180: The 10 MOST CLUTCH Quarterbacks in NFL History (Backed by Data Science)
Who’s the most clutch quarterback in NFL history — Tom Brady, Patrick Mahomes, Aaron Rodgers, or someone completely unexpected? We’ll use Python + Data Science to figure it out.
👉 Try Sphinx for free - https://www.sphinx.ai
⏱️ TIMESTAMPS
00:00 - Who’s the most clutch QB?
00:40 - Python + Sphinx AI: analyzing 1M NFL plays
02:00 - Defining “clutch” in football (data-driven approach)
03:15 - “TV Clutch” Top 10
07:50 - Using AI to processes play-by-play data
11:10 - Advanced Clutch Factor
17:00 - Advanced Top 10
24:30 - Build your own analysis
🔗 RESOURCES & LINKS
💌 Join 20k+ aspiring data analysts — https://www.datacareerjumpstart.com/newsletter
🎯 Free Training: How to Land Your First Data Job — https://www.datacareerjumpstart.com/training
👩💻 Accelerator Program: Data Analytics Accelerator — https://www.datacareerjumpstart.com/daa
💼 Interview Prep Tool: Interview Simulator — https://www.datacareerjumpstart.com/interviewsimulator
📱 CONNECT WITH AVERY
🎥 YouTube: @averysmith
🤝 LinkedIn: https://www.linkedin.com/in/averyjsmith
📸 Instagram: https://instagram.com/datacareerjumpstart
🎵 TikTok: https://www.tiktok.com/@verydata
💻 Website: https://www.datacareerjumpstart.com
📱 CONNECT WITH SPHINX
🐦Twitter/X - https://x.com/getsphinx
🔗Linkedin - https://www.linkedin.com/company/sphinx-ml/
Transcript
everyone thinks
2
:they know who the most clutch quarterback in NFL
3
:history is Tom Brady with his Super Bowl comebacks
4
:Josh Allen in the playoffs
5
:or Aaron Rodgers with impossible throws
6
:everyone has opinions but what does the data say
7
:well I'm a data scientist
8
:so I fed Python every play from the last 25 years
9
:over 1 million plays every pressure moment
10
:every game winning drive and then
11
:I let math and statistics decide
12
:who really performs the most
13
:when it matters the most and the most statistically
14
:clutch quarterback in NFL history is
15
:someone you definitely weren't expecting
16
:now analyzing
17
:patterns across 1 million football plays is not easy
18
:even with Python so I use today's sponsor
19
:Sphinx
20
:to help me process this massive dataset in Python
21
:to identify these statistical patterns that reveal
22
:who is the most clutch quarterback of all time
23
:more on Sphinx in a bit but before I reveal
24
:the most clutch quarterback of all time
25
:which I promise I will get to very soon
26
:let's start a timer for 30 seconds
27
:we have a huge problem what even is clutch
28
:it's pretty hard to define
29
:but we know it happens during crunch time right
30
:and that we can basically say
31
:is when the game is almost over and the score is close
32
:for now we'll say that
33
:clutch time is when the score is within one touchdown
34
:so that's like Seven Points
35
:and it's in the last eighth of the game
36
:or the last 7:30 of the fourth quarter
37
:or it's in overtime but even then
38
:how do we quantify clutch
39
:like numerically well
40
:we'll use two different definitions today
41
:No. 1 TV clutch and No. 2
42
:advanced analytics clutch
43
:let's go ahead and start with TV clutch
44
:coming in at No. 10 on TV Clutch is Josh Freeman
45
:surprisingly now
46
:I know he didn't have an incredible career
47
:but he definitely has some highlight moments
48
:now this one kind of surprised me
49
:I did not see Josh Freeman being on this list
50
:but he did have one really good season
51
:so maybe that's playing a big factor
52
:No.9 is the controversial Aaron Rodgers
53
:and this shouldn't be a huge surprise
54
:as Aaron has had a ton of clutch plays
55
:throughout his entire career
56
:specifically with the Green Bay Packers
57
:No.8 is Tony Romo yes
58
:that is the now lead NFL analyst for CBS
59
:who makes some very strange noises sometimes
60
:but he had some very clutch plays for the Cowboys
61
:back in the day coming in at No. 7 is Rich Gannon
62
:I'm going to be honest I had to look him up
63
:he's a little bit before my time
64
:but it looks like he had a little bit of a late
65
:career resurgence with the Oakland Raiders
66
:and ultimately
67
:won the league MVP and took them to the Super Bowl
68
:No.6 is disappointingly Deshaun Watson
69
:not much to say here so moving on to No. 5
70
:and it's the Greatest Show on Turf
71
:Mr Kurt Warner fantastic story of going undrafted
72
:to almost getting kicked out of the NFL
73
:two league MVP and Super Bowl champion
74
:No.4 is Jake Delhomme his season in 2,003
75
:LED the Carolina Panthers to the Super Bowl
76
:earning the team the nickname the Cardiac Cats
77
:due to the numerous
78
:game winning drives in the fourth quarter
79
:or overtime and those are all really things to Delhomme
80
:he set a franchise record
81
:with seven game winning drives
82
:in that 2003 season alone
83
:No. 3 is Andrew Luck and man
84
:I totally get why he retired early
85
:like I I make sense
86
:but I would have loved to see the rest of his career
87
:because he was absolutely incredible
88
:and was a very clutch coming in at No. 2
