Mariner lineup composition
Using the Baseball Musings Lineup Analysis tool (see also this link for more info) and the 2006 PECTOA projections:
The best lineup it came up with was (surprisingly)
Johjima
Sexson
Everett
Ibanez
Beltre
Betancourt
Ichiro
Lopez
Reed
at a whopping 5.016 runs a game. Fun fact: in the “best lineup” section, Sexson bats 2nd every time and Johjima or Reed bats first.
Worst lineup, at 4.792 r/g:
Lopez
Betancourt
Reed
Everett
Ichiro
Ibanez
Beltre
Johjima
Sexson
Of course, standard caveats apply: we’re using the weighted mean forecasts, it doesn’t take into account L/R advantages or tactical concerns, much less player contentment. But it’s still interesting. If for no other reason than to think that using this methodology, the difference between best and worst lineup is .2 a game, which is over 20 runs a season, which is two games in the standings!
Still, the difference between the best guess at what the actual conventional-wisdom lineup will be and the ideal lineup is only .1 runs/game.
I’ll spare everyone my super-long rant about lineup optimization, but I find this stuff fascinating. And you know who gave us the idea for this post? Peter White! Yes that Peter White!
I also find this stuff pretty interesting, just curious what the avg runs per game for the lineup that Hargrove is most likely to run out there.
It looks to me like most of the difference between best and worst is all the extra at-bats for Sexson. Hold him at a fixed position and I’ll bet you can’t find enough of a difference to matter with any permutation of the other guys.
I don’t think this tool is likely to be super far off the mark, and it’s definitely fun. However, it doesn’t seem like a very good tool for analyzing non-standard lineups. (Especially really non-standard lineups.)
By using historical data to determine the linear regression, what the regression coefficients tell us is how much one point of OBP (or SLG) from each lineup spot changes the number of runs per game under the assumption that the rest of the lineup is constructed traditionally.
So, one might argue that one of the reasons those number 3 and 4 hitters are so important is because there are speedy guys on the bases in front of them, stealing bases and getting into scoring position. Since the vast, vast majority of lineups used in the regression don’t have a guy as slow as Sexson in the number 2 spot in the order, we can’t really use this analysis to tell us anything about whether this would make a difference or not.
I really think that if you want to say something definitively about how many runs a non-standard lineup would score, you have to appeal to some sort of simulation, rather than just comparing to historical data, because the historical data tells us very, very little about non-traditional lineups. This algorithm probably gives us a pretty good idea of a lineup’s production provided it’s not too weird, though.
That lineup is contary to the traditional belief that speed is important.
As in running bases not bat speed. It seems as though the potential is greater for many double plays as potential runs scored. Although it would be interesting to read fallout from that lineup should Hargrove ever pencil it in.
And what do you guys think of putting Jojima second. I think he can do it but I think so many ABs plus catching will tire him out too quickly in the season, no?
Okay, here’s something kind of interesting I found while playing around for a little bit. Let’s say that we do want to try to take platoon effects into account. I took the 9 guys from last year with the most ABs against LHP who can fill out all 9 positions–Ichiro, Ibanez, Beltre, Winn, Sexson, Reed, Morse, Bloomquist, and Olivo. Then, I put their stats against LHP into the lineup analysis tool.
Using those stats, the best lineup scored 4.987 runs against LHP while the worst lineup scored 4.141 runs against LHP, for a pretty substantial difference of 0.9 runs/game. Granted, the worst lineup is extremely pathological. However, I think the point here is that the greater the variance in performance amongst your players, the more lineup effects will matter. So, managers that keep their lineup static no matter who is pitching, ignoring platoon effects, stand to potentially lose quite a few runs over the course of the season.
Mat’s #3 says about what I was going to say: That the regression approach loses its value with non-standard lineups. (In particular, I strongly suspect that it significantly overvalues the potential gain from nontraditional lineups; I’d put the traditional versions at maybe 4-5 runs/season below optimal.)
I mentioned my preferred approach in the Reed thread a couple days ago, and will do so again here:
Runs = Sum(i=1 to 9) PA_i * OBP_i * WeightedAvg(Slg_i+1,…,Slg_i+5)
where the weighted average of the Slg terms is according to the likelihood of batter i+j coming up in the inning with batter i still on base. (My Excel file uses .45, .30, .15, .075, and .025, but that’s just a guess– to research it and get it exact would take more time than it’s probably worth.)
