In a comment on a previous post, Brian asked about RP usage in ottoneu 4x4 leagues. I started to break down the strategy I used in past seasons and decided I needed more space than a simple comment, hence this post.
Before diving in, here is a quick primer on the RP context of ottoneu 4x4: As with all ottoneu leagues, you have five RP and five SP slots, and you are scored on ERA, WHIP, HR/9 and K. You are allowed 1250-1500 innings pitched over the season.
Brian's comment suggested relievers are overpriced and that he was better off investing in starting pitching - he even goes so far as to leave two RP slots open on his team. Interestingly, this is completely counter to what I had done in the past - making sure I had 5 very good RP to fill my RP slots AND finding a couple guys who were relieving by qualified at SP, allowing me to start 7 RP on a daily basis.
So who is right? Well, I don't have much detail on how Brian came to his conclusion, but I can at least walk you through my rationale and let you decide if it makes sense. And Brian, please feel free to post any counter-points you have - if I can do something to improve my team in the original ottoneu, I will be happy to try it out!
The first assumption I have been working with is that, when wins and saves don't count, RP are better on an inning-by-inning basis than starters. But I never really tested this assumption, so I devised a simple comparison based on 2010 stats. Using all "qualified" relievers and starters on the FanGraphs leader boards, I came up with the following numbers (SD= standard deviation):
Hardly a detailed analysis (covers 91 starters and 134 relievers) but this seems to verify my initial assumption that RP are, in fact, better per inning. But it also shows that they are more volatile. This all seems pretty straight forward, actually.
For a more specific example, let's take a look at a Player A/Player B comparison (credited to Joe Posnanski, who is, for my money, the best sports writer out there today):
Player A: 3.37 ERA, 1.16 WHIP, .56 HR/9, 8.79 K/9
Player B: 3.34 ERA, 1.14 WHIP, .56 HR/9, 8.49 K/9
Incredibly similar across the board in the ottoneu 4x4 stats. If you were offered 50 innings from one of these guys or 50 from the other, you really couldn't go wrong - B gets you .2 fewer runs, 1.7 fewer hits and walks, and one fewer K. Not much to write home about.
Well, you might imagine that since I am comparing RP and SP, I picked a RP and a SP and...you're right! Player A is one of the top earning pitchers in ottoneu - Justin Verlander. He is tied with Cliff Lee for the 5th highest paid player in 4x4 leagues at $33 on average and he has cost as much as $44. Player B is the immortal Kyle Farnsworth. He is going for, on average, $2 with a max of $3.
Again, this seems to confirm that RP put up better stats per inning than SP (no one would deny that Verlander is a better starter than Farnsy is a reliever, right?). Of course when comparing value over a season, you can't just look on a per-inning-basis. Looking at those same sets of pitchers, the starters threw 199 innings on average last year and the relievers threw 63. This means that you get roughly three times as many innings from a starter as a reliever.
Well, even if you expect Verlander to net you 4x as many innings, he gets paid 16.5 times more than Farnsworth. You need to calculate in some kind of a risk premium on RP, due to the higher volatility, and that brings down Farnsworth's salary. You could also say he is less likely to repeat his numbers. But 16.5 times more salary is a lot.
If you don't like Farnsworth, the highest paid reliever is Carlos Marmol at $12.60 on average. That is just over 1/3rd of what Verlander gets paid, in 2010 Marmol had a 2.55 ERA, 1.18 WHIP, .12 HR/9 and 15.99 K/9 - undeniably better than Verlander on a per-inning basis.
So what does that tell me? RP are an undervalued commodity. For $32.40, on average, you can have Marmol, Heath Bell and Joakim Soria. Those three should almost definitely have better stats than Verlander in similar total IP (213.1 last year compared to Verlander's 224.1). Last year Santiago Casilla had an ERA under 2 and more than a K/IP and my guess is you can have him for $1 right now if you want him.
The hard part with relievers is figuring out who to go after, due to the high volatility, but if you hit the right guys, you can put together a complete bullpen for the price of a single ace starter that well outperforms that starter.
To do this, first and foremost, I say you pay the premium for star relievers who are sure things. Heath Bell falls into this camp. So does Mariano Rivera. Hong-Chih Kuo should be in here, too. Soria as well. Feliz and Marmol probably belong. Beyond that it gets tougher, but the prices drop a lot, too.
