Steelers Sack Analytics Guru

[NorthCoast]

According to the Depot, and confirmed by Kassim himself.

Well, that was shortlived.

Tomlin and his gut win out.

[Buzz]

No need to employ an advisor the coach never listens to.

[RuthlessBurgher]

Well, that was shortlived.

Karim Kassam has been the Steelers' Analytics & Football Research Coordinator since 2015.

He's apparently now taken a data science job outside of football.

He is being replaced by a new Analytics & Football Research Coordinator named Jay Whitmire.

[hawaiiansteel]

No need to employ an advisor the coach never listens to.

yup, Tomlin doesn’t need to have validation for things he already knows...

[AzStillers1989]

Karim Kassam has been the Steelers' Analytics & Football Research Coordinator since 2015.

He's apparently now taken a data science job outside of football.

He is being replaced by a new Analytics & Football Research Coordinator named Jay Whitmire.

Sweet, let’s go Jay Whitmire!

[Oviedo]

Karim Kassam has been the Steelers' Analytics & Football Research Coordinator since 2015.

He's apparently now taken a data science job outside of football.

He is being replaced by a new Analytics & Football Research Coordinator named Jay Whitmire.

So we actually haven't gotten rid of an analytics guy on staff? My, my how the piranhas swarm when they smell a chance to criticize

[Oh wow]

So we actually haven't gotten rid of an analytics guy on staff? My, my how the piranhas swarm when they smell a chance to criticize

I love it when folks tell on themselves.

[NorthCoast]

So we actually haven't gotten rid of an analytics guy on staff? My, my how the piranhas swarm when they smell a chance to criticize

More correctly, Whitmire was already working with Kassim. So they had two on staff and are down to one.

[Oh wow]

Wait. So we had 2 guys on staff since 2015 yet folks been bashing us for not having anyone on staff because Tomlin was arrogant and trusted his gut?

[NorthCoast]

You can't dispute Tomlin's attitude towards analytics:

""I got a lot of respect for analytics and numbers, but I'm not going to make judgments based on those numbers. The game is the game. It's an emotional one played by emotional and driven men. That's an element of the game you can't measure. Often times decisions such as that weight heavily into the equation." - Mike Tomlin

[Northern_Blitz]

You can't dispute Tomlin's attitude towards analytics:

""I got a lot of respect for analytics and numbers, but I'm not going to make judgments based on those numbers. The game is the game. It's an emotional one played by emotional and driven men. That's an element of the game you can't measure. Often times decisions such as that weight heavily into the equation." - Mike Tomlin

This is definitely a Tomlin-ism. I like analytics a lot and think they should be used in sports more than they are.

But, he's also right that whatever models the analytical people use don't capture everything. They also tend to give you averages over long time periods when coaches need to make decisions based on N = 1 in the moment type things.

So in a case where the models aren't predicting overwhelming differences, I think it's reasonable to "go with the gut" of a coach when they've proven they have been successful in the past.

My guess is that it doesn't make much of a difference is you're choosing the option with a 45% win percentage over 55%. It seems unlikely that models wouldn't have that level of uncertainty or that taking a wide average might miss something important about a particular individual match up or even the emotion of the game (although I tend not to buy the emotional part too much).

But if the model is telling you that one decision is 25% and the other is 75%, it's probably a good idea to take the 75%.

[RuthlessBurgher]

More correctly, Whitmire was already working with Kassim. So they had two on staff and are down to one.

Kind of like how Shaun Sarrett was already working with Mike Munchak.

When Munchak left for the Broncos on 1/14/19, Sarrett was promoted to fill that opening the same day (just like when Kassim left, his current in-house assistant Whitmire was promoted to fill that position right away).

However, we did not hire our new assistant OL coach Adrian Klemm from outside of the organization until 3 weeks later on 2/5/19.

I imagine that we will hire a new assistant analytics guru dude from outside the organization at some point, but it probably won't be for a couple of weeks at least, just like the outside hire of Klemm for assistant OL coach.

[NorthCoast]

This is definitely a Tomlin-ism. I like analytics a lot and think they should be used in sports more than they are.

But, he's also right that whatever models the analytical people use don't capture everything. They also tend to give you averages over long time periods when coaches need to make decisions based on N = 1 in the moment type things.

So in a case where the models aren't predicting overwhelming differences, I think it's reasonable to "go with the gut" of a coach when they've proven they have been successful in the past.

My guess is that it doesn't make much of a difference is you're choosing the option with a 45% win percentage over 55%. It seems unlikely that models wouldn't have that level of uncertainty or that taking a wide average might miss something important about a particular individual match up or even the emotion of the game (although I tend not to buy the emotional part too much).

