getml.data.relationship.many_to_one,
)

star_schema

2.2 getML pipeline¶

Set-up the feature learner & predictor

We use the relboost algorithms for this problem. Because of the large number of keywords, we regularize the model a bit by requiring a minimum support for the keywords (min_num_samples).

mapping = getml.preprocessors.Mapping()

fast_prop = getml.feature_learning.FastProp(
    loss_function=getml.feature_learning.loss_functions.CrossEntropyLoss,
    aggregation=getml.feature_learning.FastProp.agg_sets.All,
    num_threads=1,
)

relboost = getml.feature_learning.Relboost(
    loss_function=getml.feature_learning.loss_functions.CrossEntropyLoss,
    num_threads=1,
)

relmt = getml.feature_learning.RelMT(
    loss_function=getml.feature_learning.loss_functions.CrossEntropyLoss,
    num_threads=1,
)

predictor = getml.predictors.XGBoostClassifier(n_jobs=1)

Build the pipeline

pipe1 = getml.pipeline.Pipeline(
    tags=['fast_prop'],
    data_model=star_schema.data_model,
    preprocessors=[mapping],
    feature_learners=[fast_prop],
    predictors=[predictor],
    include_categorical=True,
)

pipe1

Pipeline(data_model='population',
         feature_learners=['FastProp'],
         feature_selectors=[],
         include_categorical=True,
         loss_function=None,
         peripheral=['driver_standings', 'drivers', 'lap_times', 'pit_stops', 'qualifying'],
         predictors=['XGBoostClassifier'],
         preprocessors=['Mapping'],
         share_selected_features=0.5,
         tags=['fast_prop'])

2.3 Model training¶

pipe1.check(star_schema.train)

Checking data model...


Staging...
[========================================] 100%

Preprocessing...
[========================================] 100%

Checking...
[========================================] 100%


INFO [FOREIGN KEYS NOT FOUND]: When joining POPULATION__STAGING_TABLE_1 and LAP_TIMES__STAGING_TABLE_3 over 'driverId' and 'driverId', there are no corresponding entries for 68.551028% of entries in 'driverId' in 'POPULATION__STAGING_TABLE_1'. You might want to double-check your join keys.
INFO [FOREIGN KEYS NOT FOUND]: When joining POPULATION__STAGING_TABLE_1 and PIT_STOPS__STAGING_TABLE_4 over 'driverId' and 'driverId', there are no corresponding entries for 82.527910% of entries in 'driverId' in 'POPULATION__STAGING_TABLE_1'. You might want to double-check your join keys.

pipe1.fit(star_schema.train)

Checking data model...


Staging...
[========================================] 100%


INFO [FOREIGN KEYS NOT FOUND]: When joining POPULATION__STAGING_TABLE_1 and LAP_TIMES__STAGING_TABLE_3 over 'driverId' and 'driverId', there are no corresponding entries for 68.551028% of entries in 'driverId' in 'POPULATION__STAGING_TABLE_1'. You might want to double-check your join keys.
INFO [FOREIGN KEYS NOT FOUND]: When joining POPULATION__STAGING_TABLE_1 and PIT_STOPS__STAGING_TABLE_4 over 'driverId' and 'driverId', there are no corresponding entries for 82.527910% of entries in 'driverId' in 'POPULATION__STAGING_TABLE_1'. You might want to double-check your join keys.


