On Best Player Available vs. Position Scarcity
Draft season is fast approaching. That means that draft strategy will once again be a hot topic, as fantasy enthusiasts argue the merits of one philosophy versus another. But one particular debate threatens to make my head explode every single year. That is the notion of making your draft selection based on “best player available” (BPA) versus “position scarcity” (PS). The decision makes no sense to me. This is why.
The idea here is the fantasy owner must choose to employ one of these two strategies (or perhaps others) when making his pick in a snake draft. If the owner decides to go with BPA, he is simply selecting the best (most valuable?) player regardless of position. Though it’s not always made clear, the assumption is that best is determined without accounting for any sort of position adjustment. A .300-20-80-80-2 stat line is worth the exact same amount whether it’s projected to come from a catcher or a first baseman.
On the other hand, it seems that those espousing the PS strategy have their rankings in hand, but will purposely bypass a higher ranked player in order to select one from a so-called “scarce” position. Those positions, of course, are typically catcher, second base and shortstop. By executing this strategy, the players at a scarce position are given an arbitrary boost in value. I shall assume that fans of this strategy have a similar ranking list as the BPAers that ignores position, but makes those position adjustments in their head on the fly during the draft.
From my understanding, these are the two sides of the coin. I might very well be wrong and am therefore describing it incorrectly. This is because I don’t draft using this either-or scenario. And that is because the strategy of choosing either BPA or PS isn’t logical or mathematically sound.
Instead, BPA must incorporate position. Essentially, the strategy combines both. Actually, it’s not a strategy at all, it’s simply how dollar values should be generated and a resulting ranking list derived. In the second paragraph, I put “most valuable?” in parentheses and ended it with a question for a reason. When I define “best”, it’s equivalent to most valuable. And my most valuable player is going to account for the player’s stat line, of course, and his position. If your rankings and valuations don’t incorporate position, you’re only doing half the work.
The PS only strategy is similarly flawed. During the draft, there should be no need to give a player at a scarce position an arbitrary value boost. That value boost, if any, should already be accounted for in your dollar values, which would then be sorted in descending order to create your rankings. I’m by no means suggesting you then draft straight down your rankings list, obviously. But that’s an entirely different subject and one I discussed a year ago.
Let’s look at an example so we’re all clear on these various strategies. Assume a league with two teams, two positions required (catcher and 1B) and one category, home runs. Here is your player pool:
| Player | Projected HRs |
|---|---|
| 1B A | 45 |
| 1B B | 40 |
| Catcher A | 35 |
| Catcher B | 25 |
Following BPA, Owner 1 selects 1B A. He ignores position and simply selects the player expected to hit the most homers. Owner 2’s valuations incorporate position and so rather than following a BPA strategy which would lead to a selection of 1B B, he picks Catcher A. Owner 2 then picks again and takes the other first baseman, 1B B, as Owner 1 finishes the draft with Catcher B. Let’s see how they did:
Owner 1:
| Player | Projected HRs |
|---|---|
| 1B A | 45 |
| Catcher B | 25 |
| Total | 70 |
Owner 2:
| Player | Projected HRs |
|---|---|
| 1B B | 40 |
| Catcher | 35 |
| Total | 75 |
Surprise, surprise. Despite picking first, Owner 1 ended up with 70 expected homers, five short of Owner 2’s projected 75 homers. That’s because Owner 1’s rankings should have looked like this:
| Player | Projected HRs | Useful HRs |
|---|---|---|
| Catcher A | 35 | 10 |
| 1B A | 45 | 5 |
| Catcher B | 25 | 0 |
| 1B B | 40 | 0 |
Rather than comparing home run totals to every player, one must compare those totals to players at the position. You can’t draft a first baseman and slot him in as a catcher, right? So why are we comparing apples to oranges? In order to convert home runs between two positions into the same currency, we subtract out the replacement level’s expected contribution, or in this case, the last player picked. That leaves us with the player’s useful home runs, what will actually provide a positive contribution. Because what we care about is how many more home runs my catcher hits than yours does and my first baseman than yours. I don’t care if my first baseman hits more homers than your catcher because I can’t start my first baseman in the catcher slot. Again, apples and oranges.