89
:it should really surprise no one
90
:it's Patrick Mahomes don't need to say much here
91
:he is very clutch
92
:he's been on an incredible run the last eight years
93
:leading to three Super Bowl wins
94
:in five Super Bowl appearances
95
:the only surprise here is he wasn't No. 1
96
:and lastly at No. 1
97
:the most clutch quarterback via TV is Tom Tua
98
:Tonga Vaiaoloa yes
99
:somehow Tua is the most clutch quarterback of all time
100
:it's not Tom Brady question Mark
101
:I don't get it if you know much about the NFL
102
:you'll know that this result
103
:it's kind of crazy
104
:Tua isn't exactly known as a clutch quarterback
105
:and honestly kind of has a reputation of being
106
:maybe the opposite of clutch
107
:so with that
108
:let's talk about how he came up with his top 10
109
:and the analytics that LED us to this point
110
:so what is TV clutch well
111
:it's a term that our AI data science co pilot
112
:Sphinx created for us
113
:it combines things that we can easily see on TV
114
:in terms of clutchness touchdowns
115
:interceptions completion percentages
116
:those types of things
117
:as well as how often they were in clutch situations
118
:so let's rewind for a second here
119
:and I'll tell you how we got this top 10
120
:and how you can replicate this exact same analysis
121
:on your own even if you're not a programmer
122
:you're not very technical at all
123
:and you don't know Python
124
:so I start out by using NFL versus
125
:Python package called NFL Data
126
:Pie to download the last 25 years of play by play data
127
:but instead of taking hours to understand how this API
128
:works and write all the code myself
129
:I just asked Sphinx to get the data for me
130
:in plain English
131
:Sphinx then went and read all of the API docs
132
:to understand how the API works
133
:and automatically wrote the code
134
:for me to get all that play by play data
135
:now could I have done that all on my own
136
:absolutely it just would have taken me a lot of time
137
:and I'm trying to pump out
138
:really high quality episodes
139
:for you guys
140
:so it was really nice to have a little bit of a co
141
:pilot to write this code for me
142
:it ended up retrieving the last 25 seasons
143
:which averages to around 45,000 plays per season
144
:which is a total of 1.18 million rows of data
145
:and it has 300 columns that let us know who is playing
146
:where the ball is what the result of the play and a lot
147
:lot more pretty sweet right
148
:awesome dataset well
149
:of course we actually aren't interested
150
:in all 1.18 million plays
151
:only the plays that fit the clutch criteria
152
:that we stated earlier
153
:which is basically the last 1/8 of a one score game
154
:and so all I needed to do was tell Sphinx that
155
:that's what I'm interested in doing
156
:and it will create this filter for me on my data
157
:so you notice I just spell it out in plain English
158
:and tell exactly what I want
159
:and the cool part is
160
:it's actually smart enough to find the right column
161
:names to do this filter
162
:as well as do things like check for missing data
163
:without me explicitly telling it to do so
164
:and finally it writes the code to do the filter
165
:and now we have a Python data frame
166
:with only the plays that fall into our clutch category
167
:next
168
:we want to analyze the quarterback part of our data
169
:because this is play by play data
170
:which has data about
171
:everyone and everything that's going on
172
:but we're really only interested
173
:in clutch quarterback performance
174
:and you'll see in the data
175
:that we definitely
176
:have a lot of quarterback names going on
177
:as well as whether it was a completed pass
178
:whether it was a touchdown
179
:or whether was an interception
180
:yards completed all that good stuff
181
:now this is where we can ask our co pilot
182
:Sphinx to analyze all the existing plays
183
:and using only touchdowns
184
:interceptions completion percentage
185
:yards and clutch attempts
186
:we can tell it to create a TV Clutch Factor score
187
:and ask it to rank by the most clutch quarterbacks
188
:based upon that score now Sphinx
189
:our copilot gets to work
190
:and first it does the aggregation
191
:for all of the quarterback pass plays
192
:and calculates all of the completions and touchdowns
193
:interceptions and the yards gained
194
:as a reminder our data was the play by play data
195
:not quarterback clutch season stats
196
:so this step is absolutely necessary
197
:to kind of clean the data
198
:aggregate it
199
:manipulate it in a way that makes it usable for us
200
:then this code calculates what's called a Z score
201
:for each one of those different metrics
202
:if you're unfamiliar with the Z score
203
:it's basically a measure
204
:of how many standard deviations
205
:a specific data point is
206
:away from the mean of that data set
207
:indicating its position within a distribution
208
:which is basically in layman's terms
209