Staging...
[========================================] 100%

Preprocessing...
[========================================] 100%

FastProp: Trying 867 features...
[========================================] 100%

FastProp: Building features...
[========================================] 100%

XGBoost: Training as predictor...
[========================================] 100%


Trained pipeline.
Time taken: 0h:10m:31.904886

Pipeline(data_model='population',
         feature_learners=['FastProp'],
         feature_selectors=[],
         include_categorical=True,
         loss_function=None,
         peripheral=['driver_standings', 'drivers', 'lap_times', 'pit_stops', 'qualifying'],
         predictors=['XGBoostClassifier'],
         preprocessors=['Mapping'],
         share_selected_features=0.5,
         tags=['fast_prop', 'container-iDDODO'])

url: http://localhost:1709/#/getpipeline/formula1/sSpaJ1/0/

2.4 Model evaluation¶

pipe1.score(star_schema.test)


Staging...
[========================================] 100%

Preprocessing...
[========================================] 100%

FastProp: Building features...
[========================================] 100%

2.5 featuretools¶

featuretools requires some additional data preparation before we can start engineering features.

population_train_pd = star_schema.train.population.drop(["position"]).to_pandas()
population_test_pd = star_schema.test.population.drop(["position"]).to_pandas()

driver_standings_pd = driver_standings.drop(driver_standings.roles.unused).to_pandas()
lap_times_pd = lap_times.drop(lap_times.roles.unused).to_pandas()
pit_stops_pd = pit_stops.drop(pit_stops.roles.unused).to_pandas()
qualifying_pd = qualifying.drop(qualifying.roles.unused).to_pandas()
drivers_pd = drivers.drop(drivers.roles.unused).to_pandas()

Because qualifying and drivers are many-to-one joins, we can directly join them onto our table.

population_train_pd["id"] = population_train_pd.index

population_train_pd = population_train_pd.merge(
    qualifying_pd,
    on=["driverId", "raceId"],
    how="left",
)

population_train_pd = population_train_pd.merge(
    drivers_pd,
    on=["driverId"],
    how="left",
)

population_train_pd

Same for the testing set.

population_test_pd["id"] = population_test_pd.index

population_test_pd = population_test_pd.merge(
    qualifying_pd,
    on=["driverId", "raceId"],
    how="left",
)

population_test_pd = population_test_pd.merge(
    drivers_pd,
    on=["driverId"],
    how="left",
)

population_test_pd

featuretools requires us to expand our peripheral tables so they can be joined using the unique id from the population table. Luckily, we can write a simple helper function that works for all three remaining peripheral tables.

def prepare_peripheral(peripheral, train_or_test):
    """
    Helper function that imitates the behavior of 
    the data model defined above.
    """
    peripheral_new = peripheral.merge(
        train_or_test[["id", "driverId", "date"]],
        on=["driverId"],
    )

    peripheral_new = peripheral_new[
        peripheral_new["date_x"] < peripheral_new["date_y"]
    ]
    
    del peripheral_new["date_y"]
    del peripheral_new["driverId"]
    del peripheral_new["raceId"]

    return peripheral_new.rename(columns={"date_x": "date"})

driver_standings_train_pd = prepare_peripheral(driver_standings_pd, population_train_pd)
driver_standings_test_pd = prepare_peripheral(driver_standings_pd, population_test_pd)
driver_standings_train_pd

lap_times_train_pd = prepare_peripheral(lap_times_pd, population_train_pd)
lap_times_test_pd = prepare_peripheral(lap_times_pd, population_test_pd)
lap_times_train_pd

lap_times demonstrates one of the greatest dangers of featuretools. Because it is written in pure Python, featuretools requires you expand your tables so that they can be more easily joined. But this comes at the cost of increased memory consumption: In this case a table that used to have about 420,000 rows has now been expanded to over 25 million rows.

pit_stops_train_pd = prepare_peripheral(pit_stops_pd, population_train_pd)
pit_stops_test_pd = prepare_peripheral(pit_stops_pd, population_test_pd)
pit_stops_train_pd

del population_train_pd["driverId"]
del population_train_pd["raceId"]
del population_train_pd["year"]
del population_train_pd["qualifyId"]

del population_test_pd["driverId"]
del population_test_pd["raceId"]
del population_test_pd["year"]
del population_test_pd["qualifyId"]

entities_train = {
    "population" : (population_train_pd, "id"),
    "driver_standings": (driver_standings_train_pd, "index"),
    "lap_times" : (lap_times_train_pd, "index"),
    "pit_stops" : (pit_stops_train_pd, "index"),
}

