SubscribeWith Jose Valverde signing with the Tigers and Octavio Dotel apparently heading to Pittsburgh, the closer situation for most teams has been settled.
You may have noticed, however, that one of the most glaring difficulties with using computer-generated projections (CAIRO, CHONE, etc.) for fantasy baseball is the lack of saves in these projections. Recognizing that saves are much more tied to a player’s situation than to his skills, most computerized projections make no attempt to quantify the number of saves for each player.
While that represents the reality of projecting, it obviously won’t work for fantasy, which allocates 20% of the pitching points based on saves. But what can be done to fix this?
I was hoping that the FanGraphs Fan Projections would be helpful for this, but there are still lots of pitchers (Joe Nathan, for one) who don’t have enough projections to show up. As it is, there’s at least enough data to pick an appropriate scale — about 40 saves for the top closers and 15-20 for the marginal ones.
So I have attempted to come up with my own save projections. I started off with a list sorted by the number of saves from the past two years, averaged with a 2:1 weighting. From there, I tweaked the rankings to account for changes in situations and a rough estimate of skill. The results are four basic tiers of closers:
40 Saves
Joe Nathan (47 S in 2009, 39 S in 2008)
Mariano Rivera (44, 39)
35 Saves
Jonathan Papelbon (38, 41)
Francisco Rodriguez (35, 62)
Brian Wilson (38, 41)
Francisco Cordero (39, 34)
Joakim Soria (30, 42)
Heath Bell (42, 0)
Papelbon and K-Rod are near the 40 save tier, but I think they’re just a shade below the top two. Bell and Soria both require some bonus credit to get ranked this highly, compensating for Soria’s missed time in 2009 and Bell’s single year track record.
30 Saves
Trevor Hoffman (37, 30)
Jose Valverde (25, 44)
Jonathan Broxton (36, 14)
David Aardsma (38, 0)
At this level, everyone has question marks. We’ve got injuries (Valverde, Francisco, Qualls), age-concerns (Hoffman), and one-year samples (Broxton, Aardsma) clouding the situation. These guys are still pretty solid, though.
25 Saves
Brian Fuentes (48, 30)
Brad Lidge (31, 41)
Ryan Franklin (38, 17)
Bobby Jenks (29, 30)
Huston Street (35, 18)
Matt Capps (27, 21)
Rafael Soriano (27, 3)
Chad Qualls (24, 9)
Frank Francisco (25, 5)
Billy Wagner (0, 27)
Despite leading the league in saves in 2009, Fuentes was a little rocky and now has Fernando Rodney on his team. Lidge’s totals look good, but I’m also a bit concerned with his late-season breakdown. Capps and Soriano have some question marks for injuries. Basically, this is the tier where I’m not surprised if anyone loses the closer’s gig by the All-Star break.
20 Saves
Kerry Wood (20, 34)
Andrew Bailey (26, 0)
Brandon Lyon (3, 26)
Mike Gonzalez (10, 14)
15 Saves
Leo Nunez (26, 0)
Carlos Marmol (15, 7)
Scott Downs (9, 5)
Jason Frasor (11, 0)
Octavio Dotel (0, 1)
The 15-20 range are all really speculative; I expect a good number of these guys to lose the closing job by midseason. Scott Downs and Jason Frasor seem like the best candidates in Toronto, but I can’t find any information about who is more likely to get saves.
These are back-of-the-envelope calculations — not exactly mathematically rigorous but certainly better than 0’s for everyone. I’m also not watching these players nearly as close as the fans of their teams are, and so I’m hardly an expert on their situations and skills.
I’ve gone ahead and added them to the Price Guide as the default save projections for CHONE, CAIRO, and the composite projections. I’d still encourage you to edit the projections how you see fit, but I think this at least starts things off looking more reasonable.
However, I’d really like some help in refining these numbers. Anybody have any suggestions for improving these tiers?
Related posts:
As saves are second only to holds in terms of unpredictability, I think this will be more than fine.
Are you aware of any projections for save opportunities (per team) out there? The only suggestion I could make would be to factor in projected save opportunities. As an example, I think the the better offense and weaker defense in SF will result in higher scoring games. The higher the score of games, the fewer games have save opportunities. Thus, I have downgraded WIlson as a closer this year for 5X5 leagues.
Hi Mays,
These numbers are just estimates that were quickly added by hand, so they may be slightly inaccurate. A grain of salt is required.
The above projections for 2010 have 31 pitchers with 15+ saves, totaling 815 saves. 2009’s composite projections had only 21 such pitchers, who totaled just 470 saves. Last year’s total amount of saves projected (top 500 pitchers only, using all standard settings) reached 907, meaning that 48% of the projected saves came from sources below 15.
Now, I’m not suggesting that you go in and hand-add 1 save for every Russ Springer out there. The way it is now, though, top closers seem to be overvalued as this both slightly deflates the available amount of saves (projections above adding up to 815, as compared to 907 last year; in actuality, I counted 1200 saves that were recorded by an MLB pitcher last year!), but even more importantly, gives them a larger share of that total. In other words, their SD in saves too high, artificially inflating their value.
I read that you don’t want to use Marcel’s because it is a bit bearish on younger players. However, the Marcel’s on fangraphs has save projections. What if you were to run a query and have it add those projected saves to any player’s composite projection if he did not already have a save total (e.g. Mark Lowe gets 2 saves)?
I’m also of the opinion that the projected save totals for those 15+ closers are too high (c.f. the 470 vs. 815 count). Last season, only 1 pitcher was projected at 35, only 3 more at 31, and so on. I think that is probably a better model to follow by when it comes to projections, just as we only have 2 hitters this year projected to hit 40+ HR, even though the last 5 years the average has been over 6 players who actually did hit 40+ out of the park. We just can’t predict the variation, but can only use our best guesses based on talent level.
