There’s been some consternation by many (including me, in my head), about the Rangers’ easier remaining schedule.
Excluding head to head, Tx has:
10 vs. Ana
10 vs. Hou
6 vs. Sea
4 vs. TB
3 vs. KC
3 vs. Min.
3 vs. CHW
3 vs. Pit
2 vs. Mil
The aggregate winning percentage of these teams, weighted by number of games, is .443.
6 vs. Ana
7 vs. Hou
6 vs. Sea
3 vs. TB
3 vs. Bal.
4 vs. Tor.
4 vs. Det
3 vs. Cle
7 vs. Min
2 vs. Cin
Aggregate winning percentage of these teams is .451.
For purposes of figuring out the advantage of that difference, I assumed both Tx and the A’s are .59 true talent. Based on that assumption, Tx should have a .644 winning pct against .443 opposition, and the A’s would have a .636 winning pct.
That would give Tx an expected .4 game advantage from the remaining schedules. N, again excluding head to head… Not huge of course, but probably actually somewhat substantial as these things go.
This is actually quite reassuring, at least to me. The now A’s have a 2.1 game advantage (2.5 -0.4), excluding head-to-head, right?
At least for the next few hours…
no faith in Jerome Williams?
The part that concerns me is it’s easier to stay ahead than catch up. By the end of the month, because of the August schedule, we could be playing from behind and I’m mostly concerned about that.
I’m confused by your numbers (but then I’m also jet-lagged up the wazoo)
How do you get 0.4 wins? Using your 0.644/0.636 win probabilities over 44 and 45 games respectively I get 29.2 and 28.6 wins, for a delta of 0.6 wins.
Beyond that, how are you getting the win probabilities? If I assume a linear function then I’m getting Texas winning 0.647 against 0.443 opponents, and Oakland 0.639 against 0.451 opponents.
Beyond beyond that, a linear function can’t be right (since it should go to 0.0 against 1.0 opponents and 1.0 against 0.0 opponents) but if it isn’t linear I don’t think you can do the aggregating.
There was a time I considered myself a reasonably numerate fellow. Then I met you people.
“…if it isn’t linear I don’t think you can do the aggregating” made me think IYCDTWB…
You (and I) are reasonably numerate, some people here are unreasonably numerate.
At Cal I passed up Math 1A on the AP and did fine in 1B. I got into 1C (quarter system in those days), looked around at all the engineers and waved the white flag.
by waiving the whit flag you mean, ripping off one of their short sleeve button up white shirts?
Waving the Whit flag = sticking up for Blevins’ girlfriend?
The first part is just
expected number of wins = win probability x number of games
so 0.644×44=29.216 and 0.636×45=28.620, and 29.216-28.620=0.596 (which I just called 0.6).
The second part is the question of determining a team’s win probability against a particular opponent. When we say the A’s are a 0.590 team, that’s against a notional “average” (0.500) team; equally we know that they should have a win probability of 0.500 against another 0.590 team. Assuming that win probability is a linear function of opponent’s strength we can then extrapolate from these two points to any other opponent. My problem is doing that gives me different values to mikeA’s so we must be using a different model.
The third part is that the assumption of a linear function has to be wrong at some level, since the win probability should be 1.000 against a (hypothetical) 0.000 team and 0.000 against 1.000 team, whereas the linear model predicts win probabilities of 1.009 (!) and 0.009 respectively. However, if the function isn’t linear then you can no longer say simply aggregate all the individual teams to get an effective schedule strength.
The formula mikeA cited is called the odds ratio or Log5 method.
Basically, imagine that they each flip coins weighted according to their own winning percentage, and if there is a tie they each flip again. It’s a reasonable model that behaves as it should at both the middle (evenly matched teams) and the boundaries.
Also, yes, it’s non-linear, but the non-linear effects are second order, and probably not going to make any difference when the first order value is 0.4 wins.
Cool – thanks. And yes, doing everything right (including the extra game and not aggregating) the delta comes out to 0.444 wins.
As a proxy, even the adapted second order run-differential Pythagorean will do:
Which is basically just like using runs scored and scored against and gives very, very similar results.
Woo! Someone used the LaTeX plugin.
It is awesome to have it.
Kinky.
My nose started to bleed just looking at that.
1. I used the same denominator because I considered the extra game not to be a s.o.s. difference.
2. The probabilities are beyond my ability to figure out on my own. I used a formula from bill James purporting to be a way to calculate single game odds based on winning %. Formula is (a-a*b)/(a+b-2*a*b). Whatever you used might be better.
3. I don’t see the 1 and 0 thing as a problem. No baseball teams like that would ever exist, but its not incoherent,and the exercise is just using approximations of “true” winning pct. Not being able to aggregate would be counterintuitive to me…
I posted a few times yesterday about my fears, but I think the most likely outcome is the A’s start playing better, and things take care of themselves (even if it means the Wild Card), or they don’t.
Well, the problem as I see it is the Rangers have a better team.
That’s why I mentioned the Wild Card.
I took a casual look through the minor league rosters, and it looks like the only guys down there that might be able to help the A’s in September are Sonny Gray and Michael Choice.
You have no faith in a Sept Jemile Weeks miracle?
I can see Montz or Vogt hitting a monster late-September homer, Dan Johnson-style.
(it will be Rosales doing that)
lets hope angel hernandez isn’t on the crew
Ouch, that would hurt a lot.
