Editorial:
Simply put, there’s nothing you can teach me with 8 or 12 uncontrolled replays of one of my seasons that I won’t have already seen in the hundreds I would have run to test scenarios alongside the guys who made the CMs. This was not done in silos, not in my case, it was done as a team, with Scotty, and in the case of the 1983 or 1984 USFL, HOF Ed. In the case of one season I had in prototype, I had to chuck the model and try again because the midpoint itself of the model itself was an issue. I had to push it to the next year.
HJ you’ve never understood or chosen to ignore the fact I’ve told you my metrics- I am concerned about the overall balance of net yards per ball control play and turnovers per ball control play first and foremost, not overall numbers. Overall numbers are dependent on things like the pace of the season in terms of plays and the pace of the teams themselves within them; the 1957 Rams are simultaneously not as good on offense and better on defense than their overall numbers might suggest because they simply run more plays per game.
Play calling matters. In our game every coach can choose to “up” their average gain per play by taking more risks and they can choose to try to Buddy Ryan the ball downfield, using conservative end runs and flat and short passing. Both approaches can work, but they would look different in replays. In the testing we did we focused mostly on that balance- are the teams penalized as correctly as we can penalize them based on prior precedent? Can they attempt to play different strategies and if they do will the trade offs work well? What you don’t want to do is give a team a “cheat code” player that allows them to go against the team’s* or season’s outcomes. I also have constraints, the consistency of patterns or approaches used in prior seasons.
We ran hundreds of replays. All of us are engineers. We know the data. There is nothing you can show me with a cursory replay set under an unknown rule set that we have not already seen. You can hate that response but I’m not going to debate it.
Rules changes and game flow change results; I don’t chase results with new cards. That would be silly. That goes against the approach of applying prior precedent to the data, which is how all seasons are made.
Now in terms of 1974 Washington, they pose a dual set of questions you have to try to answer from prior precedent. One - while Jurgensen has the better stats, Kilmer is the starter. So you have a team with a good relief pitcher. You have a relief QB with better stats but on fewer attempts.
There are a couple of examples of this in prior seasons. In 1972 Griese was clearly the better qb on the Wideboys than Morrall, who had the better stats. This reflected the fact that he was the starter to Don Shula. Griese did play the tougher opponents and was the one called upon for the Super Bowl. On the narrow cards, the reprints, the two QBs are closer together. There’s room for more than one interpretation.
The same is true with the 1970 Raiders, where Blanda actually received AFC player of the year votes because he won games at the buzzer with either big drives in relief or game winning kicks. But Lamonica is the starter. Here Blanda’s ints are higher, so it’s easier to card to his other stats.
The other item specific to 1974 Washington is a game flow question. They have a runner who had 144 carries at 3.0 yards per carry. This kind of a runner has a tendency to bring down the whole offense and that card poses a real problem for the whole team. Again, I won’t get into the specifics because I am not obligated to, but in such cases I am very concerned with how the team is performing overall and there are some constraints that must be met.
TLDR: the results you quoted don’t really surprise me at all. And I ran controlled experiments not replays, I am an engineer trained in DOE. If I could find a handful of scenarios that covered the data I expected to me that was good enough even if that was not the CM model chosen for the final baseline. Playcalling changes results.
Now the best part! I had already talked to The Game Co. about this last week. My Statute of Limitations for my seasons is up, and so if you want to see new versions send those suggestions to them or Mike. I don’t have the time or inclination to rework stuff from ten or more years ago, so go ahead and forward those ideas on to them. If the potential future sales might support it you might have some luck!
Meanwhile I have replays to run!
-Fred
* Original rulebook from 1960s. Last line there on page 19. Applies to 1974 Kilmer and Jurgensen.
No comments:
Post a Comment