Next, scour the waiver wire for cheap relievers that are performing well are have historical peripherals that suggest great ability. RP are volatile in part due to small sample sizes on a yearly basis, so a guy who regularly has a good K/9, decent BB/9 and is putting up a good year may be a good bet. And for $1, you really can't go wrong.
Finally, keep in mind that a lot of relievers will put up ERA's below about 3.50 (68 of the 134 from the stats above which is more than the total number of RP slots in an ottoneu league), while very few SP will (31 of the 91). If you can start your share of those RP - which is actually almost 6, rather than just 5 - you are going to help your numbers. Finding guys who qualify at SP but act as relievers can be a big boost - Sean Marshall was a huge value for me last year, as was J.P. Howell a few years back.
Think about it this way: if everyone in your league starts 5 RP from among that set of 68, there will be 8 RP with sub-3.50 ERA's remaining. Assume your pitchers are 100% average in sum, and you max out your IP. Now assume that instead of just 5 of those relievers, you find 2-3 who qualify at SP and start them as well. Over the course of the season, here are your potential final stats based on the number of relievers you use:
Note that in some cases, your WHIP would go down by .01 - it actually drops by about .003, which could make a difference. But comparing Brian's 3 RP plan to my 7 RP plan, you shave .05 off your ERA, .01 off your HR/9 and add 34 K's.
All of this does rely on your ability to identify the right RP, which is a more difficult task than identifying the right SP, I think, but even factoring that in, you can really improve your pitching stats by taking advantage of what appears to be a market inefficiency.
Before diving in, here is a quick primer on the RP context of ottoneu 4x4: As with all ottoneu leagues, you have five RP and five SP slots, and you are scored on ERA, WHIP, HR/9 and K. You are allowed 1250-1500 innings pitched over the season.
Brian's comment suggested relievers are overpriced and that he was better off investing in starting pitching - he even goes so far as to leave two RP slots open on his team. Interestingly, this is completely counter to what I had done in the past - making sure I had 5 very good RP to fill my RP slots AND finding a couple guys who were relieving by qualified at SP, allowing me to start 7 RP on a daily basis.
So who is right? Well, I don't have much detail on how Brian came to his conclusion, but I can at least walk you through my rationale and let you decide if it makes sense. And Brian, please feel free to post any counter-points you have - if I can do something to improve my team in the original ottoneu, I will be happy to try it out!
The first assumption I have been working with is that, when wins and saves don't count, RP are better on an inning-by-inning basis than starters. But I never really tested this assumption, so I devised a simple comparison based on 2010 stats. Using all "qualified" relievers and starters on the FanGraphs leader boards, I came up with the following numbers (SD= standard deviation):
| Stat | SP Avg | RP Avg | SP SD | RP SD |
|---|---|---|---|---|
| ERA | 3.78 | 3.47 | .76 | 1.14 |
| WHIP | 1.28 | 1.27 | .13 | .22 |
| HR/9 | .89 | .81 | .29 | .44 |
| K/9 | 7.0 | 8.1 | 1.45 | 2.15 |
Hardly a detailed analysis (covers 91 starters and 134 relievers) but this seems to verify my initial assumption that RP are, in fact, better per inning. But it also shows that they are more volatile. This all seems pretty straight forward, actually.
For a more specific example, let's take a look at a Player A/Player B comparison (credited to Joe Posnanski, who is, for my money, the best sports writer out there today):
Player A: 3.37 ERA, 1.16 WHIP, .56 HR/9, 8.79 K/9
Player B: 3.34 ERA, 1.14 WHIP, .56 HR/9, 8.49 K/9
Incredibly similar across the board in the ottoneu 4x4 stats. If you were offered 50 innings from one of these guys or 50 from the other, you really couldn't go wrong - B gets you .2 fewer runs, 1.7 fewer hits and walks, and one fewer K. Not much to write home about.
Well, you might imagine that since I am comparing RP and SP, I picked a RP and a SP and...you're right! Player A is one of the top earning pitchers in ottoneu - Justin Verlander. He is tied with Cliff Lee for the 5th highest paid player in 4x4 leagues at $33 on average and he has cost as much as $44. Player B is the immortal Kyle Farnsworth. He is going for, on average, $2 with a max of $3.
