Ow, ow, ow
DMZ · March 29, 2007 at 11:33 am · Filed Under Mariners
Nate Silver’s got the PECOTA projected standings for the AL West up, and… they sting. The raw numbers: Angels at 86 wins, Oakland and Texas at 80, Seattle at 73.
73.
Oh, and if his playoff predictions pan out, Bob Melvin will win a World Series this year.
Or Junior Griffey.
It’s true that we can’t predict injuries for individual players. But we can make this prediction for teams: almost all of them will suffer from at least one serious injury to one of their main players, where “serious” means that either they’re disabled, or their performance suffers significantly, for a significantly long period of time.
To assume that a team will go through a season without suffering from an injury problem is … well let’s just say it goes against what history shows. We can’t know who it’ll be, but there’ll be someone who gets injured.
43: But what you call PECOTA getting lucky is simply what happens when you aggregate more data: unexpected performances tend to balance out.
I would expect PECOTA to be more accurate, on average, on predictions for individual teams vs individual players. Assuming a system is a good projection system to begin with, more data should mean more accuracy.
Is it possible to analytically compare PECOTA’s historical accuracy on projecting players vs projecting teams? Or is that a monstrous amount of work?
#46
Personally, I don’t like football futures because key injuries happen all the time — like 2 or 3 years ago when the Seahawks were starting 3 rookie middle linebackers at the same time. This last year Trufant got hurt and the Seahawks were literally signing guys who were bagging groceries the previous week — to play in the playoffs.
I don’t like basketball futures because it’s so hard to quantify that sport — I don’t like gambling with my gut that much. The Pro game is better for that than the college game though. With the pro game there’s no surprising “rookie” like Tyrus Thomas 2 years ago — he got a review like: “great leaper who may make an impact” — all of about 7 words — in a preseason magazine I had.
Finally, I’m not a fan of baseball futures because there are enough tools to know which teams *should* be good, and which teams *should* be bad. The good teams have odds that are too short (ie: don’t pay enough for the risk), the bad teams you’re throwing your money away — the Tigers won a bunch of games but they didn’t win the WS. If Pecota has you at 73 wins…forget it till next year.
Anybody know when the last time a team predicted (or projected) to win 73 games won the WS?
It happened in the NL the last two seasons, and in the majors overall in 2000. And it’s FREQUENTLY been just a couple of games away from happening. It’s not an unreasonable place to start at all. It’s much more likely than to have a extreme outlier. Baseball is a parity sport, much more so than the others. Something about those 162 games….
Gagne hits the DL for Texas. I think we need to add a win to the M’s projection.
http://sports.espn.go.com/mlb/spring2007/news/story?id=2817649
I was just in Vegas and *every* team line was high. I think if you’d total them all up you’d see an average of about 84 wins per team.
This can’t possibly be right since you can bet the under on these lines. If the average was 84, you’d make a killing betting unders, even paying 12:10. I just checked the first two online sites I found — one had an average of 81.7 wins, the other 81.8 wins.
Yes, as you said optimism and the tendency to bet over (I think my team is better than that) pushes the lines above average a bit, but the effect is slight. And since the lines are fixed, if the betting is heavy one way, they change the odds.
And I agree some of those lines look like easy money betting the under — White Sox 87.5, Angels 90, Dodgers 88.5, Yankees 97 (!).
Ok, the lines in Vegas weren’t quite as “high” as they felt — they still felt high.
The Over/Under for the M’s is 78.5 wins. Furthermore, let’s say in advance of the year you know that the M’s are going to score 650 runs and allow 720 runs, a 73 win ratio. What’s the standard deviation of wins based around the Pythagorean Formula? That is — if you bet the Under *knowing you were “right”, how often would you actually win? Would that cover the 12:10 required to play?
On whether futures bets are worth it: I wrote about this last year (specifically, the chances a team will outperform expectations) and while in general I don’t like them, you can make really good money – I my over/under picks last year would have made a ridiculous return for readers in Vegas.
The good thing I find is that if you find a team that for whatever reason public perception is way out of line with their actual fortunes, you can bet your way and get a huuuuuge cushion to be wrong. That’s pretty nice.
Check out the lines on Tampa Bay and Arizona this year. Taking the over on either one of those seems like a very safe bet.
This could possible be the worst $110M I’ve ever seen spent on a ballclub. Most of the roses collectively add up to about $30M (Ichiro, Johjima, Beltre, and the youngsters).
That means there’s about $80M worth of dreck…
But what you call PECOTA getting lucky is simply what happens when you aggregate more data: unexpected performances tend to balance out.
