Replacement Level Depth Revisited
Last week, I tried to puzzle through something that has been bothering me – how depth affects replacement level. Consider outfield and shortstop. Last season, I might be looking at Josh Reddick and Adeiny Hechavarria as my first picks off the wire. I can do something with Reddick. In fact, I actually want to own Reddick. I like his offensive capabilities, even if he’s underwhelmed in recent performances. With Hechavarria, I would desperately seek a replacement.
The post wasn’t my best – I failed to communicate my message. My first gaffe was to create a pre-draft scenario. I probably should have gone with “in a vacuum” since my goal was to explore the possibility of trading $20 for $17. Despite my lack of clarity, the post did generate a lot of useful comments. Today, I’ll try to summarize and incorporate the dialogue.
Let’s start today’s effort by reframing the scenario. It’s post-draft, but the season has yet to begin. Your team has the correct number of outfielders but no shortstop. You can trade a $20 outfielder for a $17 shortstop. In that event, you’ll grab an outfielder off the waiver wire. Alternatively, you can select a shortstop off the wire. Your offensive stats are balanced such that you’re indifferent as to which categories improve/decline (just go with it). Patrick adequately described the pros and cons of either choice.
Even though you’re indifferent to which categories improve/decline, you’re still trying to maximize your overall performance. You want to take the option that delivers the best expected value subject to variance and your risk aversion.
The commenters to last week’s article frequently cited replacement level as a reason to avoid spending on scarce positions. If Troy Tulowitzki comes with a 40 percent chance to land on the disabled list, it could be better to invest in a similarly priced outfielder. Maybe Carlos Gomez? If Gomez lands on the disabled list, I can go plug in Reddick. If Tulo hits the skids, I have to reach for Hechavarria. My actual preference will depend on my risk tolerance since Tulo is a higher variance play with more upside (even if that’s debatable, let’s just say it’s true for the sake of the example).
Alternatively, you can swap Ian Desmond into the equation if you want a less injury prone shortstop. Based on early ADP data, Desmond (31st pick) is comparably priced to Tulo (27th). Unfortunately, you could be trading a lot of production for his track record of health. Michael Brantley (26th), Hanley Ramirez (28th), Josh Donaldson (32nd), Ryan Braun (33rd), Victor Martinez (34th), and Buster Posey (35th) are among the players in the same cohort. All of them project to outperform Desmond’s statistics by a wide margin.
Does Desmond plus Reddick match Braun plus Hechavarria? That depends on your projections. Steamer says team Desmond produces a fantasy slash of 128/35/135/23/.253 compared to team Braun’s 126/28/128/22/.264. Not much difference between the two. Of course, you could also have different choices for a replacement. I picked Reddick and Hechavarria because they bounced on and off waivers in a 12-team, deep roster league last year. Maybe your options are actually Dayan Viciedo and Wilmer Flores. Or maybe you project a lot more out of Braun and think Reddick will continue to flail.
There’s no foolproof, “correct” answer here. Your job is to maximize your expected performance subject to your preferences and tolerances. There is certainly such a thing as being “more correct,” but the backbone of building a roster is to create a logical process and follow it.
As commenter DaBears pointed out (emphasis mine):
Replacement level valuation systems poorly account for distribution at any point in the value spectrum. Intuitively, it seems there must be a difference between a position where values (pre- or post-positional adjustments) are bell-curve shaped, top-heavy or bottom-heavy.
Distribution of talent is exactly what is bothering me. Shortstop drops off very quickly. Outfield more closely resembles a lightly sloped curve with a tail. There are plenty of top talents, more mid-level talents, and a long tail of useful parts until the eventual plunge where playing time evaporates.
My hope was to derive some logical means to incorporate this into dollars. As multiple commenters pointed out, dollars are an insufficient means of capturing a player’s total value. There are many reasons, but the most important is that value is relative. Just as Mike Trout is worth more money to the Angels than the Phillies, Billy Hamilton might be worth a lot more to me than you based on our rosters. For me, Hamilton might represent a nine point gain in steals (and some points lost elsewhere). If you own Gordon, Gomez, and Ben Revere, then you’re already likely to win steals without Hamilton.