entities_test = {
    "population" : (population_test_pd, "id"),
    "driver_standings": (driver_standings_test_pd, "index"),
    "lap_times" : (lap_times_test_pd, "index"),
    "pit_stops" : (pit_stops_test_pd, "index")
}

relationships = [
    ("population", "id", "driver_standings", "id"),
    ("population", "id", "lap_times", "id"),
    ("population", "id", "pit_stops", "id")
]

featuretools_train_pd = featuretools.dfs(
    entities=entities_train,
    relationships=relationships,
    target_entity="population")[0]

/home/patrick/.local/lib/python3.9/site-packages/featuretools/entityset/entity.py:462: UserWarning: index index not found in dataframe, creating new integer column
  warnings.warn("index {} not found in dataframe, creating new "

featuretools_test_pd = featuretools.dfs(
    entities=entities_test,
    relationships=relationships,
    target_entity="population")[0]

featuretools_train = getml.data.DataFrame.from_pandas(featuretools_train_pd, "featuretools_train")
featuretools_test = getml.data.DataFrame.from_pandas(featuretools_test_pd, "featuretools_test")

featuretools_train.set_role("win", getml.data.roles.target)
featuretools_train.set_role(featuretools_train.roles.unused_float, getml.data.roles.numerical)
featuretools_train.set_role(featuretools_train.roles.unused_string, getml.data.roles.categorical)

featuretools_train

90.725472% of all entries of column 'MAX(pit_stops.lap)' are NULL values.
90.725472% of all entries of column 'MAX(pit_stops.milliseconds)' are NULL values.
90.725472% of all entries of column 'MAX(pit_stops.stop)' are NULL values.
90.725472% of all entries of column 'MEAN(pit_stops.lap)' are NULL values.
90.725472% of all entries of column 'MEAN(pit_stops.milliseconds)' are NULL values.
90.725472% of all entries of column 'MEAN(pit_stops.stop)' are NULL values.
90.725472% of all entries of column 'MIN(pit_stops.lap)' are NULL values.
90.725472% of all entries of column 'MIN(pit_stops.milliseconds)' are NULL values.
90.725472% of all entries of column 'MIN(pit_stops.stop)' are NULL values.
90.725472% of all entries of column 'NUM_UNIQUE(pit_stops.year)' are NULL values.
90.934554% of all entries of column 'SKEW(pit_stops.lap)' are NULL values.
90.934554% of all entries of column 'SKEW(pit_stops.milliseconds)' are NULL values.
90.934554% of all entries of column 'SKEW(pit_stops.stop)' are NULL values.
90.757032% of all entries of column 'STD(pit_stops.lap)' are NULL values.
90.757032% of all entries of column 'STD(pit_stops.milliseconds)' are NULL values.
90.757032% of all entries of column 'STD(pit_stops.stop)' are NULL values.
90.725472% of all entries of column 'MODE(pit_stops.DAY(date))' are NULL values.
90.725472% of all entries of column 'MODE(pit_stops.MONTH(date))' are NULL values.
90.725472% of all entries of column 'MODE(pit_stops.WEEKDAY(date))' are NULL values.
90.725472% of all entries of column 'MODE(pit_stops.YEAR(date))' are NULL values.
90.725472% of all entries of column 'NUM_UNIQUE(pit_stops.DAY(date))' are NULL values.
90.725472% of all entries of column 'NUM_UNIQUE(pit_stops.MONTH(date))' are NULL values.
90.725472% of all entries of column 'NUM_UNIQUE(pit_stops.WEEKDAY(date))' are NULL values.
90.725472% of all entries of column 'NUM_UNIQUE(pit_stops.YEAR(date))' are NULL values.
90.725472% of all entries of column 'MODE(pit_stops.year)' are NULL values.