Again, this seems to confirm that RP put up better stats per inning than SP (no one would deny that Verlander is a better starter than Farnsy is a reliever, right?). Of course when comparing value over a season, you can't just look on a per-inning-basis. Looking at those same sets of pitchers, the starters threw 199 innings on average last year and the relievers threw 63. This means that you get roughly three times as many innings from a starter as a reliever.
Well, even if you expect Verlander to net you 4x as many innings, he gets paid 16.5 times more than Farnsworth. You need to calculate in some kind of a risk premium on RP, due to the higher volatility, and that brings down Farnsworth's salary. You could also say he is less likely to repeat his numbers. But 16.5 times more salary is a lot.
If you don't like Farnsworth, the highest paid reliever is Carlos Marmol at $12.60 on average. That is just over 1/3rd of what Verlander gets paid, in 2010 Marmol had a 2.55 ERA, 1.18 WHIP, .12 HR/9 and 15.99 K/9 - undeniably better than Verlander on a per-inning basis.
So what does that tell me? RP are an undervalued commodity. For $32.40, on average, you can have Marmol, Heath Bell and Joakim Soria. Those three should almost definitely have better stats than Verlander in similar total IP (213.1 last year compared to Verlander's 224.1). Last year Santiago Casilla had an ERA under 2 and more than a K/IP and my guess is you can have him for $1 right now if you want him.
The hard part with relievers is figuring out who to go after, due to the high volatility, but if you hit the right guys, you can put together a complete bullpen for the price of a single ace starter that well outperforms that starter.
To do this, first and foremost, I say you pay the premium for star relievers who are sure things. Heath Bell falls into this camp. So does Mariano Rivera. Hong-Chih Kuo should be in here, too. Soria as well. Feliz and Marmol probably belong. Beyond that it gets tougher, but the prices drop a lot, too.
Next, scour the waiver wire for cheap relievers that are performing well are have historical peripherals that suggest great ability. RP are volatile in part due to small sample sizes on a yearly basis, so a guy who regularly has a good K/9, decent BB/9 and is putting up a good year may be a good bet. And for $1, you really can't go wrong.
Finally, keep in mind that a lot of relievers will put up ERA's below about 3.50 (68 of the 134 from the stats above which is more than the total number of RP slots in an ottoneu league), while very few SP will (31 of the 91). If you can start your share of those RP - which is actually almost 6, rather than just 5 - you are going to help your numbers. Finding guys who qualify at SP but act as relievers can be a big boost - Sean Marshall was a huge value for me last year, as was J.P. Howell a few years back.
Think about it this way: if everyone in your league starts 5 RP from among that set of 68, there will be 8 RP with sub-3.50 ERA's remaining. Assume your pitchers are 100% average in sum, and you max out your IP. Now assume that instead of just 5 of those relievers, you find 2-3 who qualify at SP and start them as well. Over the course of the season, here are your potential final stats based on the number of relievers you use:
| RP Used | ERA | WHIP | HR/9 | K |
|---|---|---|---|---|
| 1 | 3.77 | 1.28 | .89 | 1175 |
| 2 | 3.75 | 1.28 | .88 | 1184 |
| 3 | 3.74 | 1.28 | .88 | 1193 |
| 4 | 3.73 | 1.28 | .88 | 1201 |
| 5 | 3.71 | 1.28 | .87 | 1210 |
| 6 | 3.70 | 1.28 | .87 | 1219 |
| 7 | 3.69 | 1.28 | .87 | 1227 |
| 8 | 3.67 | 1.28 | .86 | 1236 |
Note that in some cases, your WHIP would go down by .01 - it actually drops by about .003, which could make a difference. But comparing Brian's 3 RP plan to my 7 RP plan, you shave .05 off your ERA, .01 off your HR/9 and add 34 K's.
All of this does rely on your ability to identify the right RP, which is a more difficult task than identifying the right SP, I think, but even factoring that in, you can really improve your pitching stats by taking advantage of what appears to be a market inefficiency.
I had figured the same thing you had early on - that on an inning-by-inning basis, over the course of the major leagues, relievers would probably average out to be a little more effective than starters.
ReplyDeleteThere's a point where you say that relievers are an undervalued commodity, and I agree to an extent. Their undervalued as a single asset. But as a collection, they're valued close to what they should be. Let's go back to your comparison of Verlander vs. Marmol/Bell/Soria.
