Project Luther A quantitative approach to casting decisions Sarah Cullem
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Project Luther A quantitative approach to casting decisions
Sarah Cullem
Create a system to select actors for new films based on their relative impact on the film’s potential domestic revenue
Objective:
Jason Schwartzman
Rachel McAdams
Justin Long
Dream Team 1
Zach Galifianakis
Scarlett Johansson
Jonah Hill
Dream Team 2
Jason Schwartzman
Rachel McAdams
Justin Long
Dream Team 1
Zach Galifianakis
Scarlett Johansson
Jonah Hill
Dream Team 2
? ? ?
? ? ?
Can we find a way to tie to our decision to the financial return of
our film?
PROCESS OVERVIEW• Scrape and clean data (Box Office Mojo & OMDb API)
• Select scoring method for actors in a film to include as a regression feature
• Select additional features for modeling
• Select best performing model based on test & train error
• Apply findings to selecting casting for films
Actor Scoring Example: Rachel McAdams
Rachel’s score for prediction is the average domestic gross
for every prior film
Film Scoring Example: The Family Stone
Score = 110.13 * 45.03 * 35.62 * 24.54 = 4334869
Log(Score) = 15.28
The log of the score showed the strongest relationship with Domestic Total Gross
The Family Stone Log(Product of Actor Scores)
Dom
esti
c To
tal G
ross
Product of Actor Scores
Log(Product Actor Scores)
Log(Product of Actor Scores)
Other features in the model
Film Budget (in Millions)
Theaters
Days in Release
Run Time in Minutes
Domestic Total Gross (in Millions)
0
1250
2500
3750
5000
1 2 3 4 5
MSE TrainMSE Test
Model Selection: Mean Squared Error
0
0.175
0.35
0.525
0.7
1 2 3 4 5
0.23
0.38
0.49
0.66 0.69
Model Selection: Adjusted R Squared
Coef
ficien
t Ban
ds
-3
0
3
6
9
12
7.89
4.592.32 1.89 1.89
Lower BoundCoefficientUpper Bound
Actor Scoring: Coefficients & ConfidenceP
-Valu
e
0.0
0.1
0.2
1 2 3 4 5
0 0.0010.08
0.18 0.17
= 1.89 + 0.56
+ 0.78 + 0.02Log(Product of Actor Scores) Budget
TheatersDays in Release
Note: the intercept in the model equation is -96.4
Domestic Gross
Model 4: Interpretation
= 1.89 + 0.56
+ 0.78 +
Every 100 added theaters adds $2M more revenue
Note: the intercept in the model equation is -96.4
+ 0.02Log(Product of Actor Scores) Budget
TheatersDays in Release
Domestic Gross
1.89 + 0.56
+ 0.78
Every 10 days more on the release adds $7.8M in revenue
=
Note: the intercept in the model equation is -96.4
+ 0.02Log(Product of Actor Scores) Budget
TheatersDays in Release
Domestic Gross
1.89 + 0.56
+ 0.78
Every $10M increase in budget adds $5.6M to revenue
=
Note: the intercept in the model equation is -96.4
+ 0.02Log(Product of Actor Scores) Budget
TheatersDays in Release
Domestic Gross
1.89 + 0.56
+ 0.78
=
Every 1% increase in actor scores adds ~$1.9M to revenue
Note: the intercept in the model equation is -96.4
+ 0.02Log(Product of Actor Scores) Budget
TheatersDays in Release
Domestic Gross
Domestic Gross
1.89 + 0.56
+ 0.78
=
Every 1% increase in actor scores adds ~$1.9M to revenue
Note: the intercept in the model equation is -96.4
+ 0.02Log(Product of Actor Scores) Budget
TheatersDays in Release
Maintain a scorecard with the latest revenue
score for each actor, updated as new films
are released
Jason Schwartzman
Rachel McAdams
Justin Long
Dream Team 1
Zach Galifianakis
Scarlett Johansson
Jonah Hill
Dream Team 2
Jason Schwartzman
Rachel McAdams
Justin Long
Dream Team 1
Zach Galifianakis
Scarlett Johansson
Jonah Hill
Dream Team 2
20.8 82.8 43.3
114.0 88.8 93.1
Jason Schwartzman
Rachel McAdams
Justin Long
Dream Team 1
Zach Galifianakis
Scarlett Johansson
Jonah Hill
Dream Team 2
20.8 82.8 43.3
114.0 88.8 93.1
$21.2M
$25.9M+22%
QUESTIONS
Related Documents