featuretools_test.set_role("win", getml.data.roles.target)
featuretools_test.set_role(featuretools_test.roles.unused_float, getml.data.roles.numerical)
featuretools_test.set_role(featuretools_test.roles.unused_string, getml.data.roles.categorical)

featuretools_test

90.191042% of all entries of column 'MAX(pit_stops.lap)' are NULL values.
90.191042% of all entries of column 'MAX(pit_stops.milliseconds)' are NULL values.
90.191042% of all entries of column 'MAX(pit_stops.stop)' are NULL values.
90.191042% of all entries of column 'MEAN(pit_stops.lap)' are NULL values.
90.191042% of all entries of column 'MEAN(pit_stops.milliseconds)' are NULL values.
90.191042% of all entries of column 'MEAN(pit_stops.stop)' are NULL values.
90.191042% of all entries of column 'MIN(pit_stops.lap)' are NULL values.
90.191042% of all entries of column 'MIN(pit_stops.milliseconds)' are NULL values.
90.191042% of all entries of column 'MIN(pit_stops.stop)' are NULL values.
90.191042% of all entries of column 'NUM_UNIQUE(pit_stops.year)' are NULL values.
90.351581% of all entries of column 'SKEW(pit_stops.lap)' are NULL values.
90.351581% of all entries of column 'SKEW(pit_stops.milliseconds)' are NULL values.
90.351581% of all entries of column 'SKEW(pit_stops.stop)' are NULL values.
90.239204% of all entries of column 'STD(pit_stops.lap)' are NULL values.
90.239204% of all entries of column 'STD(pit_stops.milliseconds)' are NULL values.
90.239204% of all entries of column 'STD(pit_stops.stop)' are NULL values.
90.191042% of all entries of column 'MODE(pit_stops.DAY(date))' are NULL values.
90.191042% of all entries of column 'MODE(pit_stops.MONTH(date))' are NULL values.
90.191042% of all entries of column 'MODE(pit_stops.WEEKDAY(date))' are NULL values.
90.191042% of all entries of column 'MODE(pit_stops.YEAR(date))' are NULL values.
90.191042% of all entries of column 'NUM_UNIQUE(pit_stops.DAY(date))' are NULL values.
90.191042% of all entries of column 'NUM_UNIQUE(pit_stops.MONTH(date))' are NULL values.
90.191042% of all entries of column 'NUM_UNIQUE(pit_stops.WEEKDAY(date))' are NULL values.
90.191042% of all entries of column 'NUM_UNIQUE(pit_stops.YEAR(date))' are NULL values.
90.191042% of all entries of column 'MODE(pit_stops.year)' are NULL values.

We train an untuned XGBoostRegressor on top of featuretools' features, just like we have done for getML's features.

Since some of featuretools features are categorical, we allow the pipeline to include these features as well. Other features contain NaN values, which is why we also apply getML's Imputation preprocessor.

imputation = getml.preprocessors.Imputation()

predictor = getml.predictors.XGBoostClassifier(n_jobs=1)

pipe2 = getml.pipeline.Pipeline(
    tags=['featuretools'],
    preprocessors=[imputation],
    predictors=[predictor],
    include_categorical=True,
)

pipe2

Pipeline(data_model='population',
         feature_learners=[],
         feature_selectors=[],
         include_categorical=True,
         loss_function=None,
         peripheral=[],
         predictors=['XGBoostClassifier'],
         preprocessors=['Imputation'],
         share_selected_features=0.5,
         tags=['featuretools'])

pipe2.fit(featuretools_train)

Checking data model...