:how much better or worse is a quarterback
210
:from the average quarterback
211
:based on these stats basically
212
:this is what you need to know
213
:a Z score of 0 means you're very average quarterback
214
:a Z score of 1 means you're above average
215
:and a Z score of negative 1 means you're below average
216
:for that particular stat
217
:our python code then creates a formula
218
:where the touchdown Z scores
219
:the yard Z scores
220
:and the completion percentage scores are all good
221
:and interceptions the scores are bad
222
:and out of the other end of this formula
223
:comes our top 10 list from earlier
224
:with the most clutch quarterbacks
225
:where Tua Tonga by Lowa
226
:somehow is the most clutch quarterback of all time
227
:now if this is your first time seeing Sphinx
228
:hopefully you realize how cool it is
229
:and how useful it can be
230
:it's basically a co pilot for anyone working with data
231
:it currently ships as a VS code extension
232
:that interfaces with Jupiter
233
:and other compatible notebooks
234
:it runs in your environments alongside you
235
:which first makes it safe and second
236
:makes it easy to use
237
:and it can access data through Python APIs
238
:or it even has MCP
239
:capabilities with things like snowflake
240
:or Databricks or Big Query or Salesforce
241
:whatever you're using
242
:and you can learn more about Sphinx
243
:and get started for free
244
:using the link in the description down below
245
:now once again if you know anything about football
246
:you'd really question how on earth
247
:can the current quarterback of the Miami Dolphins
248
:Tua Tonga Vealua
249
:be the most clutch quarterback of all time
250
:and is it like
251
:possible that Tony Romo can even be in the top 10
252
:and the answer to that is I don't know
253
:it's hard to know but we only really use basic stats
254
:like touchdowns and completions and interceptions
255
:earlier and that is a little bit basic
256
:the good news is that the NFL stat heads
257
:all of these NFL data analysts and data scientists
258
:and all these smart people
259
:have created other stats that are able to capture
260
:how unique and how clutch a play is
261
:and one of the things they created is called the win
262
:probability added or WPA for short
263
:and it's a bit hard to understand
264
:and it would be an entirely
265
:separate video to explain it in full
266
:but basically every play in a football game changes
267
:the chance that a team wins
268
:or loses but specifically wins
269
:WPA measures how much that play changes your odds
270
:so for example in 2024 the Washington Commanders
271
:were losing to the Chicago Bears
272
:15 to 12
273
:the commanders had the ball with two seconds left
274
:on their own 48 yard line
275
:they basically had a current chance of winning of 17%
276
:Jaden Daniels is playing quarterback and as you can see
277
:he drops back he's has to throw this
278
:now basically they have to score a touchdown
279
:so he has to throw it 60 yards 17% chance of winning
280
:they score a touchdown
281
:they win and you can kind of see that
282
:he has to run away from all these guys
283
:trying to sack him goes to the other side of the field
284
:he's now look at this
285
:he's at his 30 yard line he has to throw it 70 yards
286
:there's the ball being thrown
287
:it's in the air flies up
288
:you'll see that
289
:the ball pops out and is caught for a touchdown
290
:and Washington wins the game
291
:now
292
:that is a very high WPA because the time has expired
293
:there's no time left they'll now be at 18 points
294
:and they have beaten the bears pretty much
295
:no matter what happens cause there's no time left
296
:so obviously that play was a very high WPA
297
:now a WPA of 0
298
:would be a play that does not change
299
:the effect of the game at all
300
:for example most kickoffs that result in a touchback
301
:have a WPA of zero
302
:cause it's a standard play that happens multiple times
303
:the game and doesn't really affect what happens
304
:it's really just like okay
305
:the ball starts on
306
:like the 30 yard line or whatever it is
307
:a negative WPA
308
:would be when the offense makes a really
309
:costly mistake where they were going to win the game
310
:but now they're probably not
311
:for example this Josh Allen interception
312
:in this situation it is overtime
313
:the bills are playing the Minnesota Vikings
314
:the Vikings are winning 33 to 30
315
:but Buffalo the bills
316
:have the ball with Josh Allen as quarterback there
317
:is a minute left and they're on the 20 yard line
318
:so all Josh Allen has to do is score a touchdown
319
:and the game is over and the bills win
320
:and Josh Allen's really good
321
:so there is a current 74% chance of winning
322
:when this ball is snapped
323
:you'll see that Josh Allen gets the ball
324
:he looks he looks pump fakes throws interception
325
:the Vikings have the ball and they slide
326
:and now Vikings end the game
327
:because you only get one possession
328
