(07-23-2020, 10:58 AM)speculadora Wrote:another LIE from FAKE NEWS mainstream media!
since Trautner brought me here let me just make a quick point. when you're looking at test sims, you have to remember teams aren't testing against their opponent's next strategy. they're testing against the previous. so you should really be comparing four numbers here:
1) baseline (both teams previous week strategy)
2) home team vs road team's previous week
3) road team vs home team's previous week
4) actual results
since we're talking about test simming the best any team is looking to do is maximize their odds against their opponent's previous week strat. although there's probably a certain amount of prediction you can do of what your opponent will do, it's mostly throwing darts. comparing the baseline to the following week ignores the fact that each team's changes interact with the others in sometimes unforeseen ways. this isn't to say you ignore the ultimate result of both team's changes, but rather that you should hold constant what you can first.
here's a rough example of this using the four sims I laid out above, numbers expressed as home team win%
1: 70%
2: 74%
3: 66%
4: 68%
so here you'd see that both teams tests improved 4% relative to the baseline, but the ultimate outcome skewed more toward the road team's tests. so with your methodology this would count as -2% for the home team. but the reality of this is that both teams believed they improved, but the road team's changes were more effective.
the road team improved their odds 6% relative to the home team, while the home team improved their odds by 2% relative to the road team. despite the home team slipping by about 2 points, there's a real argument that they still improved their odds of winning this game relative to doing nothing.
and what you might wind up wanting to do is breaking the question of "who's the best" down into two categories:
- improvement relative to baseline (+4% for each team above)
- improvement relative to opponent changes (+2% for home, +6% for road above)
there's probably some other math you could do to further normalize here but I've gone on way longer than I intended already
nerd lmao
![[Image: xboy623.gif]](https://sig.grumpybumpers.com/host/xboy623.gif)
![[Image: jeffie43bottom.gif]](https://sig.grumpybumpers.com/host/jeffie43bottom.gif)