Staging...
[========================================] 100%

Preprocessing...
[========================================] 100%

Checking...
[========================================] 100%


WARNING [COLUMN SHOULD BE UNUSED]: 90.725472% of all entries in column 'mode(pit_stops.year)' in POPULATION__STAGING_TABLE_1 are NULL values. You should consider setting its role to unused_string.
WARNING [COLUMN SHOULD BE UNUSED]: All non-NULL entries in column 'COALESCE( min(lap_times.lap), 1.000000 )' in POPULATION__STAGING_TABLE_1 are equal to each other. You should consider setting its role to unused_float or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
WARNING [COLUMN SHOULD BE UNUSED]: All non-NULL entries in column 'COALESCE( min(pit_stops.stop), 1.000000 )' in POPULATION__STAGING_TABLE_1 are equal to each other. You should consider setting its role to unused_float or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
WARNING [COLUMN SHOULD BE UNUSED]: All non-NULL entries in column 'COALESCE( mode(lap_times.weekday(date)), 6.000000 )' in POPULATION__STAGING_TABLE_1 are equal to each other. You should consider setting its role to unused_float or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
WARNING [COLUMN SHOULD BE UNUSED]: All non-NULL entries in column 'COALESCE( num_unique(lap_times.weekday(date)), 1.000000 )' in POPULATION__STAGING_TABLE_1 are equal to each other. You should consider setting its role to unused_float or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
WARNING [COLUMN SHOULD BE UNUSED]: All non-NULL entries in column 'COALESCE( mode(pit_stops.weekday(date)), 6.000000 )' in POPULATION__STAGING_TABLE_1 are equal to each other. You should consider setting its role to unused_float or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').
WARNING [COLUMN SHOULD BE UNUSED]: All non-NULL entries in column 'COALESCE( num_unique(pit_stops.weekday(date)), 1.000000 )' in POPULATION__STAGING_TABLE_1 are equal to each other. You should consider setting its role to unused_float or using it for comparison only (you can do the latter by setting a unit that contains 'comparison only').


Staging...
[========================================] 100%

Preprocessing...
[========================================] 100%

XGBoost: Training as predictor...
[========================================] 100%


Trained pipeline.
Time taken: 0h:0m:7.319103

Pipeline(data_model='population',
         feature_learners=[],
         feature_selectors=[],
         include_categorical=True,
         loss_function=None,
         peripheral=[],
         predictors=['XGBoostClassifier'],
         preprocessors=['Imputation'],
         share_selected_features=0.5,
         tags=['featuretools'])

url: http://localhost:1709/#/getpipeline/formula1/DDRwHL/0/

pipe2.score(featuretools_test)


Staging...
[========================================] 100%

Preprocessing...
[========================================] 100%

2.6 Studying features¶

We would like to understand why getML outperforms featuretools. In particular, getML's FastProp is based on an approach that is very similar to featuretools.

To investigate this matter, we first take a look at the importance of the features FastProp has learned:

names, importances = pipe1.features.importances(target_num=0)

plt.subplots(figsize=(20, 10))

plt.bar(names

name	driverStandingsId	raceId	driverId	points	position	wins	positionText
role	unused_float	unused_float	unused_float	unused_float	unused_float	unused_float	unused_string
0	1	18	1	10	1	1	1
1	2	18	2	8	2	0	2
2	3	18	3	6	3	0	3
3	4	18	4	5	4	0	4
4	5	18	5	4	5	0	5
	...	...	...	...	...	...	...
31573	68456	982	835	8	16	0	16
31574	68457	982	154	26	13	0	13
31575	68458	982	836	5	18	0	18
31576	68459	982	18	0	22	0	22
31577	68460	982	814	0	23	0	23