But that’s exactly what I mean by getting lucky. The systematic agorithm takes in a wide variety of variables and possiblities and projects an outcome based on what it knows at the time. But if the variables change over the course of the season, and last season’s July / August / September / October team was not the same as the one that PECOTA originally projected, how can we credit PECOTA’s outcome as being accurate when it was?
I’m just saying that looking at PECOTA’s past results for teams and applying it to current projections for teams is irresponsible unless one looks at the entirety of the data.
And before anybody else jumps in here to make offhanded comments, I’m not knocking PECOTA at all, nor am I claiming USSM is pessimistic. I’m knocking our interpretations of the results of PECOTA’s projections and our general laziness to question when it’s needed.
Huh- so, basically, PECOTA is projecting the AL West to be the 2007 version of the NL West in 2005 or NL Central of 2006- the division being collectively pretty weak and well under .500.
Since that aligns pretty much with what I’ve been thinking since the offseason, OK, then.
As for the Mariners being terrible, well, PECOTA hates Ichiro, always has, probably always will. I take the 73 as the probably low end scenario- Guillen’s a shambles, all the veterans decline collectively, and so on. The problem is that for a team payroll that has 9 digits, we should be looking at a win total that’s much closer to three digits than what we’re getting.
What this DOES seem to say is nobody in the AL West is going to blow the doors off anyone else…
I guess I have a couple responses:
– Obviously, you have to acknowledge that doing this is pretty much taking a lot of forecasts and playing time guesses and putting it all in a blender, and doesn’t reflect what will happen
– To the issue of teams not being the same, that’s certainly true, but for the most part, team composition doesn’t change that much in the aggregate. No team gears up or tears down entirely in any single year.
Look at this as more of a start-of-season rough cut: “If these guys perform at their weighted mean forecasts and we’re more or less right about who gets what playing time, here’s the relative strength of the teams.”
I don’t view these as having any real predictive value – but they are still interesting.
Derek, I’d be interested to reread your post on betting but wasn’t able to find it. Do you have a link handy?
Accuracy with Pecota is dependant on context. Offensively, i’m very, very confortable with it’s projections (correlations are over .70 typically). Pecota can’t handle pitching nearly as well though (projection correlate to actual roughly around .45) which makes sense I guess.
So if there is any saving grace here, its that the Ms can make up wins because Pecota has missed on the pitching side of things. Looking at the staff though, I’m thinking that’s more a case of hope springs eternal.
Then there is some marginal stuff you can consider that might pick up a few wins. Having Guillen’s arm in right with Ichiro’s in center might be collectively good for 7-8 runs (maybe a win?). Then you can hope that the two do something special in high leverage situations that could get you another win or two…
Of course that California sliding into the ocean (while Texas is at Oakland) idea sounds like a winner…
Careful what you wish for though, the Ms might be at Anahiem that weekend….
DMZ,
Okay. I guess I’m just thinking about this shit too hard.
The NL? yes. The entire MLB? No. The NL is crappy and had/has zero outstanding teams. When it happened in 2000, the Giants had a .599 win %.
PECOTA had the Yankees maxing out at 93 wins, I’d be shocked if one team wasn’t over .600. I take those results with a salt lick.
This may explain why, 2 years after canceling my season tix, they’re STILL sending me postcards begging me to come back and telling me I wouldn’t lose my previous season ticket holder priority?
Nah, what it explains is why I’ll still be getting those postcards begging me to come back every year for the foreseeable future…
PECOTA is not predicting that no team will play .600 ball. PECOTA is not projecting any specific team to do so. Those are not the same thing. PECOTA is saying that no team is built to play .600 ball under their ordinary circumstances: that for a team to do so they will have to play over their heads.
I’m sure that if you looked at all the projected seasons that PECOTA ran, in the majority of them some team played over .600 ball. But it was a bunch of different teams, no team doing so often enough to set that as their level.
I’ll bet almost every year the best team and the worst team are better and worse than any of PECOTA’s projections. They’re telling us what each team is, more or less, at their average performance. Of course some teams will play better than their average. And of course PECOTA can’t predict which teams will do so. Their projections still tell you something useful about the relative strength of the teams, and a team projected to win 73 is a lot less likely to overachieve to .600 ball than a team projected to win 93.
Pecota also assumes that the same players will be on the final roster as the starting roster. We all know that this is not true. Every year some teams hold fire sales and some teams buy players for the stretch. I actually expect the M’s to hit around 68 wins after they do a player dump in July. Bavasi will get canned around mid July and the new GM will gut the team, trying to get anything he can for our current overpriced squad.
We’ll be playing with the kids in August & September with the expected results.