Post by eric on Feb 6, 2019 21:39:30 GMT
Gonna put some stats in here, discussion welcome.
Leaguewide stats (2012 not including robots).
This tells us that the average human is a better player than before, to a level that is unlikely to be statistical noise.
.
Regression of pts/tsa from 2012 to 2011, players of >1k pts in each (54 total):
This tells us that particular humans did not see nearly as much improvement as the average human; on average a human player saw their pts/tsa increase by .025 as opposed to .050ish.
This tells us that the increase in average human efficiency was evenly due to particular humans getting better and other, worse particular human scrubs getting their walking papers going fishing on vacation. Not naming names (Brain Winter).
The regression also tells us that better scorers were more likely to get worse. This part is probably regression to the mean; Vince Carter probably didn't drop from 1.22 to 1.16 because of robots, but because his true talent level wasn't 1.22 in the first place and last year was a fluke, further borne out by him being a 1.15 type the previous five seasons. In the same way Yao is probably not really a 1.30 player and probably wasn't really a 1.19 player last year, but is probably still the 1.23 he was the rest of his career.
Since many players improve in TC and players always improve in labs (but not Germany, sorry Daniel Taurasi) my read is that the robot players are pretty well tuned to represent league average.
.
We can also look at the robot teams overall. The Eastern Conference robots were designed to lose every possible game, and since per the Donovan McNabb rule games cannot end in ties they had to add up to at least eight wins in the games they played each other. They in fact won 4 + 3 + 1 = 8 games.
The Western Conference robots were designed to finish .500 and ended up 536-612 (.467). I'm pretty satisfied with that.
.
.
I expect the contraction concentration of talent to wane slightly over time.
-We only have 12 GMs putting points into players instead of 29.
-Draft classes are going to be reduced in size, making the second round TC generated stud that much less likely to appear.
-Contracts will remain wonky for awhile, but that has more to do with which teams have talent, not how much talent is in the league.
.
.
All in all I think this went really well! Looking forward to more seasons. :D
Leaguewide stats (2012 not including robots).
year REB A PF ST TO BL PTS FG% FT% 3P% PTS/TSA
2012 8.72 3.98 3.07 1.25 1.52 .77 19.48 .493 .831 .403 1.152
2011 8.11 3.83 3.10 1.16 1.75 .82 18.28 .474 .813 .387 1.104
2010 8.25 3.84 3.10 1.15 1.75 .84 18.48 .473 .811 .393 1.103
2009 8.27 3.84 3.12 1.14 1.71 .82 18.51 .471 .819 .388 1.102This tells us that the average human is a better player than before, to a level that is unlikely to be statistical noise.
.
Regression of pts/tsa from 2012 to 2011, players of >1k pts in each (54 total):
This tells us that particular humans did not see nearly as much improvement as the average human; on average a human player saw their pts/tsa increase by .025 as opposed to .050ish.
This tells us that the increase in average human efficiency was evenly due to particular humans getting better and other, worse particular human scrubs getting their walking papers going fishing on vacation. Not naming names (Brain Winter).
The regression also tells us that better scorers were more likely to get worse. This part is probably regression to the mean; Vince Carter probably didn't drop from 1.22 to 1.16 because of robots, but because his true talent level wasn't 1.22 in the first place and last year was a fluke, further borne out by him being a 1.15 type the previous five seasons. In the same way Yao is probably not really a 1.30 player and probably wasn't really a 1.19 player last year, but is probably still the 1.23 he was the rest of his career.
Since many players improve in TC and players always improve in labs (but not Germany, sorry Daniel Taurasi) my read is that the robot players are pretty well tuned to represent league average.
.
We can also look at the robot teams overall. The Eastern Conference robots were designed to lose every possible game, and since per the Donovan McNabb rule games cannot end in ties they had to add up to at least eight wins in the games they played each other. They in fact won 4 + 3 + 1 = 8 games.
The Western Conference robots were designed to finish .500 and ended up 536-612 (.467). I'm pretty satisfied with that.
.
.
I expect the contraction concentration of talent to wane slightly over time.
-We only have 12 GMs putting points into players instead of 29.
-Draft classes are going to be reduced in size, making the second round TC generated stud that much less likely to appear.
-Contracts will remain wonky for awhile, but that has more to do with which teams have talent, not how much talent is in the league.
.
.
All in all I think this went really well! Looking forward to more seasons. :D