As mentioned elsewhere the Rangers DFA’d Rosales today. I assume he’ll clear waivers and declare free agency, Billy will re-sign him and he’ll be in Oakland in September.
Yay!
Ah. Though maybe now he becomes a Devil Ray.
Im picturing a big CY25 hit at some point. Maybe even a hot Sept or post season.
I’m doing what I can to help manifest that reality.
I see a Johnny Damon style 2 HR game and good playoffs.
Good stuff, thanks for doing it!
How did you come up with the TTL of 0.59 for each team? OAK has WP% of .577 and TEX of .558. If you regress with 70 games* of .500, you come up with .547 and .536 respectively.
Or did you use second or third order wins?
*I think that is the point of r=.7, IIRC and I’m to lazy to calculate now
And plugging these numbers into the Log5 formula would give Oakland .595 winning percentage the rest of the way and Texas .592.
i’ll take it
I wasn’t trying to be accurate with those, just chose a plausible number, probably was too high.
I used the same number for both teams because the point of the exercise was to estimate the advantage just from different s.o.s. I wasn’t trying to do a projection for the rest of the season. So whatever ttl I chose had to be the same for both.
OK, thanks. Nothing wrong with that, was just curious.
One could use Fangraphs ROS team projection or BBPro if one were so inclined. Then again, if you’re doing that might as well simply use their playoff odds.
Just a dumb question- has anyone ever tried to adjust for home vs. away in the odds ratio method, or the second-order pythagorean method? I’m only wondering because the A’s have especially drastic home/away splits due to the pitching staff’s high fly ball rates. I haven’t looked at the schedule, but if there are a larger weight of home or away games for the A’s, it might change things a bit in either direction.
I would guess that something like the BPro playoff odds, where they Monte Carlo the rest of the season would account for home/away at some level. On the other hand, it seems like all but the most extreme team splits would be pretty hard to measure, given that you only have 162 games in the full season (and for something like that playoff odds page, you are projecting from less than a full season). But it would make sense to at least include the (well measured) league-wide home/away split.
That 14.4% 7-day drop in the A’s playoff likelihood is pretty ugly.
If you look at it historically, you will see that home teams win at around .530 clip. This happens mostly because of the batting in the bottom of the inning, but certainly to some extent because of more rested players, familiarity with the field and any roster-to-ballpark construction, such as you mention.
However, you are mistaken in assuming that the A’s have a drastic home away split this year. The A’s have a home winning percentage of .636 and an away winning percentage of .518. That’s added home winning percentage of .552, which is just a tad above this year’s average of .546, and ranks them just in the middle of the pack:
LOLmets… and LOLcubs, while we’re at it.
Have the home and road schedules been comparably hard?
The A’s ones? I’m not sure, but it is rather easy to check
I’ll have those numbers for you coming right up.
If you only wrote this in italics
Heh…yeah, I should have done that.
So easy I was hoping someone would do it for me.
As tango always says, it is a great opportunity for a young researcher!
Which rules me out.
In fairness, that split was bigger pre-break.
True, .570, but still good for only 9th in MLB…
Your ranking confuses me. Wouldn’t it be Home WPct – Away WPct?
It’s home winning percentage divided with overall winning percentage and then multiplied with a .500 record so it brings everybody to the same level.
Or in other words, it says if each of these teams had an overall .500 record, and won more games at home at the same rate they are doing it now, that would be their home winning percentage (and subsequently 1 – that would be their away winning percentage)
That seems unnecessarily overly complicated. And it seems to ignore away altogether.
It’s not complicated at all. And it doesn’t ignore away altogether, because away is included in overall winning percentage.
It is the number that answers the question What would this team’s home winning percentage be if it were an overall .500 team? The other one is not.
Not that the Angels worry me any more, but since Pujols went down:
That’s 22.1% of the pitches off the outside part of the plate, 21.3% off the inside part.
Just in case any of you like attending Gold Rush Days in Old Sacramento, they’re going to be doing recreations of it from roughly 150 years ago, looking like twice daily.
Keep an eye on the site for scheduling.
http://sacramentogoldrushdays.com/
Uh, doing recreations of Old West baseball, that is.
I saw the “suffragists” walking the boardwalk on Saturday when I was down there, and I so wanted to tell those ladies about female supremacy and lesbianism, but then I remembered I’m not actually Dr. Who…
muahahahahaha
I’ve kinda moved on from second wave feminism anyway. I’m ensconced in the 3 1/2th wave now.
Although I’m still a HUGE fan of Shulamith Firestone and I think she should be up there with the other predictors of the singularity, Vinge and Kurzweil. Her prediction was more for a post-gendered post-biologically reproductive internal&externally enhanced human society rather than a post-human society, but she was close enough for 1972 for me to round up…
just wait. 13th doctor will have to be trans, if the kerfuffle over the 12th possibly being a woman is any indicator.
12th as already announced. It’s a dude.
If it was The Dude I might be interested.
yep, that’s what i was basing my thought on. they’re gonna have to ease into a gender change because reaction seemed to stop it cold this time.
crap:
Has better arm than Coco
and Garza.
Best part is dad not even knowing his kid threw it back.
I love how the father isn’t even aware as he’s thanking the dude who gave it to him.
Didn’t that kiddo throw out the first pitch recently.
yes
I mean the kid did the right thing, it was Austin Jackson’s home run
Ooops, I stand corrected, it wasn’t a home run ball
they threw out the first pitch in may
I like the older boy in the OAKLAND jersey sitting a couple of rows up.