But if the model is telling you that one decision is 25% and the other is 75%, it's probably a good idea to take the 75%.

Spot on comments NB. Few people talk about the uncertainty in these methods. And given some of the weak correlations that exist in sports, there is maybe a level of trust that will be difficult to gain.

My guess is that Tomlin being an intelligent coach, might look to analytics if he himself was not sure in a given situation. But I would further guess he doesn't feel himself in that position all that often.

[Captain Lemming]

You can't dispute Tomlin's attitude towards analytics:

""I got a lot of respect for analytics and numbers, but I'm not going to make judgments based on those numbers. The game is the game. It's an emotional one played by emotional and driven men. That's an element of the game you can't measure. Often times decisions such as that weight heavily into the equation." - Mike Tomlin

I like Tomlin you all know this.
That said, I do not agree with this.

I would suggest that the most "analytics" driven team in the league is the current dynasty.

I truly believe they are the one NFL team built on baseballs "Moneyball" principles, which are based entirely on analytics.

Tomlins biggest in game weakness as a coach is going on HIS emotions rather than data driven odds.

[NorthCoast]

I like Tomlin you all know this.
That said, I do not agree with this.

I would suggest that the most "analytics" driven team in the league is the current dynasty.

I truly believe they are the one NFL team built on baseballs "Moneyball" principles, which are based entirely on analytics.

Tomlins biggest in game weakness as a coach is going on HIS emotions rather than data driven odds.


I agree with the Captain. I like Tomlin as well but
it is his weakness and I think he struggles to acknowledge it.

[Northern_Blitz]

I like Tomlin you all know this.
That said, I do not agree with this.

I would suggest that the most "analytics" driven team in the league is the current dynasty.

I truly believe they are the one NFL team built on baseballs "Moneyball" principles, which are based entirely on analytics.

Tomlins biggest in game weakness as a coach is going on HIS emotions rather than data driven odds.

If I had to pick between extremes "pure emotion" vs. "pure analytics", I'd take pure analytics without question.

That said, I think that the analytical tools we have probably still neglect or fail to capture everything in the sport. So, I think coaches are forced to make lots of decisions based on emotion / gut instincts (unless you video tape your opponent's practices :p).

I also think that the NFL is probably the sport that is hardest to apply analytics because so much of success and failure is based on team co-ordination instead of individuals competing against each other.

I think the order of sports easiest to apply analytics to is something like:
- MLB (almost all individual competition)
- NBA (dominated by elite individuals and defenses are still more individual vs. individual)
- NHL (more free flowing, fewer set plays, hard to quantify issues like strength of competition vs. strength of line mates)
- NFL (the easy thing about the NFL is that every play is a set play. The hard thing is that defense and offense are super complex. It's hard to even grade defenses in coverage or O-linemen blocking because you never know their actual assignments if you don't know the play).

I think that the analytics can tell us how to improve win percentage for things like, never punt if you're past your own 40 or always go for 2. These are "analytical" in the same way the NBA used analytics to determine that it's actually a good idea to shoot threes. But these kinds of decisions get coaches fired, so they don't run teams like that in the NFL (and many here would **** on Tomlin if they did).

[hawaiiansteel]

Tomlin Says He’s Trying To Embrace Technology And Analytics

By Alex Kozora
Posted on June 19, 2019

It’s funny what you come across during the offseason. Two days ago, we’re discussing the Pittsburgh Steelers losing Karim Kassam, head of the team’s analytics department, and juxtaposing that with Mike Tomlin’s seemingly distaste for that kind of number crunching.

But perhaps Tomlin is changing his tune. Last week, we referenced a leadership interview he conducted with L3 Leadership. He covered a variety of topics but when asked about how he’s personally grown as a leader, he pivoted to a surprising point.

“I’m getting better in a lot of ways,” he said in the interview that we’ll include below. “The use of technology, the proper use of technology in my job. I think we all are faced with those challenges when we’re in a job for an extended period of time. How the job changes based on information, the speed in which you have, the information, the amount of information that you have. The amount of information and the speed of the information and the speed in which decisions need to be made on my job are very different than 13 years ago when I got this opportunity.”

Though the league is still archaic in many ways, hey, they still use fax machines, there’s no question technology has evolved. During games, no longer do teams snap a bunch of photos of the plays from that drive. It’s all immediately recorded and accessible by tablet on the sideline. Player data has boomed with everyone, starting with their Combine workout, wearing trackers that measure speed, steps, and a litany of other valuable information. Analytic departments have emerged when they would’ve been shunned by front offices when Tomlin was hired in 2007. And briefly, Tomlin seemed to acknowledge that fact.