It’s this relativity that defeats my attempt to account for the distribution of talent with a dollar value. Fortunately, it’s not too difficult to use strategy and tactics to adjust for changing dollar values. While there probably is a mathematical solution, it would change with each pick and every add/drop. For me, it’s not worth the effort.
We’re left with an unsatisfying answer to our original question. Would I trade a $20 outfielder for a $17 shortstop? It depends.
You can follow me on twitter @BaseballATeam
In your example of Team DEsmond versus Team Braun, there is one more consideration beyond noting that the performances are similar. Are we likelier to find an OF or SS on the waiver wire that can exceed your replacement level player (Reddick/Hechavarria)? It’s always easier for me to find an OF to exceed expectations, so I need Team Braun to outperform Team Desmond to justify a Braun purchase.
I define replacement value as 150% of the starting spots. So there should be 12 starting catchers in a 12 team league, so 6 valuable backups, meaning the 18th catcher is replacement value, which has a value of $0 auction dollars. 90th outfielder is $0. 18 SS’s and 18 2B’s, but then after using their individual weights, i then find the 54th best MI and use that number as a factor for all 2B’s and SS’s, in addition to their individual weights, the 54th MI should be $0. Could possible adjust that later but im happy with it for now. You would be amazed what kind of circular logic formulas in excel can solve regarding position eligibilty and replacement level.
Also, as thin as SS is, you only need 1 SS, and 1 MI. You need 5 OF’s, and a lot of depth. SS is still more scarce, but scarcity is closer than most think.
But that isn’t always true. In a 12 team league it might be the case that there are not, for example, 6 useful backup catchers. Maybe there are only 3, and by naming #18 your replacement level you are setting the floor too low. I don’t care how many players get drafted in my league, I care how many are WORTH drafting. If some fool wants to pick up a useless player just to say he’s got a backup at that position that should not affect my valuations.
This is a concern of mine, and something i have considered altering in my Spreadsheet. But the 17th best Catcher is Yasmani Grandal, and i think he has value, mine has him at $3. with about for guys less than $1 (My formula allotted a few more catchers some value as utililty depth. Even still, less than $1 is still unrosterable since the cheapest you can sign a player is $1.
Backup doesnt necessarily mean what you think. Its not an extra guy just so you dont have to bother with waivers. Some of these guys are youngons with breakout potential. I own a $4 Marcus Semien, he is a 3rd SS for me, but he is more than depth, he is a guy that could breakout and allow me to trade my starter.
There are any number of reasons to draft a sub-replacement level player. A prospect with great lottery ticket potential but whose playing time concerns drive down his projections is one such. Nonetheless, the basic framework of your values has to built on _something_ and arbitrary numerical cutoffs is likely to mess up your floor.
I agree they are arbitray cutoffs, and i try to stay away from arbitrary frameworks of any kind when approaching a project like this. Maybe i come up with a better approach, but accurate Dollar values are based on replacement value, which makes this article a great discussion. I am all ears to a better concept, but in the meantime, although arbitrary, it is reasonable.
Also, you mentioned prospects. I have a few theories on how to put a price tag on a player that will not play in 2015, using age, scouting grades, and projected arrival year. I also considered using the tool provided by Ottoneu which gives us the average actual salary, which should be pretty accurate in determining market value of prospects. But currently, i base my dollar values on around 36 roster spots and $400 per team used. So, if you want to roster prospects, you will do so with the 4 remaining roster spots and any dollars left over after the auction.
Your premise is still not quite in line with how player valuation is supposed to work. Values are values, and if a $17 player seems more valuable than a $20 player, it could be that your values need tweaking. One fix I make is to average the +/-2 players around the replacement level player to find the ‘true’ replacement level. This moves the replacement level further down for positions like catcher, where the 15th or 16th catchers might be so terrible (few PAs projected) that the value curve really dives off. Meanwhile OF is unaffected because the curve is basically linear and smooth, as you’ve pointed out.
