03-25-2020, 11:04 AM
(This post was last modified: 03-25-2020, 11:14 AM by iStegosauruz.)
[div align=\\\"center\\\"]Background and Methodology – [/div]
I simulated 14,400 games so that you wouldn’t have to.
Before I get too deep into the weeds about this study I want to shout-out my Birddogs teammate @Forty Jordy. The core idea behind this study was his, I just executed it.
Forty has wondered a while whether penalties play a role in home field advantage. He has been tracking penalties on the year and has also found that offensive line players have about 50 penalties as a position group on the year while all other position groups peak around 20. Essentially, if penalties play a role in your ability to win on the road and you can reduce the amount of penalties coming from your offensive linemen then you should be able to increase the amount of games you win on the road.
I realized I had a good amount of this work already done. I had simulated 50,422 games in the last two weeks looking at the differences between human offensive linemen and bot offensive linemen. In doing all of those simulations I would not only have data for a team’s penalties at home and on the road, I would also be able to see if any of the changes I made in offensive tackle build contributed to lower penalties. When I ran the simulations for my offensive tackles over 650TPE experiment I maxed strength, pass blocking, run blocking, endurance, and speed. I got very close to maxing agility. With the data I already had I would be able to tell if there was a difference between the number of penalties a team has on the road compared to the number they have at home and also see if any of those attributes helped lower the amount of penalties.
If those attributes didn’t impact penalties, I speculated that intelligence would be the attribute that did. I decided that if it was necessary for me to test the value of intelligence I would run four sets of 100 game simulations – two sets at home and two sets on the road – for a team with their offensive line at the base intelligence (55) and then scale that intelligence up to 70, 80, and 90 to see the impact it would have.
I used the Otters as the control team as well. I used the current attribute levels of their offensive linemen and only added intelligence. I did not reduce any other attribute or make any other changes.
[div align=\\\"center\\\"]Results – [/div]
The first step was trying to see if a team has more penalties on the road than they have at home. I took the data from my previous offensive line studies and calculated the average number of penalties per game they had the average number of penalty yards they accrued per game in each situation.
[div align=\\\"center\\\"]
[/div]
There is a definite increase in penalties for teams on the road than teams at home. The control team in this experiment was the Orange County Otters. They averaged 4.78 penalties per game on the road across all TPE/Weight levels and 2.85 penalties per game at home across all TPE/Weight levels. The difference of 1.93 penalties per game between showcases that there is a difference in simulation between the penalties a team puts up at home versus on the road.
What this data doesn’t show is a correlation between increased weight, strength, pass blocking, run blocking, endurance, speed, or agility and lower penalties per game. This means that there must be a different attribute that impacts a team’s penalties per game. I hypothesized that it is intelligence, so I tested the Orange County Otters against every team with their offensive line having a scaling amount of intelligence. I started with all of their linemen having 55 intelligence because that is the default in the “athletic linemen” archetype and the lowest a lineman can have. I tested values all the way up to 90 intelligence which is the maximum any of the linemen archetypes can have.
[div align=\\\"center\\\"][/div]
For the Otters at home, as the intelligence of their offensive line increased the amount of penalties and penalty yards per game they were accruing decreased drastically – from 2.67 penalties per game on average costing them 20.97 yards per game at the lowest intelligence value to 1.54 penalties per game on average costing them 12.19 yards per game at the highest intelligence value. This is a difference of 1.13 penalties per game and 8.78 penalty yards per game.
[div align=\\\"center\\\"][/div]
At the same time as their penalties per game and average penalty yards per game were deceasing, their winning percentage was increasing – from 71.50% at the lowest intelligence value to 74.22% at the highest intelligence value. This is a difference of 2.72% that can be attributed to their offensive linemen maxing their intelligence.
[div align=\\\"center\\\"][/div]
For the Otters on the road, as the intelligence of their offensive line increased the amount of penalties and penalty yards per game they were accruing decreased drastically – from 4.41 penalties per game on average costing them 34.64 yards per game at the lowest intelligence value to 3.36 penalties per game on average costing them 26.8 yards per game at the highest intelligence value. This is a difference of 1.05 penalties per game and 7.84 penalty yards per game. The differences between the highest intelligence value and lowest intelligence value were smaller for the away runs than the home runs.
[div align=\\\"center\\\"][/div]
At the same time as their penalties per game and average penalty yards per game were deceasing, their winning percentage was increasing – from 48.25% at the lowest intelligence value to 50.14% at the highest intelligence value. This is a difference of 1.89% that can be attributed to their offensive linemen maxing their intelligence. The increase in win%, on average, also makes them a favorite against every team in the league while they’re on the road.
[div align=\\\"center\\\"]Conclusions –[/div]
1. By having their offensive linemen max intelligence teams can win 4.61% more games on the year. This difference is only for offensive linemen archetypes that can build up to 90 intelligence. For archetypes that cannot pass 80 intelligence the increase in winning percentage is between 1.09% and 3.52%,
2. Intelligence has less value in reducing penalties on the road than it does at home, but it still has a fairly pronounced impact.
3. Teams should be prioritizing as many human offensive linemen as possible because those linemen can earn past the highest bot levels – 550TPE and 750TPE – which means they can put more TPE into intelligence than teams have the ability to with bots.
4. I'm going to assume that intelligence has the same impact on other positions as well. I chose not to test it for this particular study because I didn't want to have juggle too many variables. Offensive line consistently gets more penalties than other position groups, so testing it for them would have the biggest impact. I'll test it for other positional groups at a later point.
[div align=\\\"center\\\"]Notes – [/div]
1. Once again this was a lot of sim work – around 4.8 hours not factoring in the amount of time it took to change team in the simulation, update the control team’s offensive line, and build the spreadsheets.