name	driverId	number	driverRef	code	forename	surname	dob	nationality	url
role	unused_float	unused_float	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string	unused_string
0	1	44	hamilton	HAM	Lewis	Hamilton	1985-01-07	British	http://en.wikipedia.org/wiki/Lew...
1	2	nan	heidfeld	HEI	Nick	Heidfeld	1977-05-10	German	http://en.wikipedia.org/wiki/Nic...
2	3	6	rosberg	ROS	Nico	Rosberg	1985-06-27	German	http://en.wikipedia.org/wiki/Nic...
3	4	14	alonso	ALO	Fernando	Alonso	1981-07-29	Spanish	http://en.wikipedia.org/wiki/Fer...
4	5	nan	kovalainen	KOV	Heikki	Kovalainen	1981-10-19	Finnish	http://en.wikipedia.org/wiki/Hei...
	...	...	...	...	...	...	...	...	...
835	837	88	haryanto	HAR	Rio	Haryanto	1993-01-22	Indonesian	http://en.wikipedia.org/wiki/Rio...
836	838	2	vandoorne	VAN	Stoffel	Vandoorne	1992-03-26	Belgian	http://en.wikipedia.org/wiki/Sto...
837	839	31	ocon	OCO	Esteban	Ocon	1996-09-17	French	http://en.wikipedia.org/wiki/Est...
838	840	18	stroll	STR	Lance	Stroll	1998-10-29	Canadian	http://en.wikipedia.org/wiki/Lan...
839	841	36	giovinazzi	GIO	Antonio	Giovinazzi	1993-12-14	Italian	http://en.wikipedia.org/wiki/Ant...

name	raceId	driverId	lap	position	milliseconds	time
role	unused_float	unused_float	unused_float	unused_float	unused_float	unused_string
0	1	1	1	13	109088	1:49.088
1	1	1	2	12	93740	1:33.740
2	1	1	3	11	91600	1:31.600
3	1	1	4	10	91067	1:31.067
4	1	1	5	10	92129	1:32.129
	...	...	...	...	...	...
420364	982	840	54	8	107528	1:47.528
420365	982	840	55	8	107512	1:47.512
420366	982	840	56	8	108143	1:48.143
420367	982	840	57	8	107848	1:47.848
420368	982	840	58	8	108699	1:48.699

name	raceId	driverId	stop	lap	milliseconds	time	duration
role	unused_float	unused_float	unused_float	unused_float	unused_float	unused_string	unused_string
0	841	1	1	16	23227	17:28:24	23.227
1	841	1	2	36	23199	17:59:29	23.199
2	841	2	1	15	22994	17:27:41	22.994
3	841	2	2	30	25098	17:51:32	25.098
4	841	3	1	16	23716	17:29:00	23.716
	...	...	...	...	...	...	...
6065	982	839	6	38	29134	21:29:07	29.134
6066	982	840	1	1	37403	20:06:43	37.403
6067	982	840	2	2	29294	20:10:07	29.294
6068	982	840	3	3	25584	20:13:16	25.584
6069	982	840	4	26	29412	21:05:07	29.412

name	raceId	year	round	circuitId	name	date	time	url
role	unused_float	unused_float	unused_float	unused_float	unused_string	unused_string	unused_string	unused_string
0	1	2009	1	1	Australian Grand Prix	2009-03-29	06:00:00	http://en.wikipedia.org/wiki/200...
1	2	2009	2	2	Malaysian Grand Prix	2009-04-05	09:00:00	http://en.wikipedia.org/wiki/200...
2	3	2009	3	17	Chinese Grand Prix	2009-04-19	07:00:00	http://en.wikipedia.org/wiki/200...
3	4	2009	4	3	Bahrain Grand Prix	2009-04-26	12:00:00	http://en.wikipedia.org/wiki/200...
4	5	2009	5	4	Spanish Grand Prix	2009-05-10	12:00:00	http://en.wikipedia.org/wiki/200...
	...	...	...	...	...	...	...	...
971	984	2017	16	22	Japanese Grand Prix	2017-10-08	05:00:00	https://en.wikipedia.org/wiki/20...
972	985	2017	17	69	United States Grand Prix	2017-10-22	19:00:00	https://en.wikipedia.org/wiki/20...
973	986	2017	18	32	Mexican Grand Prix	2017-10-29	19:00:00	https://en.wikipedia.org/wiki/20...
974	987	2017	19	18	Brazilian Grand Prix	2017-11-12	16:00:00	https://en.wikipedia.org/wiki/20...
975	988	2017	20	24	Abu Dhabi Grand Prix	2017-11-26	17:00:00	https://en.wikipedia.org/wiki/20...