I think the other thing you’re hitting on here somewhat abstractly is the concept of tiers. It’s always a good draft strategy to make sure you draft a player at a position before a big drop-off occurs. The opportunity cost of passing up on Tulo/Desmond/Hanley is a lot higher than passing up on a top OFer because you can generally take an appropriately valued OFer at any point in the draft without worrying too much about tiers.
Otherwise, I say trust the spreadsheet. If you’re going to disagree with the values you might as well go the Larry Schechter route and not bother making positional adjustments.
“It’s always a good draft strategy to make sure you draft a player at a position before a big drop-off occurs.”
Why? As long as you’re getting X amount of value of profit per pick, the ranking within the position doesn’t matter. If I “miss” the end of a tier, I just wait a bit until the next tier comes up. By definition, I am guaranteed to get a player for every position that is above replacement level.
Assuming the last guy in the tier is near the top of your list, it is a very good strategy to preserve value.
Again I’ll ask…. why?
In theory yes, but in practice the draft is a game of playing your valuations vs. your league-mates. Getting guys before tier drop-offs also increases the likelihood that you are ‘beating the market’.
“Getting guys before tier drop-offs also increases the likelihood that you are ‘beating the market’.”
I’m still not seeing any reasoning behind the logic.
I understand the last guy in a tier strategy seems logical, i also understand why it isnt logical.
http://i.imgur.com/AsvANp2.jpg Tiers, visualized. At the end of a tier, you’re most likely getting something you value above the market (assumes your values are accurate).
A very simple example assume a 3 player draft. (Salaries are used for value)
1st baseman OF
$40 tier 1 – Player A $40 – tier 1 Player D
$39 tier 1 – Player B $33 – tier 2 Player E
$38 tier 1 – Player C $33 – tier 2 Player F
If you draft 1st and select Player A you will end up with Player A ($40) + Player F ($33) for $73 of value.
If you draft 1st and select Player D you will end up with Player D ($40) + Player C ($38) for $78 of value.
Both that graph and the 6-team example use faulty logic.
The graph is a misrepresentation of the value of players picked in a draft. It is predicated on a) strict tiers with all players being identical in value and b) there only being one tier to choose from. In reality, there is a difference in value between players of the same position in the same tier, and there are likely 6-8 different tiers simultaneously being drafted from. When you mush all of that together you get a smooth curve. The stepped curve does not represent the value of players that are taken in a draft.
As for the example using 6 players for a 2 team draft, you are making an argument for replacement player theory, not for tiers. If you properly value the players then you have an $8 OF, $3 1B, $2 1B, two $1 OFs, and a $1 1B. Using replacement player theory, the obvious choice is the $8 OF. This has nothing to do with tiers.
Your unwillingness to accept simplified examples is making it impossible to explain this to you…
I’m more than willing to be proven wrong if tiers truly are useful. I’ve been proven wrong before, and I’ve changed my mind before. I’ve never seen a valid argument for tiers.
Like I said, there are 6 to 8 tiers being drafted from at any given time (C, MI, CI, OF, SP, RP, and maybe you want to break down MI and CI into their components) and there is indeed a range of values within each tier. The consequences of missing out on, say, a second tier 2B is that….. I get a third tier 2B. And so on. The goal should be to maximize value. Using tiers unnecessarily restricts my freedom to draft the most value.
When it gets near the end of the draft and I have only one OF spot left and one MI spot left, then yeah, it might make sense in that scenario to use tiers. If there is one OF left in the penultimate OF Tier, and three 2Bs left in the penultimate 2B Tier, (obviously I always have access to an OF or 2B from the lowest tier at each position), I would choose the OF and hope that one of the three 2Bs make it back to me. But this is a rare situation that occurs maybe once, if ever, throughout a draft, and requires no tier preparation because it can easily be done on the fly if the situation arises. I almost always have the option to simply draft a player from another position if there is nobody of value to be taken at SS, for example.
Tiers are one of the biggest fallacies in fantasy baseball, almost as bad as position scarcity. They sound nice on the surface, but they just don’t stand up to scrutiny.