Now let’s look at the PS strategy. Assume this time that this is your standard 12-team league with normal rosters and 5×5 categories. Below are the players sitting at the top of your rankings, with dollar values calculated that are not adjusted for position, as your pick is up.
| Player | Dollar Value |
|---|---|
| 1B B | 20 |
| 3B B | 19 |
| OF A | 19 |
| 1B A | 18 |
| SP B | 18 |
| 3B A | 17 |
| OF B | 16 |
| SP A | 16 |
| Shortstop A | 15 |
Following a PS strategy, you might think long and hard about drafting SS A. Sure, your values say he’s worth $5 less than your top player still on the board, 1B B, but he plays a scarce position, man! You then decide you don’t want to be left choosing between Jed Lowrie and J.J. Hardy and draft SS A. You give up $5 in value according to your calculations, with the thinking that you could still draft a decent player at 1B, 3B and OF later.
Was this the correct move? Maybe, maybe not. A shortstop’s stat line could very well deserve a boost of more than $5, which would have made the choice correct, or perhaps the true boost was just $2, which would have made the pick wrong. But you don’t know because you’re just guessing! How do you decide where to draw the line? When is the value gap too great to go for the scarcer position and just go with the top value? $6? $7? $10? A valuation system that accounts for position will do the math for you so there’s no need to guess. Nor will there be a need to give an additional position scarcity boost, because that has already been done by your trusty spreadsheet.
So in the end, the best strategy is technically a combination of both. Most of the time, you’re going to want to select the player at the top of your rankings, which many dub the BPA strategy. But that assumes your valuations are calculated correctly. And to be correct, they absolutely must account for the player’s position.
Mike Podhorzer is the founder of ProjectingX IQ, an advanced fantasy baseball analytics platform that transforms projection data and in-season performance signals into actionable intelligence. He is the 2015 Fantasy Sports Writers Association Baseball Writer of the Year and three-time Tout Wars champion. He is the author of the eBook Projecting X 2.0: How to Forecast Baseball Player Performance, which teaches you how to project players yourself. Follow Mike on X@MikePodhorzer and contact him via email.
What about statistic scarcity? I’m looking for big power at the top of the draft because it seems to come with too much downside later in the draft.
Yup, should also be factored into your valuations. Power is much scarcer now than it had been when I first started playing fantasy in the 90s.
I thought we had “The Position Scarcity Post To End All Position Scarcity Posts” 🙂
Using Tristan H. Cockcroft ESPN rankings he has Jose Reyes at 46th overall, Starlin Castro at 83rd, and Carlos Gonzalez at 45th. Assume that you agree with these rankings and that Reyes and Gonzalez are the top 2 on your board.
Do you draft Reyes because of the huge drop off to the next guy or do you stick with the BPA?
You remember that article! It’s impossible to give you a perfect answer without seeing the rest of the rankings, but sounds like you go Reyes, which follows he strategy I outlined in the article I linked to from last year. That’s not a position scarcity answer, it’s a value distribution within the position thing.
There was no need to look at any other position scarcity articles because it. 🙂 You should of named last year’s article: “Mathematical Principles of Natural Philosophy of the Snake Draft”
Is there a way to dynamically recalibrate values in the draft based upon who was drafted and/or who is expected to be drafted?
I believe draft software does that for you, but I’ve never used anything more than my spreadsheet with players ranked in descending order of dollar value.
Thanks.
I overall agree with all of this logic. I am not honestly sure what I do, but I find this all interesting. From posts the past few days, it seems the problem involves 2 things. 1) What if you already have a 1B but your next best player available is also a 1B? 2) What if players draft more SS than you expect, causing the value of your replacement SS to plummet. This could also happen if other SS are plucked off the wire, then yours gets injured, causing replacement SS to be worse than your initial valuation. In the latter scenario(s), I think there is confusion about how much value to boost SS, assuming it has a much steeper decline in replacement level compared to OF.
It won’t. There is very little differentiation at or near replacement value. One or two extra players won’t effect it much, if at all.
I posted this as a comment to Cwik’s article yesterday, but I really want to see everybody put their money where their mouth is with all the draft strategy/player valuation theory. Do a draft (or multiple drafts, because small sample size) where everybody will be scored using 2014 stats (or 2015 projections if you prefer). After the draft, add up everybody’s roster using standard 5×5 scoring. See who wins, and what kind of draft strategy really gets the best roster. See who loses, and mock them without mercy.
Not that I disagree with you – your points make sense. But I think this is something we can actually test and I’d love to read about the results, so why not give it a shot?
I read those comments and love the idea, but actually creating all the rules so there’s no subjectivity and then going through with the drafts seems far too difficult.