:I think at this time during the NFL
329
:but even if that wasn't the case
330
:they could just kneel and the game would be all over
331
:so the game officially is over
332
:Josh Allen has thrown a game losing interception
333
:to make his team lose that would be extremely low
334
:uh WPA
335
:in fact they went from a 74% chance of winning to a 0%
336
:so that's negative point seven four not good
337
:now that we understand WPA
338
:we know that a high positive value means you are clutch
339
:and a low
340
:negative value means you are the opposite of clutch
341
:the play by play data that we actually downloaded
342
:has WPA for every single play
343
:which is really awesome for us
344
:because that means that we can ask Sphinx
345
:to find the 10 most clutch quarterbacks
346
:based on WPA so we can literally ask Sphinx you know
347
:find these 10 quarterbacks
348
:give us a total WPA an average WPA and sort the table
349
:descendingly by the total WPA
350
:and Sphinx will create its game plan
351
:over here on the left hand side
352
:not a football game plan but a data science game plan
353
:and then start writing the code itself to
354
:to do this analysis
355
:and finally it will give us our results
356
:so starting at No. 10 we have Aaron Rodgers
357
:once again he was No. 9 on our TV clutch list
358
:so this is not a huge surprise here
359
:anyway you slice it
360
:Aaron Rodgers was a pretty clutch quarterback
361
:is he now I don't know
362
:but back in the day he was
363
:No.9 is another current quarterback
364
:and that is Mr Cool Joe Burrow
365
:and there's a reason he has that name you guys
366
:this guy performs when it means the most
367
:and is able to make the big throws
368
:when the game is in his hands
369
:No.8
370
:is a name that I hadn't heard of in a really long time
371
:and it honestly really surprised me
372
:Carson Palmer if you're anything like me
373
:you haven't thought about that in a while
374
:but while he was playing
375
:he had some pretty clutch moments
376
:No.7 was on our previous list as well
377
:and it's Andrew Luck once again
378
:this guy was on the trajectory of becoming
379
:one of the best quarterbacks
380
:of all time and it's a shame
381
:we didn't get to watch the rest of his career
382
:No. 6 is Matthew Stafford
383
:also a current quarterback and it's well deserved
384
:he is one of my favorite clutch plays ever
385
:and he's really like
386
:done a lot of clutch things in his career
387
:both as the Lions quarterback
388
:and now as the Rams quarterback
389
:No. 5 is NFL legend Peyton Manning
390
:yes before he did all the broadcasting stuff
391
:he was a really good quarterback
392
:and made some absolutely clutch plays
393
:No. 4 is Patrick Mahomes
394
:and I was honestly kind of like
395
:I can't believe he's so low on this list
396
:but once again very clutch quarterback
397
:has been on an absolute tear with the Chiefs
398
:and has made a lot of the biggest throws
399
:in NFL history No.3 is Big Ben Roethlisberger
400
:a Steelers legend was in it forever
401
:made some great clutch plays
402
:No. 2 on this list actually kind of surprised me
403
:but it's kind of surprised
404
:we haven't seen his name so much at all anymore
405
:because he was such a good quarterback during his day
406
:and that is Drew Brees
407
:legend for the New Orleans Saints
408
:always seems to make really good completions
409
:like his percentage was always so high
410
:wasn't doing anything fancy ever
411
:but like pass after pass after pass
412
:he just like charged down the field
413
:and No. 1 the No. 1 clutch quarterback of all time
414
:as defined by total WPA across his entire career is
415
:drum roll please
416
:the goat himself
417
:Mister Tom Brady so in the end
418
:data science
419
:kind of already proved what we already knew
420
:deep down right there you have it
421
:the 10 most clutch quarterbacks
422
:as proven by data science
423
:now I challenge you to make your own clutch list
424
:especially if you hate this clutch list
425
:let me know in the comments down below if you hate it
426
:and why you think I'm wrong
427
:but you can create your own by downloading Sphinx
428
:using the link below
429
:and you can get my own Jupiter notebook
430
:and play around with my code template
431
:create something similar with football and actually
432
:you know change the parameters of clutch
433
:or use a different metric
434
:instead of WPA you can use something called EPA
435
:which is another one of those crazy NFL advanced stats
436
:or you can do something totally different
437
:and do it with the NBA or with soccer
438
:or with cricket or you don't have to do sports at all
439
:that is the beauty of something like Sphinx
440
:is it helps you create awesome analysis
441
:that would take you hours on your own
442
:and probably will still take you hours with Sphinx
443
:but a lot faster I'm excited to see what you make
444
:and I'm excited to see you
445
:hit the subscribe button as well right
446
:see you in the next one