I agree with Skin Blues. Tiers are good to squint at but the name of the game is getting the necessary stats. Projected dollar value is a good proxy for that, but you have to make sure you’re not loading up too much on one stat at the expense of another.
Once you realize you’re buying stats, not players, you’re good to go
Tiers are terrible. They drive you to bad decisions by aggregating unlike players together. They also encourage the mistaken belief that tiers of the same level have equivalent value to each other.
Hear, hear
Like most things, “it depends”. I typically chart my tiers on a ledger sheet so I’ll have all hitters on the same sheet. Then I might leave a gap so (for sake of argument), my “Tier 2” 2Bs are level with the “Tier 3” outfielders.
So my tiers might look like:
C 1B 2B SS OF
1 1 1 1 1
2 2 2 2
3 2 3 3
3 4 3 4
4 5 4 4 5
(my league is a bit different also in that position eligibility is determined on a week-to-week basis: A position player needs to play 2 games at a position the previous week to be eligible there. So, for example, Miguel Cabrera was never eligible at 3B in our league last year except for Week 1)
OK, that didn’t line up properly, so it probably makes no sense now.
I have no idea about the dynamics of a league like that with the position eligibility. Is it a huge headache keeping track of it all? I guess most of the time it’s straightforward, but if guys make a position change throughout the year it seems like it’d screw up your plans, or if your 1B is filling in for an injured OF for a week or two. Ben Zobrist must be a nightmare in that league! I’m actually in favour of just giving each player a single position to use for the entire year and not giving out any additional positions regardless of where they play, so sort of the other extreme I guess.
For a normal roto at least, lining the tiers up any which way still doesn’t make them useful. Any drafting method can work due to the variance of outcomes, and it’s not like using tiers will destroy your chances of winning. But it’s suboptimal because all it does is restrict your ability to build the best team possible. There’s no advantage gained by it, but there very well could be a few times that it’s a hindrance if you take a player too early simply because there was on a run in that particular tier.
ie: All the good SS might be getting gobbled up and there’s a huge dropoff after the one guy left in this tier, so you draft a $12 Jimmy Rollins and pass up on a $16 Adam Laroche at 1B since there are a handful of $15+ 1Bs still on the board. So you still get one of those $15+ 1Bs, but you could have had two of them and put one at UT or CI, or you could have had Laroche (a $16 1B) and a $15 OF. Instead, you have a $12 SS and a $15 1B. $27 of value instead of $31. Then you get a $1 1B in the final round rather than a $1 SS, between which there is very little difference in value. When/how is that $4 of lost value recouped?
I was one of the original commenters who said you were incorrectly defining replacement level by wanting to trade $20 for $17. I think that you have taken a decent approach here at answering for reasons why you might do that but I think this really only applies to a category league (which most people play) and not a point league (which fewer people play). $ values are harder in category leagues because you need find a balance between your various categories which leads to different prices for the same players. Billy Hamilton is simply worth more to the team currently in last place in steals than he does to the team in second place in the category. In this case you are convoluting category needs and position needs. Yeah it might make sense to trade a $20 OF for a $17 SS but it might also make sense to trade a $20 OF for a $17 OF depending on your needs.
Points leagues take away this somewhat artificial construct and allow you to focus on the simple goal of maximizing your points by staying within your budget.
“In this case you are convoluting category needs and position needs”
perhaps scarcity should be weighted more in terms of category needs instead of positional needs. then apply what you need/what’s scare to players who present them as good values.
That should be step one in creating Dollar Values anyways. But just because a Stolen Base is scarce (debatable), it doesnt matter if you are in 1st place by a wide margin. At some point, current standings should affect your personal dollar values. Although i would argue it doesnt alter his trade value, only his value relative to your own team needs.
Again, you need to recalculate your values based on what is actually available on the WW. So if you can get a $5 OF on the WW, then you need to subtract 5$ from the value of your OFs. If the WW SS is -11$ then you need to add $1 to your SS values.
I agree. Dollar value spreadsheet is a work in progress… but i love what i have so far. I might alter it to account for the realtime waiver wire… but that’s a project for April.