If you think Nathan and Rivera are the two best bets, I would put them at 35 each. Papelbon and Rodriguez could be around 31-33 each. A couple of the next at 30, and so on. That would still give the upper tier more than was projected from it last season, but it would also be significantly closer to that level.
Josh
I’d be curious to know what formula FanGraphs uses to derive Saves from the projections that the users input (because saves is not one of the “input-able” categories when you go to make projections).
Kind of along the same thoughts as #2, what would it do for the value of the Matt Thorntons & Chris Perezes of the world if they were allotted 5 saves for their projection. I’m thinking set up guys with good skills behind less reliable closers. Brandon League, Ryan Madson, JP Howell, maybe even Daniel Bard or Joey Devine. This might fall to far into the realm of speculation, but projecting a few saves might do more to assess these types true values.
@John: Last year, Derek Carty tried to see if anything correlated to a team getting more save opportunities:
http://www.hardballtimes.com/main/fantasy/article/are-saves-predictable/
Basically, save opportunities are very slightly correlated with wins, but not with anything else he tried (e.g. runs scored). In my rankings, I’ve put little or no value on team quality on offense or defense, and I’m not sure how much doing so would help.
@Josh: I didn’t mention it in the article, but I did add 1-10 saves for several other guys who are either a) good pitchers or b) in a good situation. For example, with Billy Wagner a somewhat risky proposition, I think I allocated 10 saves to Takashi Saito.
Even so, the projections won’t match reality (with lots of pitchers picking up a handful of saves each). I’m just guessing at what could be expected before the season, and I’d argue that saves for most of the sub-fifteen-save relievers couldn’t be expected in January.
You’re also right that my initial projections are coming in a bit higher than the Marcels, and I’m still not fully convinced my way is best. My thinking was that saves might not need to be regressed as strongly as the Marcels do, since they are such a situation stat.
That is, Marcel doesn’t understand the situation, and so it is trying (through regression) to account for what looks like a great deal of variation year-to-year. If we know the situations, then we understand some of the variation (e.g. Hoffman leaves SD and Heath Bell goes from 0 to 42 saves) and we can be more aggressive in making predictions.
@Nick: I did project a handful of saves to guys I didn’t mention above:
Fernando Rodney
Kevin Gregg
Takashi Saito
Matt Lindstrom
J.P. Howell
Jim Johnson
George Sherill
Brad Ziegler
LaTroy Hawkins
Ryan Madson
Joel Zumaya
I like Matt Thornton a lot and probably should have added him to that list. Brandon League is in a good situation in Seattle, but I’m concerned from his time in Toronto that he wasn’t trusted with closing duties even when the job was wide open.
Mays,
I think I better understand your reasoning behind the projections now. Certainly, the situation does dictate much of the results in terms of saves. However, I’d argue that this is already the reason why we’re giving large save totals to these pitchers in the first place (even in cases where they aren’t the best reliever on their teams).
I do think that Marcel’s does have a bit too much regression in this stat – and I would imagine they would have very conservative estimates on new closers in particular, although I haven’t looked into that. But at the same time, I don’t think the outlook needs to be quite as positive as the aforementioned numbers. Let’s take Rivera for instance. You have him at 40, and Marcel’s has him at 28. Over 13 years as a closer, he’s had 40+ 7 of those times, so it’s pretty much 50/50, historically.
My recommendation in the previous comment was to put him at 35. I think that leaves some room for the possibility of regression in the stat due to injury, ineffectiveness, fewer opportunities, etc. If we were purely thinking historically, I’d put down probably 38-39 (depending upon seasonal weights, etc.). However, I do think we need to temper the predictions a bit, just as would be done with other categories (even if not quite as much).
Nathan would be similar, I would probably have him down for about 39-40, without any regression, and 35 sounds fair using a modest amount of consideration for potential problems.
That’s great to hear that there are already some saves added for those non-closers. My main concern on this point isn’t whether Mark Lowe gets those 2 saves and fraction of a dollar added to his projected valuation, but rather, that the reality that there will be a certain amount of additional saves available can keep the value of the current closers in check.
All this said, I really don’t intend to be picky, and am thankful to be able to use your service (and the ideas behind it as well). At this point it essentially is nit-picking, but I do think if the values can be improved by even a dollar here and there, it would be another positive.
Josh
What if we did a historical study over the past 10 years or so, and found pitchers who had averaged 40+ saves over the previous two-year period, and who also began the following season as a closer. Would they still average 40+ over the next season, or would it be 38? 35? 30? I have no idea, but it may be somewhat helpful in figuring out how much “unforeseen” problems are affecting the average situation.
I’m not sure how to do something like this without going back manually and looking for potential matches.
Soria didn’t miss time in 2008, but he did miss some in 2009.
@rbt: You are right, and 2009 is what I meant to say. It’s fixed, now.
Did you do any more work on this? The FanGraphs projections now have a lot more data, and maybe could be incorporated into your projections.
@Matt: I’ve been tweaking the numbers continually since this post went up, based on input from readers and from sources like the Fan projections.
Unfortunately, now that FanGraphs has projections for everyone, some of the projections are already becoming dated. (38 saves for Joe Nathan is seeming optimistic at the moment….) Nothing major, but their numbers would still have to be tweaked.
Without one authoritative source, I’ll probably keep pulling from as many as I can and encouraging Price Guide users to utilize the “customize projections” feature.
True. Thanks for the update.