SubscribeJohn left this comment on one of my recent posts about SGPs:
Mays, I’m sure you’ll see, if you haven’t already, that Tim at RotoAuthority has released his projections and dollar values based on the SGP method.
It seems like a good opportunity to test your methodology against SGPs. I for one, would be curious to see how they compare.
The dollar values John is referring to are here. And since he asked, let’s look at what happens when I run Tim’s projections through the Price Guide.
When I built fantasy dollar values with the Price Guide using the exact same league settings that Tim uses for SGP, I get pretty much what I expected: Very little difference.
By my count, for the 418 players Tim projects, the Price Guide’s values are within $2 on 290 of them. I count 354 players that are within $3 (85%). When the Price Guide builds values for the exact same fantasy league that SGPs were designed for, you should notice very little difference.
But where do the methodologies diverge? One of the biggest differences was Ubaldo Jimenez, projected for 4.49 ERA, 1.51 WHIP, 10 W, 155 K, 182 IP. Tim’s SGPs put him at $1; the Price Guide says -$7. The SGPs really like his win and strikeout totals, enough that they will overlook some pretty nasty ratios. For the Price Guide, the negatives in ERA and WHIP outweigh any positives in the counting stats. SGPs love the counting stats, but the Price Guide hates the rate stats.
The only other player that they are that far apart on is Ben Sheets, whom the Price Guide has at $6, and RotoAuthority puts at -$2 — another $8 difference. Accounting for surgery that will cause Sheets to miss at least half of the year, Tim projects a line of 3.60 ERA, 1.21 WHIP, 5 W, 60 K, 75 IP. This is the same situation as Jimenez, except in reverse! The Price Guide loves the good rate stats; SGPs hate the bad counting stats.
Basically, the Price Guide’s strategy is to not chase wins and strikeouts, because doing so costs you ERA and WHIP. It advocates drafting middle relievers like Hideki Okajima ($1 to RA’s -$3), Scot Shields ($2 to RA’s -$1), and Edwar Ramirez ($1 to RA’s -$2). It says that’s what people should do.
Standings Gain Points show you what people actually do. Namely, they will draft a starter with a 1.51 WHIP if it can get them a few extra K’s.
Keep in mind that all of this is talking about a league with the exact same settings as the SGP dollar values were built for. But what happens if you try to apply SGP values to an even slightly different league?
Related posts:
Mays,
This is great! I find your conclusion especially fascinating, that the Price Guide shows what you should do, but SGPs reflect what people actually do.
This is a great insight in constructing a pitching staff. I appreciate your willingness to test this out!
When Tim first published his pitcher dollar values, at first glance they all looked quite inflated. Sure enough after he posted his entire spreadsheet, I realized his hitting/pitching split was 58/42! No leagues will ever spend 42% of their money on pitching so my initial suspicion was correct that all his pitchers are overvalued.
I believe Tim said that he doesn’t actually choose the split, the values just come out however they do. Is this another flaw of SGP or is this telling us what pitchers’ true value is, which is much higher than we pay for during the auction? For the method I use, I actually choose the split and go with 67/33 because my league always spends about that and most other leagues do too.
Mike – Mays and Tim both let the projections and the system figure out the proper hitting/pitching split depending on the player values it computes. Mays explains his here: http://www.lastplayerpicked.com/examining-the-7030-split/.
I believe that they come up with a higher pitching split because projections aren’t knowledgeable about the increased injury and luck risk inherent with pitchers. While projections can give you a number that it thinks a player will reach on average, it is unable to account for the fact that there is greater variance in pitching projections than there are in hitting projections (due to BABIP, LOB%, HR/FB, defense, bullpen, etc). I believe this is also part of why actual drafters “undervalue” catchers as well, as opposed to their projected statistics; catchers have a great injury risk and therefore their projections have greater variance.
In order to reduce the risk of picking a dud who gets injured, people will simply pay less for catchers and pitchers. Even though a few lucky souls will end up with a steal if their catcher or pitcher stays healthy all season. People are simply risk averse in will pay less for riskier players even though on average they will perform up to their projections. However you only have one team and one season so you don’t want to be that guy who ends up in last (because all of your risky players got injured/unlucky) even though you might end up in first (because all of your risky players stayed healthy/lucky). I guess where you take these risky players will ultimately depend on your risk preference. Do you invest in stocks or us bonds?
I think that one of the reasons that SGPs undervalue pitching rate stats is that in order to determine the value of a good ERA/WHIP in an SGP system is that you have to assume that each owner has the same number of IP. This is because in order to evaluate a rate stat you need to first turn it into a counting stat. In an SGP system, you have to find the number of ERs and BBs + Hs and then find a (negative) value for each ER and BB or H. So if you assume that each team puts out 1200 IP, then the number of ERs and BBs + Hs is large so the value of a good (or bad) ERA is diminished.
The Price Guide does things differently. It doesn’t assume a certain number of IP so the way that it values ERA and WHIP is based on whether or not a pitcher is better or worse than average and weights that by the number of IP that pitcher is projected for. In other words, a good way to construct your pitching staff is to get a high quality ace or two, an elite closer or two, and then a bunch of low ERA/WHIP, high K rate middle relievers with win vulture potential and get by with the minimum number of IP you can. This is what the price guide recommends. You can also throw a bunch of innings and rack up the counting stats and hope you get an above average ERA/WHIP, which you have a good chance of doing if you can pick favorable matchups to go with an ace and a closer or two. Since ERA/WHIP is often clustered very close in final standings, this is an approach that a lot of people do since it allows you to try to maximize your counting stats while hoping you get slightly better than average in ERA/WHIP.
If you were in a league with no (or limited) transactions, the price guide style is clearly superior. In a real situation, the price guide style will likely leave you struggling in W, K, and SVs and you have to sacrifice some ERA/WHIP to compete in the counting stats.
To test this theory, I would bet that a hitter with low counting stats and a high average is affected in the same way. Take a look at Joe Mauer, Dustin Pedroia, Placido Polanco, and Christian Guzman in both systems and see if there’s a greater than average dollar difference there.
@Mike: FYI, the Price Guide gave a 64/36 split for the RotoAuthority projections.
@Molson: I almost mentioned Mauer, because he was the only hitter where there was a greater than $4 difference. For the players you mentioned:
Player: $RA, $PG
Mauer: $27, $22
Pedroia: $22, $20
Polanco: $8, $6
Guzman: $3, $2
So SGPs were actually consistently a bit higher on BA specialists than what the Price Guide thought. So I think your theory for pitching sounds good, but I don’t know if the batting data confirms it.
What was the average innings that the price guide ended up with when using the Tim’s projections?
I’m curious as to if the differences would be lessened further if there was an innings minimum imposed on the price guide, say 1200 IP? That’s what my league uses, anyway…
@Nick: It came up with 1239 IP per team.
The Price Guide has no awareness of IP minimums.
Interesting. That’s more than I thought it would have.
I know the PG doesn’t understand minimums. I meant changing the SP/RP/P numbers until you get to around 1200 innings. But I guess you were already there!
To clarify: I entered 6 SP and 3 RP to match the number of starters and relievers to what Tim had. (I think he actually had about 6.5 SP and 2.5 RP per team.)
Otherwise, as you anticipated, the IP would have been way too low.