March 3, 2017
PECOTA, the PFM, My Bid Limits, and You
This column is designed to address the questions I commonly get about my published bid limits here at Baseball Prospectus, how they’re different from the prices in our Player Forecast Manager (PFM), and most importantly why they are useful. It would be impossible to address every question our readers have had about my pricing modeling versus what the PFM is and what it does, so I’ll start with a few of the most common questions our readers have had about both.
Are you planning on using the awful Q&A format for this article?
I wasn’t before, but I am now, tough guy.
(5,000 words of heated argument with Mike and the imaginary questioner REDACTED.)
Let’s start with the basics. What is the PFM? How do I get started?
Pictured above are what the PFM default settings look like. Figure A shows where you can change settings for how the PFM allocates money for a variety of factors including inflation, hitting/pitching dollars, size of your fantasy league, and player pool. Figure B shows where you can alter roster composition and your league’s categories. Most of the content in this piece will focus on Figure A. If you are in a 9x9 league with 35 player active rosters, you are better off using the PFM than my bid limits and can stop reading. Most of my work is geared toward 5x5 “traditional” Rotisserie formats.
Returning to Figure A, for redraft leagues like the “expert” leagues I turn the inflation settings “off” and turn the SGP setting “on”. I then adjust the league and number of teams as needed. Then I click “Submit PFM Parameters” and within seconds a list of prices is generated for whatever league format I am using.
That sounds so easy! Why don’t you just use the dollar values in the PFM for your auction?
To understand why I don’t recommend using the PFM, it is important to take a few steps back and understand what the PFM is and how it derives its values.
The statistics in the PFM are from the 50th percentile projections from PECOTA. The projections are reasonable baselines of what each player can do and the earnings assigned in the PFM are as well (especially if you turn on the SGP feature, as recommended above). However, while the baselines are reasonable, bidding solely using these baselines is a recipe for a sixth-place finish.
The auction marketplace does not behave the way the PFM or any other projection model that assigns dollar values recommends. You are not bidding against the PFM, but rather against 11 or 14 other fantasy managers in your specific auction. Can you swim against the current and build a team using the PFM? Yes, but I believe that you are putting yourself at a tactical disadvantage by doing so.
To put this as simply as possible, the PFM is grading everyone on the curve. The expert marketplace is not building a curve into its bids.
Table 1: Top 10 PFM Projected 2017 versus Top 10 PFM Earners in 2016, AL-only
The expert market isn’t modeling their bids on future projections. Rather, it is building its prices based on historical performance of the Top 10 players overall. The experts don’t know who the 10 best players are going to be, but they’re paying the Top 10 players as if they do know. The PFM is correct about regression overall. The 10 most expensive players in 2016 (the CBS/LABR/Tout column) earned $29.58 per player. The expert market took a six-dollar loss on the 10 most expensive players. I’d guess if you asked most of the experts if they were OK with this, they would answer affirmatively.
Am I missing something? At the risk of stating the obvious, isn’t losing money on your investment bad?
Of course it is. To continue stating the obvious, you are better off getting $40 of stats from a $40 investment than you are getting $30 worth of stats. In this hypothetical example, the fantasy manager who gets the $40 ROI has an advantage over the fantasy manager who gets the $30 ROI. But the ROI isn’t linear. The winning bidders on Altuve and Betts both turned a profit. The winning bidder on Bautista took a big loss. The other seven players in last year’s top 10 landed somewhere in the middle.
When you bid on a player, there is the goal of trying to buy that player. There is a secondary goal, however, which is to prevent the other fantasy managers in your league from purchasing that player at a favorable price. The market pushes everyone’s prices high enough so that none of the perceived best players are bargains. Even with the market’s best efforts, a couple of bargains will occur (as with Altuve and Betts last year) but overall the market succeeds in making sure that nearly all the top players are losses.
If the market did decide to start behaving like the PFM in 2016, I would change my bid limits accordingly. But if only one team starts behaving like the PFM, it doesn’t impact the market at all. In fact, it makes it very difficult for the team that decides to use the PFM to purchase any players.
How so? Is the PFM incorrectly budgeting its auction dollars?
