Monday, May 17, 2010

rotating league name week 5 recap

Far too involved league analysis:

Couldn't decide what I wanted to do for this week's update, so I just did way too much for way too little.

The table below shows expected win values next to actual win totals and the difference between the two.

Here's how we got there. I found the average values and standard deviations for each category. I then classified each weekly category result and assigned that result a point value (high outlier +2, above average +1, below average 0, low outlier -2). I figured you had a reasonable expectation to win a category by being above average in it, with bonuses/punishment for extreme showings. I'm pretty sure I treated negative categories correctly. If you're at all interested, I used 1.75sd for the outlier break, because there were very few above the standard 1.96. I then added up all those categories and weeks and got an expected points total (451), which I then scaled to the number of possible wins (360 to this point). The actual total of wins is lower than 360 due to ties.

Anyway, here's the result:

A negative difference means based on your teams performance, your team would have fared better against an average team than against the teams it actually faced. The biggest caveats are that this is actually less precise than the other way I did expected wins, as this assumes an on/off for winning or not, while the other rates your chance to win a category given others' performance. (old way explained here). This way was slightly easier to look at the season as a whole though, instead of just one week. It also helps show how much luck and opponent can affect the standings.

And since Shane is still kicking names and taking ass, here's the ongoing best (with a new best and worst from week 5):

Brew Plop Week 5

Griping it up:

Because I lost epically to my fantasy nemesis (that's right, it's gotten that bad after this loss and last years final), I had time to stew and think of several ways to make myself feel better.

First up, a new thing I hadn't though much of until this weekend:

Through week 5 in our league, there have been 1561 runs scored and 1436 RBI compared to 2060 hitter's strikeouts (only 1710 pitching K's thanks to the minimalist style of several). Since there have been more -K than either R or RBI, it would seem that a "good week" would be one where you had more R or RBI than -K. Likewise, since there are more R and RBI combined than K (by a lot), a "bad week" would be one where your hitter's strike out more than those two stats combined.

The above table shows the number of weeks a team had more R or RBI than -K. It also shows a negative count of how many times a team had less combined RBI and R than -K. At the bottom of the table, you see the average, min and max number of wins had when having a "good week" or a "bad week", as narrowly defined for this post.

There are nine instances of good R week. In six of those instances, the team managed 7 or more wins (win totals of 6,5, and 3-my terrible week 5- were the others). There are also nine instances of good RBI week. In five of those, the team won 8 or more, but the remaining four were just okay (6,5,5,3-see above).

There are 13 total "good weeks", meaning that there were 5 weeks where a team was good in both R and RBI. Those five weeks resulted in win totals of 10 (dalek), 8 (edit), 8 (unicorn elbow), 8 (hops), and 3 (who do you think). In general, if you have a double good week, you win the week. It also appears that R is a better indicator than RBI for how well your team might do.

On the flip side are "bad weeks", which seem to be more straight forward. If you have a "bad week", your team will most likely win only 4 games.

Because Jon asked for it (and I'm still feeling sorry for myself), here are week 5's expected wins compared to actual.

Of note, I also tried incorporating in ties this time (by counting each tie as half a win), but it didn't change the overall picture by much.

Not a lot to comment on that doesn't make me feel angry and bitter. Interesting though to see when the match-up projects less than 12 wins, who picks up the extras (Dunbars and Reynolds this week).

Lastly, here's a longer term view of the expected wins.

Lots of fluctuations here that you can sort through for yourself. All the same caveats apply to this as to regular roto rankings (i.e., doesn't adjust for extreme weeks, no account for punting a category).

Video updated: My glasses were falling down and I look kinda wonky. I'll try to get someone better to do it next time.