2. As always, my work is always open source. You can find it here.
I simulated 14,400 games so that you wouldn’t have to.
Before I get too deep into the weeds about this study I want to shout-out my Birddogs teammate @Forty Jordy. The core idea behind this study was his, I just executed it.
Forty has wondered a while whether penalties play a role in home field advantage. He has been tracking penalties on the year and has also found that offensive line players have about 50 penalties as a position group on the year while all other position groups peak around 20. Essentially, if penalties play a role in your ability to win on the road and you can reduce the amount of penalties coming from your offensive linemen then you should be able to increase the amount of games you win on the road.
I realized I had a good amount of this work already done. I had simulated 50,422 games in the last two weeks looking at the differences between human offensive linemen and bot offensive linemen. In doing all of those simulations I would not only have data for a team’s penalties at home and on the road, I would also be able to see if any of the changes I made in offensive tackle build contributed to lower penalties. When I ran the simulations for my offensive tackles over 650TPE experiment I maxed strength, pass blocking, run blocking, endurance, and speed. I got very close to maxing agility. With the data I already had I would be able to tell if there was a difference between the number of penalties a team has on the road compared to the number they have at home and also see if any of those attributes helped lower the amount of penalties.
If those attributes didn’t impact penalties, I speculated that intelligence would be the attribute that did. I decided that if it was necessary for me to test the value of intelligence I would run four sets of 100 game simulations – two sets at home and two sets on the road – for a team with their offensive line at the base intelligence (55) and then scale that intelligence up to 70, 80, and 90 to see the impact it would have.
I used the Otters as the control team as well. I used the current attribute levels of their offensive linemen and only added intelligence. I did not reduce any other attribute or make any other changes.
[div align=\\\"center\\\"]Results – [/div]
The first step was trying to see if a team has more penalties on the road than they have at home. I took the data from my previous offensive line studies and calculated the average number of penalties per game they had the average number of penalty yards they accrued per game in each situation.
[div align=\\\"center\\\"]
[/div]
There is a definite increase in penalties for teams on the road than teams at home. The control team in this experiment was the Orange County Otters. They averaged 4.78 penalties per game on the road across all TPE/Weight levels and 2.85 penalties per game at home across all TPE/Weight levels. The difference of 1.93 penalties per game between showcases that there is a difference in simulation between the penalties a team puts up at home versus on the road.
What this data doesn’t show is a correlation between increased weight, strength, pass blocking, run blocking, endurance, speed, or agility and lower penalties per game. This means that there must be a different attribute that impacts a team’s penalties per game. I hypothesized that it is intelligence, so I tested the Orange County Otters against every team with their offensive line having a scaling amount of intelligence. I started with all of their linemen having 55 intelligence because that is the default in the “athletic linemen” archetype and the lowest a lineman can have. I tested values all the way up to 90 intelligence which is the maximum any of the linemen archetypes can have.
[div align=\\\"center\\\"][/div]
For the Otters at home, as the intelligence of their offensive line increased the amount of penalties and penalty yards per game they were accruing decreased drastically – from 2.67 penalties per game on average costing them 20.97 yards per game at the lowest intelligence value to 1.54 penalties per game on average costing them 12.19 yards per game at the highest intelligence value. This is a difference of 1.13 penalties per game and 8.78 penalty yards per game.
[div align=\\\"center\\\"][/div]
At the same time as their penalties per game and average penalty yards per game were deceasing, their winning percentage was increasing – from 71.50% at the lowest intelligence value to 74.22% at the highest intelligence value. This is a difference of 2.72% that can be attributed to their offensive linemen maxing their intelligence.
[div align=\\\"center\\\"][/div]
For the Otters on the road, as the intelligence of their offensive line increased the amount of penalties and penalty yards per game they were accruing decreased drastically – from 4.41 penalties per game on average costing them 34.64 yards per game at the lowest intelligence value to 3.36 penalties per game on average costing them 26.8 yards per game at the highest intelligence value. This is a difference of 1.05 penalties per game and 7.84 penalty yards per game. The differences between the highest intelligence value and lowest intelligence value were smaller for the away runs than the home runs.
[div align=\\\"center\\\"][/div]
At the same time as their penalties per game and average penalty yards per game were deceasing, their winning percentage was increasing – from 48.25% at the lowest intelligence value to 50.14% at the highest intelligence value. This is a difference of 1.89% that can be attributed to their offensive linemen maxing their intelligence. The increase in win%, on average, also makes them a favorite against every team in the league while they’re on the road.
[div align=\\\"center\\\"]Conclusions –[/div]
1. By having their offensive linemen max intelligence teams can win 4.61% more games on the year. This difference is only for offensive linemen archetypes that can build up to 90 intelligence. For archetypes that cannot pass 80 intelligence the increase in winning percentage is between 1.09% and 3.52%,
2. Intelligence has less value in reducing penalties on the road than it does at home, but it still has a fairly pronounced impact.
3. Teams should be prioritizing as many human offensive linemen as possible because those linemen can earn past the highest bot levels – 550TPE and 750TPE – which means they can put more TPE into intelligence than teams have the ability to with bots.
4. I'm going to assume that intelligence has the same impact on other positions as well. I chose not to test it for this particular study because I didn't want to have juggle too many variables. Offensive line consistently gets more penalties than other position groups, so testing it for them would have the biggest impact. I'll test it for other positional groups at a later point.
[div align=\\\"center\\\"]Notes – [/div]
1. Once again this was a lot of sim work – around 4.8 hours not factoring in the amount of time it took to change team in the simulation, update the control team’s offensive line, and build the spreadsheets.
2. As always, my work is always open source. You can find it here.