In terms of balancing the ledger, no. Just like my bids do, the PFM’s top 276 bids add up to $3,120 in 12-team leagues and its top 345 bids add up to $3,900 in 15 team leagues. If you follow the PFM to its logical conclusion, you will spend all your auction money. The problem comes with how that money is allocated.
I don’t necessarily mind that the PFM is telling you to avoid the top ticket items. My bids are somewhat more conservative than what LABR and Tout are paying and I often finish auctions without having a player over $25 on my roster. Tout Wars NL was an extreme case of this last year, where I didn’t spend over $17 on a player. It takes discipline, but you can build your roster this way if you choose to do so.
Where the PFM fails in auctions that are like the LABR and Tout auctions are in two areas:
1) Scarcity is expensive
Do you remember that Dyson projection above? Twenty-seven dollars in earnings is reasonable for Dyson if he matches his PECOTA projection. It is also a lot to spend for a hitter who has never logged more than 337 plate appearances in a single season. Dyson went for $16 in the CBS auction last month. He will probably won’t crack $20 in the LABR auction tomorrow night in Arizona. While any smart fantasy player recognizes Dyson’s ceiling, the risk of paying $27 for someone who could become a part-timer if he doesn’t hit is considerable.
This phenomenon extends to closers as well. If the market for closers collapses (as it has in the expert leagues the last few years) it doesn’t do you any good for buy four closers for $15 apiece just because they are “bargains”. You don’t want to pack your roster with excess value in two categories. You want to win your league.
2) The players at the bottom are “too expensive” in the PFM
Just like the market pays the players at the top a little bit more than what they are projected to earn, it cheats the players at the bottom of the player pool. In mixed leagues, this is because there is a good deal of talent that typically costs one dollar at the end of an auction that is interchangeable with the free agents who aren’t purchased. Even in mono leagues, the “free talent” that isn’t purchased at auction often surpasses the value of the players purchased in the only league endgame for $1-3.
The PFM isn’t attempting to account for this phenomenon. It is simply assigning value to players based on PECOTA’s 2017 projections. We know that many of the players projected for modest value in the PFM will disappear or have subpar years that will put them outside of the top 276 or top 345. This is a case where human intervention is a good idea.
Can Scooter Gennett earn $6.06? Sure, but I can get that kind of production from many in-season pickups. I’ll pay a dollar…that is, if I even must own Scooter Gennett on my fantasy team. The same can be said for Thomas Pham ($5.19), Cory Spangenberg ($4.96), and Andre Ethier ($4.13). Daniel Nava ($4.02) probably won’t make the Phillies, but even if I were metaphysically certain that Nava would crack the Opening Day roster I would not pay more than a dollar.
Most fantasy players direct the money that the PFM says that Gennett, Pham, Spangenberg, and Ethier “should” get elsewhere. Often, it is directed to the top players. My methodology directs some of it toward the top but also to players who I like in the second and third tier.
Let me guess. The entire point of this 2,000-word article is that your bid limits are correct.
No! The bid limits are merely a guideline for your auction. I happen to think that they are a reliable guideline, but they are not the final word in what you should pay for a player.
If you’re busy and don’t have time to put any prep work in, I think you can use my bid limits and do well in your league. But if you do have the time to tweak my bid limits, I strongly urge you to do so. Every league is different. Each league has anywhere from 10 to 15 different personalities who will approach their auctions differently. You know your leagues better than I do.
Does your league put more value on saves than my bid limits do? Add some money to your closer pool. Do you think my bid on Matt Carpenter is too conservative? Push some money toward Carpenter. Are you in a keeper league where rookies are at a premium? Raise your bid limits on rookies; you don’t want to get caught in July without chips to move for the following season. Just don’t forget that for every player you add money to you must take money from others.
The bid limits are a better guideline than the PFM for conventional 5x5 auctions. This does not mean that they are the final say. I’m glad if you use my bids as a starting point but then move players up or down based on your league or your preferences. I started out using someone else’s bids before I eventually developed the confidence to use my own prices. Questioning my process – even if you agree with it completely – is a great first step toward becoming an expert player in your own right.