There's no arguing that the Voltron pitcher is a much, much better pitcher than Verlander. But there are two considerations I have to keep in mind:
(1) Voltron takes up 3 roster slots, Verlander takes up 1.
(2) Voltron will appreciate by $6 next year, Verlander will appreciate by $2.
In 2011 - you may be paying $32.40 for Voltron's production compared to Verlander's at $33. But in 2012, you'd be paying $38 to Verlander's $35. And in 2013, you'd be paying $44 to Verlander's $37. I'm not saying that it's not worth the difference in price, but I do think that (and the loss of two roster slots) is worth thinking about. When tying that to reliever volatility, I think that this "proven-closer" method has it's flaws as well as its benefits.
But when you talk about using $1-$3 relievers in the same way...and perhaps using them on a one-year basis, then tossing them...I see the wisdom of the approach. You're right in that the trick is finding high-peripheral guys that are poised for success. And you're right in saying with the small sample sizes and year-to-year changes, this could be a great deal more difficult than finding reliable starters.
For reference, here's what my idea was in the beginning: get 5 "elite" starters, a couple of backups, and fill out my RP slots with really good $1 relievers - the same ones you've spoken of. If I didn't spend more than $5 on relievers (I spent $8), and didn't use up too much of my roster space, I'd have more money to spend on a few really great starters, plus I'd have more roster slot flexibility to stock up on prospects.
I'll admit that my draft of starters didn't go quite as planned, but I wound up with seven starters that don't embarass me, and I wound up spending a total of $107 on them. After cutting a couple of my relievers (I made a few mistakes early on), I wound up with ten active pitcher slots (swelled to 11 when Britton was called up), and only $115 (plus $5 for Britton) spent on active pitching.
I still think that if I would've executed my plan a little better, it would provide me with the best chance to win both now and in the future - but I'm going to seriously consider this 6-7 reliever setup going forward. I can't deny that the reliever advantage in the rate stats is substantive - the trick is not spending too many roster slots on them, nor spending so much cash that I can't improve my team in other areas.
Do you mind sharing how much you're spending on these relievers? I'd like to see how your budget figures out.
Thanks again!
I agree that on a per-inning basis, RP perform better than SP. But, I went with a strategy of targeting cheaper RP rather then elite ones given the salary disparity between the Bells/Marmols of the world and the next couple of tiers, although the volatility is higher as you move down the list.
ReplyDeleteI ended up with Bailey, K-Rod, Valverde, Hensley, Marshall, and Howell after the draft for $14 and added Crow last week for $2. The problem is that none of these guys qualify at SP, which will only surface once Bailey and Howell return.
Adding these to my starting staff of Kershaw, Lee, Price, Danks, Cahill, Sanchez, Kuroda, and Britton (total cost of $116), and I like my chances to have a top-3 staff in my league, if not the best.
By the way, where can you go to see the average player salaries across all ottoneau leagues? That would be very useful info.
ReplyDeleteZach, you can get those numbers at ottoneu.fangraphs.com/averageValues. Also, I would keep an eye out for SP/RP qualified relievers and consider picking them up and cutting one or two of the seven guys you currently have. Every inning you get out of one of the dual-role relievers is basically one less inning from your worst starters. I'd think a few mediocre relief innings would be a nice replacement for when Britton has 2-3 starts each against the Red Sox and Yankees.
ReplyDeleteBryan, my bullpen has been a debacle this year. Thornton has been terrible, Soriano has not been much better and Saito is hurt. But in past years, I have been able to snag guys like Sean Marshall or J.P. Howell for $1. Typically the guys who qualify at both roles are either prospects who aren't quite ready or failed starters. The former tend to already be owned and not that cheap (see Feliz a couple years back), but the latter tend to be free agents. Going back to last year, I was spending $8 on Bell, $5 on Thornton, $4 on Soriano, $5 on Billy Wagner, and $1 on Saito, along with Marshall for $2 and a rotating cast of characters including the likes of Jose Contreras for $1 or $2 as my swing guys. That's a $27 bullpen (probably more like $30-$31 if you include cap penalties on guys I tried and dumped) which probably gave me about 450 IP with very very good stats. Verlander, on the other hand, cost me $42 (after a trade) last year. Lincecum cost $60. So for half the price of Lincecum and 2/3rds the price of Verlander, I had 450 excellent innings.