“Analytics is something that’s fast evolving in our business.”

That was all he said on the topic but the “seemingly pro analytic” spot on the Tomlin Bingo card has been left unchecked for a long time. So it seems like there’s progress there in regards to him knowing its value.

While losing Kassam is a blow to that element, the Steelers will still have a dedicated analytics employee. Jay Whitmire was hired by the organization this past December and will take over Kassam’s role for 2019. Separately, but going back to the rapid evolution of technology, the team also hired a Virtual Reality intern to help all players but chiefly quarterbacks, especially backups who can simulate reps they miss out on during the regular season.

You can check out the interview below. We’ve linked it to begin at the point Tomlin talks about technology and analytics but the whole 12 minutes are worth watching.


https://steelersdepot.com/2019/06/tomlin-says-hes-trying-to-embrace-technology-and-analytics/

[Oh wow]

I like Tomlin you all know this.
That said, I do not agree with this.

I would suggest that the most "analytics" driven team in the league is the current dynasty.

I truly believe they are the one NFL team built on baseballs "Moneyball" principles, which are based entirely on analytics.

Tomlins biggest in game weakness as a coach is going on HIS emotions rather than data driven odds.
Did Billicheat use analytics when he went for it on 4th down inside his 40 vs the Colts when they ended up losing?

Did he use it when he didn’t call a timeout and played chicken with Pete Carrol in the SB?

I think those were gut decisions based on emotion and feel for the game.

Now maybe analytics come into play when it comes to being the GM but that’s not Tomlin’s job. IMO it’s much easier to influence and build a team when you are the HC and the GM. Players know they can be moved much quicker when the coach also controls the purse strings.

I think where the Pats excel (besides having Brady) is flexibility and a willingness to keep doing something until a team stops it.

Granted, I have no idea if that’s considered analytics or not.

[Northern_Blitz]

Did Billicheat use analytics when he went for it on 4th down inside his 40 vs the Colts when they ended up losing?

Did he use it when he didn’t call a timeout and played chicken with Pete Carrol in the SB?

I think those were gut decisions based on emotion and feel for the game.

Now maybe analytics come into play when it comes to being the GM but that’s not Tomlin’s job. IMO it’s much easier to influence and build a team when you are the HC and the GM. Players know they can be moved much quicker when the coach also controls the purse strings.

I think where the Pats excel (besides having Brady) is flexibility and a willingness to keep doing something until a team stops it.

Granted, I have no idea if that’s considered analytics or not.

My guess is that the 4th down call was what the analytics would have suggested.

Similarly, analytics probably also suggested that Carrol dialing up a passing play on the goal line was a good call.

I argued that analytics likely also suggested that we kick an onside kick (we did) that many here didn't like.

Going with the probabilities doesn't mean that you'll get the result you want. It means that in the long run, you should get favorable outcomes more often than not.

However, my guess is that (i) NFL seasons and (ii) most coaching careers aren't long enough to make up the "long run". I think this is a place where the Steelers general practice of stability of coaching should give us another advantage over other teams that make knee-jerk reactions when it comes to coaching staff.

[Oh wow]

My guess is that the 4th down call was what the analytics would have suggested.

Similarly, analytics probably also suggested that Carrol dialing up a passing play on the goal line was a good call.

I argued that analytics likely also suggested that we kick an onside kick (we did) that many here didn't like.

Going with the probabilities doesn't mean that you'll get the result you want. It means that in the long run, you should get favorable outcomes more often than not.

However, my guess is that (i) NFL seasons and (ii) most coaching careers aren't long enough to make up the "long run". I think this is a place where the Steelers general practice of stability of coaching should give us another advantage over other teams that make knee-jerk reactions when it comes to coaching staff.

Pretty sure analytics had us going for 2 more often that first year it was implemented.

I know Bill Simmons or one of these ESPN Page 2 guys would always bash teams for punting on 4th and 3 or less if the team was past it’s on 40.

Maybe that’s just odds but I think if we fully embraced analytics we probably wouldn’t like the decisions being made.

[NorthCoast]

The following was written after the Steelers beat the Browns 24-9 on Nov. 20th, 2016. In this case Tomlin's gut agreed with the analytics:


https://www.cmusportsanalytics.com/in-defense-of-mike-tomlin/

In Defense Of Mike Tomlin





CMU Sports Analytics Club (https://www.cmusportsanalytics.com/author/admin/) - November 21, 2016

Written by Adam Tucker (@adamtb182 (https://twitter.com/adamtb182))
Main Takeaways


Expected point gain of going for it on fourth down gives key insights as to why Mike Tomlin chose not to kick a field goal at the end of the half of yesterday’s game.
Under the lens of expected point gain, are NFL coaches really acting rationally to maximize points scored?

