Lap Times vs Rolling Average of Lap Times

[MooshuBeef]

Thanks everyone for the help, finally got to a notable place with using the API and learning Python :D

This compares lap times to a rolling average over the whole race, and allows for tuning of the "fast laps" to account for races like Monaco.

import fastf1.plotting as p
import seaborn as sns
import pandas as pd
from matplotlib import pyplot as plt
import csv

p.setup_mpl(misc_mpl_mods=False)

#What Race and Driver are we looking at?
driver = 'ALO'
year, grand_prix, session = 2023,'MONACO', 'R'

#set a threshold of fast laps for cleanliness
# Typically it's 1.1, but for messy races like Monaco 2023, 2.0 is needed
fastLapThreshold = 2

# Load the race session and create the plot
race = f.get_session(year, grand_prix, session)
race.load()
#fig, ax = plt.subplots(figsize=(8, 8))

# Get all the laps for a single driver
# Filter out slow laps (yellow flag, VSC, pitstops etc.) as they distort the graph axis
driver_laps = race.laps.pick_driver(driver).pick_quicklaps(fastLapThreshold).reset_index()

# Seaborn doesn't have proper timedelta support
# So we have to convert timedelta to float (in seconds)
driver_laps["LapTime(s)"] = driver_laps["LapTime"].dt.total_seconds()

# Calculate the running avg lap time, up to 100 laps (covers all grand prixs)
driver_laps["RunningAvgDriver"] = driver_laps["LapTime(s)"].rolling(window=100, min_periods=1).mean()


###############################################################################
# LET'S MAKE A GRAPH
###############################################################################
#ENABLE THIS FOR HORIZONTAL
fig, axes = plt.subplots(1,2 , figsize=(16, 8), sharey=True)

#THIS IS FOR VERTICAL PLOT STYLE
#fig, axes = plt.subplots(2,1 , figsize=(8, 12), sharex=True)


fig.suptitle(f"{driver} at {race.event.year} {race.event.EventName} - {race.name}"
#I enable the below when I want the sub header
#"\n"
#"\n"
#r"Lap Time Vs Rolling Average Lap Time"
)

# Lap Time goes to plot 1
sns.scatterplot(ax=axes[0], data=driver_laps,
                x="LapNumber",
                y="LapTime(s)",
                #ax=ax,
                hue="Compound",
                palette=p.COMPOUND_COLORS,
                s=100,
                linewidth=0,
                legend='auto')
axes[0].set_title("Lap Times")

# Rolling Average Lap Time to plot 2
sns.scatterplot(ax=axes[1], data=driver_laps,
                x="LapNumber",
                y="RunningAvgDriver",
                #ax=ax,
                hue="Compound",
                palette=p.COMPOUND_COLORS,
                s=80,
                linewidth=0,
                legend='auto')
axes[1].set_title("Rolling Avg Times")

#Add the grid lines and despine
axes[0].grid()
axes[1].grid()
sns.despine(left=True, bottom=True)

plt.tight_layout()
plt.show()

[Casper-Guo]

I have not seen this approach to analyzing dry/wet races before. Can you please explain a bit about the intuition behind this rolling average or how we can interpret your graphs?

[MooshuBeef]

It's not specifically to better analyze dry/wet races but a step towards the analysis I want to do: seeing how some drivers excel / fall behind even when on the same tire strategy, and seeing if there may be insights within each sector. Gonna see if rolling averages help me see anything in the future