I agree that it’s a non-trivial amount of work, but what’s the challenge with the rules? The hardest thing I see is to figure out how to account for playing time, but that seems manageable – you get every game from your starters, plus prorated stats from your backups to get to 162 games. There’s some fudging necessary when you get to a Zobrist-type player you can swap in all over your lineup, but I think there’s an acceptable solution if you can accept some level of abstraction (basically, assuming none of your starters miss games on the same day). For maybe too much abstraction but great simplicity, you could use Steamer/600 and ignore backups altogether.
There are some biases – you can’t play matchups, stream pitchers, chase saves on the waiver wire, etc. Pitchers are overvalued relative to a real draft because there’s no volatility (to the point that an offense-only draft might make more sense). There are any number of elements of real fantasy that are missing, so this isn’t a perfect simulation of a real fantasy draft. But it’s still a relevant exercise in optimizing a roster across multiple categories and positions and answering some of these questions with data instead of theory.
If you find the people to do the draft (and it’s not like there’s a shortage of people willing to do mock drafts), I’ll volunteer to nail down the rules, arbitrate disputes, and tabulate scores.
Yes, I have spent way too much time thinking about this.
Email me. I’m interested in discussing further.
Email sent. Very interested in this.
If you want to get real crazy, if you have pre-2014 draft valuation sheets, use those, then play out the 2014 year. Otherwise you can plan an article for December 2015 and run a few sims now based on current cheat sheets, then in the downtime after the season, see how that would have yielded. As for players times, I would assume each player is used in his most “scarce” position. Another option instead of using actual played seasons would be to use the current values of the fangraphs depth charts.
Did this happen with the strategy mock draft? Would be very cool and would love to see the results.
Thanks
Not yet. Still discussing possibly doing it, but seems like a lot of work.
I am interested in the idea.
What if you just drafted hitters? Or automate where pitchers are being picked?
For example pitcher ranks
5. kershaw
10.Felix
15. Bumgarner
The person who picks 5th gets Kershaw, the person who picks 10th gets Bumgarner etc. Then you make other rules like 2 pitchers max in 5 rounds, 4 in 10 etc.
In reality, the best way may be to a self drafts with X numbers of teams using a certain strategy, Y number of teams using a different one etc. The best strategy may be very dependent on what other owners strategy is.
Including pitchers at all just adds unnecessary variation, just do an offense only draft. It isolates the question we’re actually interested in.
I posted something very similar on Zola’s Mastersball post a couple of days ago about the value of a kind of regression draft to test theory. 2014 stats would be the easiest way to go. A few years back there was an all-timer draft that happened using historical seasons on the NFBC message boards. As I recall, it ended up not being so much an exercise in drafting theory but morein game theory and analysis. The winner early on got a feel for the shape of the draft and started moving tactically behind a strong strategic analysis.
The game theory element is definitely present, and it’s hard to separate a good draft strategy from a skillful application of that strategy. But it’s somewhere to start, at the least. My thinking is that you learn something from the distribution of results, not just the overall winner, and ideally you’d look at multiple drafts to get more data points and start looking at trends instead of instances.
If you end up putting a draft together I’d love to participate. oaktownsteve@gmail.com
It probably will take many drafts using different years projections to find conclusions. It will be interesting to see if optimal draft strategy depends on the composition player pool.
I am wondering if it better to use a point system to eliminate drafting a player for a categorical need.
Don’t use projections! Just use season totals, keep the random variation down, projections just wind up telling us which team drafted players who underperformed and overperformed. We need to draft from absolute knowledge of player performance for this to work. Projections are important to the fantasy player because they tell us what to expect from a player, but we don’t need or want them in determining the optimal strategy, we want to know, if we had absolute knowledge of the accuracy of our projections, who would win the league, the best way to do that is through hindsight of a given year and see which strategy, given known outcomes is actually the best.
Have you ever tried using tiers in your personal rankings? I found it very useful for football last year and am going to try it with baseball this year, although I’m thinking it’s going to be a tougher with far more positions to fill. Anyways, I personally found it helpful for judging positional depth quickly when you are doing a traditional snake draft with only 90 seconds to pick.
I’m a big tier guy. It really helps me shape my auction or draft on the fly. There’s tons written on tiers, so I’m not going to get into it, but absolutely make tiers. It prevents getting too attached to specific guys, which isn’t a big deal in a draft, but in an auction can really allow you to conserve a few dollars here and there.
If you calculate values, there’s no need to guess at tiers. You’ll see the drop off at a position when sorting everyone by dollar value. Tiering is also doing half the work.
No tiers, and only using reasonable positional adjustments which are mathematically derived and accounted for in the dollar values… this is a breath of fresh air. Any chance you’ll be taking a few turns on the Rotographs podcast this season??