Is it premature to ink in Johjima in the two hole? I mean he hasnt even taken an at bat in a spring training scrimmage yet.
I think comment #4 has a point. Ignoring speed (since it seems from a glance at the website that only OBA and SLG are used in construction) might lead to biased results. Ichiro going first to third on a single is valuable… or having someboday steal a base with good success rate for that matter.
This might be more interesting in analyzing PAST lineups (since we’re not adding performance projections). Where would Ichiro have been best in 2005? 2001?
EC has it right. Using actual past performance, what would have been the best ordering of past lineups?
If we did that historically, we might be able to build a better model going forward that wasn’t reliant upon having a traditional build.
I would also be really interested in seeing this done for past lineups. And Ichrio 7th is mind-boggling, even if a computer thinks that is correct.
Ichiro could still go first to third on a single batting fifth. That said, I think this lineup tool is kinda fun for the off-season, but its biggest flaws are that it doesn’t take into account the double play or L/R matchups (both of which have been said before, I guess), and I think those are both going to have real effects on the number of runs scored.
It seems to want you to bat your best hitter second, which is something I’ve seen people advocating since Barry Bonds’ 2002; it also makes the valid point that it may not be wise to wedge your worst hitter in the nine hole, right ahead of your best hitters. Other things (such as batting Everett third) seem counter-intuitive to the point of warranting further exploration, if you ask me.
All in all, I’m not sure regressions are the best way to determine lineup construction. Not that can I can come up with a preferable alternative.
“And Ichrio 7th is mind-boggling, even if a computer thinks that is correct.”
Don’t forget that PECOTA doesn’t think Ichiro will have a great season next year. That’s likely why he ends up so far down in the lineup, moreso than something funny with the lineup analysis tool.
Ichiro seventh isn’t so mind-boggling when you consider his PECOTA projection is 308 / 343 / 406. Zips is a little more optimistic, with 319 / 366 / 429. Given that this lineup was constructed with PECOTA weighted mean projections, I don’t think it’s absurd to bat Ichiro 7th. If he bats something like .308. he’s just not a particularly good hitter.
“If he bats something like .308. he’s just not a particularly good hitter.”
He may not be as valuable as the money we are paying him, or contributing what we really need from him, but i think its a stretch to say he is not a good hitter at a .308 clip.
No, but really. A .749 OPS (what PECOTA is projecting him for) is pretty bad for a corner OF. Granted, speed and defense helps, and Safeco hurts that, but the AL average OBP last year was .330, and average SLG was .424…a .754 OPS. Granted, Ichiro’s OBP is more valuable than the missing SLG- but the average AL OBP and SLG also includes SS, 2B, CF and C that are leftwards on the defensive spectrum, where an awful lot of bad hitters live.
Ichiro hitting a shade above .300 with those power/OBP numbers is a VERY mediocre hitter for a corner OF. It would also be his worst season ever, by a lot…so, basically, PECOTA hates him for some reason.
PECOTA seems to hate guys who are on the wrong side of 30, and starts projecting big drop-offs in production.
Although, Ichiro is such an amazing athelete and doesn’t seem to be losing a step (from what I can see), so it’s tough not to see him producing at the rate he has for several more years.
Projetion systems work really well with groups of players. Look at the Marcel system – you couldn’t build a rougher system, and yet it projects groups of players spectacularly well.
With individuals, there’s more variance. There’s even more variance with guys whose value is tied up in really volatile stats, like batting average.
When Ichiro doesn’t have an exceptionally high batting average, his OBP becomes middling and his SLG% drop well below what you want from a corner outfielder getting on at a 35% rate. I’d agree that he doesn’t appear to be losing a step, but PECOTA’s probably seeing his dramatic drop in batting average and not expecting much more out of him. When he bats .308, he doesn’t possess the power or on-base skills to be particularly useful. What makes him good is that he’s liable to bat well better than .308 most years.
C. Joseph (15) and e.c. (17): If Ichiro as VERY mediocre when hitting 300, why was he 6th for all ML RFs in VORP last year, with an EQA well above the league average for that position?