There are very few moments that bring a football fan to the edge of his or her seat like when a team goes for it on fourth down. The coach often gets praise for when this gutsy call pays off, but the same people who would praise the coach are also ready to lambast him for when it fails. Pittsburgh Steelers fans are no stranger to these feelings; Coach Mike Tomlin often goes for it on fourth down and attempts many two-point conversions. Although his decision yesterday to go for the touchdown twice at the end of the second quarter was not on fourth down, he still had to make many of the same considerations as a fourth down play because the drive would have been over if the play was unsuccessful.
Tomlin’s two options to score were to either attempt a field goal or go for it. In order to evaluate his decision making, I will use data from 2009-2015 NFL season and figures calculated by Konstantinos Pelechrinis in his paper “Decision Making in American Football: Evidence from 7 Years of NFL Data” (http://pitt.edu/~kpele/mlsa16.pdf) to evaluate success rates of field goals and attempted fourth down conversions, the costs of failing to convert and turning the ball over, and using the previous two metrics to calculate the net gain (or loss) of going for it on fourth down. Let us examine the effectiveness of a field goal first.
Field Goal Success Rate
As you can see in Figure 1, the success rate of a field goal decreases as distance increases. Not controlling for distance, the average success rate of a field goal attempt is 85.5%.[i] (https://www.cmusportsanalytics.com/in-defense-of-mike-tomlin/#_edn1) However, one can also observe that, for attempts near the goal line, the success rate is very close to 100% according to Figure 1, as was the case for the Steelers.
FIGURE 1: Probability of a successful field goal given distance (credit to Konstantinos Pelechrinis)
https://cmusportsanalytics.files.wordpress.com/2016/11/figure1.png
Fourth Down Conversion Success Rate
Likewise, fourth down conversion success rate decreases as yards to go increases as seen in Figure 2. The overall success rate of fourth down conversion is 73% when controlling for yards to go.[ii] (https://www.cmusportsanalytics.com/in-defense-of-mike-tomlin/#_edn2) However, 55% of the dataset for these situations occurs when the offense only has one yard to go, which was the case for the Steelers, and this success rate figure jumps up to an 89% in this situation.[iii] (https://www.cmusportsanalytics.com/in-defense-of-mike-tomlin/#_edn3)
FIGURE 2: Probability of a fourth down conversion given yards to go (credit to Konstantinos Pelechrinis)
https://cmusportsanalytics.files.wordpress.com/2016/11/figure2.png
Cost of Turning the Ball Over
The biggest downside of going for it on fourth down instead of attempting a field goal or punt is turning the ball over to the opponent and risking giving them a scoring drive. The following figure illustrates the probability of a drive ending in a TD, field goal, or a failure to score points as a function of drive starting position. These probabilities are important in the final calculation for the expected gain of going for it on fourth down.
FIGURE 3: Probability of the result of a drive given the beginning yard line (credit to Konstantinos Pelechrinis)
https://cmusportsanalytics.files.wordpress.com/2016/11/figure3.png
Expected Benefit of Going for It on Fourth and Conclusion
Using the above data, expected point gain of going for it on fourth down (E[P]) is calculated by E[P] = E[P+] - E[P-], where E[P+] is the expected point benefit of a successful fourth down conversion and E[P-] is the expected point cost of a failed fourth down conversion.[iv] (https://www.cmusportsanalytics.com/in-defense-of-mike-tomlin/#_edn4) E[P+] is a function of the fourth down conversion rate given a yard line, and E[P-] is a function of the success rate of a field goal and probability of an opponent’s drive leading to a touchdown or field goal given a yard line.[v] (https://www.cmusportsanalytics.com/in-defense-of-mike-tomlin/#_edn5) A more rigorous derivation can be found here (http://pitt.edu/~kpele/mlsa16.pdf). Using the above, a graph for E[P] can be created, as seen in Figure 4. In conclusion, at the goal line the value of E[P] is over 3, which is greater than the expected points from a field goal at the same yard line. Therefore, Mike Tomlin made the correct call twice yesterday (11/20/16) to attempt to convert for a touchdown instead of opting for an attempted field goal. However, more broadly, one can see that E[P] is positive for a majority of the field and leads to questions as to why coaches don’t go for it on fourth more often. The rationality of coaches and whether they seek to be averse to risk or maximize expected points is outside the scope of this paper, but something that could be researched further.
FIGURE 4: E[P] given distance to goal line when 4th down (credit to Konstantinos Pelechrinis)


https://cmusportsanalytics.files.wordpress.com/2016/11/figure4.png