I actually thought about talking to the guys about being on every so often. I’ll go do that!
How do you factor in the value of flexibility offered by players who are eligible at multiple positions?
I value the player at the position he is worth the most, and would use the multiple eligibility as a tie-breaker when deciding between 2 equally valued players.
“Rather than comparing home run totals to every player, one must compare those totals to players at the position.”
Wrong. You need to look at the share of home runs the player generates compared to the total home runs that are expected to be generated across all players at all positions for the season. Only then can you adequately control for the positional scarcity that is engendered through MI/CI/Util positions.
Well yes, you’re correct. But in my example, it was just that position in a 2 position league. Technically you need to compare to all positions, but after adjusting the stat line to account for replacement at that position.
You are, then, basically taking the PS point, but using a strawman PS to insist that you aren’t. PS drafters aren’t, as you’re suggesting, simply throwing values out the window and insisting on drafting a 2B in round two, regardless of best-ness. Rebutting the PS position by saying that PS should be incorporated into value isn’t really rebutting the PS position. Rather, the “arbitrary” bump that you’re claiming PS’ers are giving scarce-position-player X, you are also giving to that player, but you’re calling it “incorporating into value”. PS’ers are also starting with value, and then adjusting based on position. The only drafter your argument works against is someone who insists on drafting a 2B in round 1, a SS in round 2 and a C in round three, regardless of the in-draft situation. While it’s possible that this person exists, she’s certainly not a member of a whole cadre of drafters who go into a draft thinking nothing value (in absolute or relative terms) who need an article telling them that’s stupid.
This is why the argument is, ultimately, a nullity. BPA’s are probably more likely to be the ones not incorporating position into their assessments, as both camps are likely starting with some ranking/dollar value. But saying that PS’ers should be incorporating position into their value calculations is redundant. That’s exactly what they’re doing.
It’s a response to those choosing either or. Saying they “go position scarcity”. One shouldn’t have to make this decision because position adjustments should already be reflected in one’s dollar value projections. If this is already done as you suggest, then PSers aren’t really PSers, they are doing it right all along, drafting the best player available, which by my definition equates to theist valuable player on the board.
I think the primary problem with BPA vs PS is that it’s overly reductive. There are many more inputs that inform a good selection at any point in the draft including: projection, remaining positional player pool, current roster construction at that point, current available statistical pool per your projections, current statistical analysis of your own roster (steals heavy, power light, etc.), competitor roster construction, game theory and analysis of competitor strategy, perceived risk portfolio of current construction, risk tolerance and so on.
One thing I like to point out is that the second any player at any position is selected in a draft, any positional adjustments you made in considering draft rankings become mathematically out-of-date. A draft is a tremendously complex activity. In a 15 team NFBC format, you’re looking at 450 players being distributed. The potential number of team permutations is vast.
I love to make the analogy of IBM Deep Blue vs Gary Kasparov. The computer used a lot of what they call brute force analysis based on immense computational power to evaluate millions of potential outcomes and then select from the best. There’s a zen saying, “in the beginner mind there are many paths, in the master’s mind only few.” The way Kasparov could compete with Deep Blue is that experience led the way to eliminate out of hand all but a few of the million permutations.
Applying this to a draft, that’s why mock drafting is actually not the masturbatory exercise it seems to be. It simply building repetitions so that our subconscious, where decisions are percolated, can do it’s thing. When you put prescriptions or rules of thumb like BPA or PS on a draft, the decisions will generally be weaker as they block out the mind’s inate ability to understand complexity and operate in the widest possible context.
The problem I have with the claim that a “correct” valuation system does the math for you, is that it assumes all other players are 1) acting rationally and 2) using the same valuation system as you are. What if I think I can draft $20 1B followed by $15 SS, instead of $17 SS followed by $16 1B, but there are more SS selected than expected before my next pick? Then I might now end up with $20 1B + $7 SS, whereas if the unexpected behavior was in reverse, I might get $17 SS + 13$ 1B.
To look at it from a different type of example, but the same basic idea, if I have a player valued at $20, but I have a good reason to believe nobody else has them valued at more than $10, it wouldn’t be optimal for me to select them as soon as they are the highest valued player available. I should select a player I value at $19, and wait a round or two to draft my $20 player.
This isn’t strictly an argument against the author’s POV, which I generally agree with, it’s just a way of saying that taking PS into account above-and-beyond the “correct” BPA is completely valid and necessary in order to maximize your total draft value.
I feel like the BPA vs PS thing is just a measure of how risk averse you are and not whether one is a better strategy. I feel like when you go for BPA you are taking on a greater amount of risk than those who draft for position. I know I feel much more settled whenever I cross 2B, SS, and C off the list because the risk of getting stuck with a dud at those positions is much higher the longer you wait. Of course, with BPA, you get a bigger reward if the player performs at a high level and, in essence, gives you 1st round value from a 2nd or 3rd round pick.
Agree completely, however, I’ll adjust a component which ends up differing from most of the projection systems, i.e., number of AB’s based on injury risk. (My “guesses” have been better after 20 years of playing FLB.) This alters the projections which are adjusted on my spreadsheet.
This usually results in passing on a first tier catcher because they either go for too much money in auction or the ADP is too high in relationship to other players available.
This seems to be the 20th topic on this post, but not one has provided a step by step processing to generating your own valuations. They always only include snippets or link to other articles which themselves don’t paint the full picture.
Can we please just get an article that systematically goes through how to generate league specific valuations?
sorry “post on this topic”
I don’t get into that because many smarter than me have already come up with reasonably good valuation systems.
Maybe I’m crazy, but I draft by sketching out the team I want. Compare that to what rankings are out there and ADP, then swap out players for what I should actually be able to get. I then get a rough idea of what my draft looks like round by round, then at positions that are high risk to not be available, have a fallback in later rounds.
I don’t do that exactly, but I do make a note of players I like at their ADP, as well as players I think are going way too early. I won’t plan my whole draft around specific players, but if I know that only 3-4 players at a position are likely to give me what I consider good value in the draft, I’ll try not to miss out on them.
I guess I’m not entirely clear on what you’re advocating for here. That said, I’m sort of leaning towards moderating my anti-positional adjustment view. It seems to me that the value of a positional adjustment is the value of avoiding garbage. So you look at catcher, and there are a couple of good players, a few more averageish players, and a whole lot of garbage. The positional adjustment approaches to dollar valuations for an auction draft that I’ve seen have added the difference between the positional average and the overall average on top of the valuation. So if Posey is 1.5 standard deviations above the average player, but catcher is 1 standard deviation below the overall player pool, we value Posey as a 2.5 standard deviations better player.
I see a huge problem with this, like a rising tide, it lifts all boats. So the crap at the bottom suddenly doesn’t look so crappy, while the top talent suddenly looks absolutely incredible. What I think is needed is a system that shows us precisely how horrible the bottom of the catcher pool is, while giving you a positional adjustment that tells us the value of avoiding the crap. I have no idea how you would do that. But your positional adjustment should not have the effect of making Brayan Pena look like a viable draft selection, Brayan Pena is what you’re trying to avoid. At the same time, should avoiding Brayan Pena really boost Posey’s value? There are lots of ways of avoiding Brayan Pena that don’t involve drafting Posey or Lucroy, so it seems to me that the upper-middle catchers should be valued more highly while the top catchers probably shouldn’t and the bottom catchers absolutely should not get any bump at all, we need to see in our spreadsheets exactly how bad Brayan Pena is, and positional adjustments should be based on the value of the fear that should put into the heart of my team. Posey is not better than McCutchen, we shouldn’t value him as better than McCutchen, but perhaps there is value in acquiring Yadier Molina or Willin Rosario in order to avoid Brayan Pena.
It’s not about avoiding the crappy catchers. It’s about the fact that in a standard 12 team league that starts 2 catchers, 24 catchers must be drafted, or bought at auction. Therefore, correct valuations would value 24 catchers at at least $1. Without a positional adjustment, you’re going to have 10 catchers (example) worth negative value, but you MUST spend at least $1 to purchase them. So to push them into positive territory, you boost their dollar value and all other catchers. Those extra dollars are then taken away from other positions.
I think we’re saying the same thing in radically different ways. As a practical matter, it doesn’t really matter if they’re valued at $1 or <$1 though, if you don't positionally adjust you'll have catchers for whom you know that $1 of value relative to the entire player pool is more than they'll contribute to your team. So you say "ok, I'm taking a loss in value on this crappy catcher because there's nobody decent left." Since there's not an alternative choice to inform your decision making, it doesn't really matter at that point. You have to take the best catcher you can come up with for $1.
It does matter though because you have to spend that $1, and that $1 is therefore not allocated to another player. So all those dollars need to come from somewhere and must be accounted for when valuing players.
So how does one figure out the math behind assigning players a dollar value?