I’ll concisely re-state a few points I made on the last article:
– The demand for replacement OF’s is much higher than for replacement SS’s (5x more dedicated positions = 5x more likely that an injury will occur, or a demotion will happen, etc)
– Dual-eligible players that qualify in OF are routinely used at all five infield positions rather than OF, ie: Gattis at C, Zobrist at SS, Trumbo at 1B, etc. This removes even more of the OF depth.
– Flex positions (CI/MI/UT) are a buffer against position scarcity since they are very disproportionately filled with players that qualify in the OF and 1B
– The value that you can expect to extract from any replacement player is minuscule given a) the low chance of occurrence, b) length of time a replacement is typically needed, and c) the difference between value of replacement players.
If a full year of “Replacement SS A” is worth -$4 and a full year of “Replacement OF B” is worth -$1, then the difference between those two players during the few weeks they might possibly be needed on the active roster is a very small fraction of $1. It does not even come close to bridging the gap between a $20 player and a $17 player.
And as I said in the other post, in a realistic league setting the replacement players for MI and OF will be comparable in both depth and in quality. The hypothetical situation of 24 SS’s being drafted compared to 75 OF’s does not come anywhere close to the reality of a mixed league.
“And as I said in the other post, in a realistic league setting the replacement players for MI and OF will be comparable in both depth and in quality.”
I haven’t found that to be the case, mainly whatever flaws exist in these types of players, it’s generally easier to extract positive value from the OF than the SS at these levels. Usually it’s some extreme (L/R H/A) split in the OF that the SS doesn’t have, and more people to platoon the OF with. So maybe if you’re in a position to give every PA to these players it’s fair, but not if you’re selective. Perhaps the number of bench slots in the league are what’s causing the discrepancy?
It’s definitely different depending on league type. In weekly leagues the platoon thing really doesn’t matter. Whereas it may be slightly beneficial in daily leagues, but even still, roster spots have a ton of value in daily leagues. If you’re using a spot on a player that you only use half the time then you’re foregoing the value of using relief pitcher. In leagues that have an innings cap, relief pitchers are extremely valuable if you have room on the roster for them. And when it comes to waiver wire players, their purpose is generally to be an injury replacement, anyway.
And like I said, the demand for replacement OF’s is a lot higher than the demand for replacement SS’s. With 5 OF slots and only one SS slot, that’s on average 5 times more OF’s that will be injured and need to be replaced. So even if there is slightly more supply of replacement OFs, there is a lot more demand. And the value provided by these guys is very minimal.
You hit the nail on the head because it is all about maximizing performance and value is relative to each situation/owner. A mathematical formula will likely to a good result but it would need to take into account hundreds of factors before it would be useful to the above average player.
For example, lets say I was playing chess against someone and I was able to use a computer for one move. The computer suggested Move X and let’s assume it is the best move. Is Move X best for my chess game? It depends on my opponent and how well I can follow up after I play Move X. So the best move in a vacuum (Move X) may not be the best move for my situation.
Couldn’t you just project your total team roster scoring/category standing with the trade + specific OF replacement player and compare it to without the trade + specific replacement SS? This is what I consider when making trades post-draft. As you say, accounting for this pre-draft (without a full roster to project) is not possible. After the draft these relative valuations become easier.
SS…..tulow, reyes, desmond and hram……only a few people will be getting better SS than the rest. unless you can get desmond dont bother.
if you play in a league with just a SS and now MI dont bother getting a SS early. the better SS stats can be had with a later OF.
My situation is a little unique because its a shallow league, 10 team h2h with no MI/CI, 3 OF and 1 UTL. Having looked at the depth of all of the positions, it certainly looks like SS and 2B is extremely top heavy whereas the OF is very deep. In a league where we only carry 3 OF, would it be too crazy to take my SS/2B very early and filled out my OF afterwards?
I’m thinking of targeting guys like Desmond and Rendon in the first few rounds and then grabbing guys like Marte/Springer/Dickerson afterwards. Hypothetically speaking it could look something like Goldschmidt/Rendon/Donaldson/Desmond/Springer/Dickerson/Marte. So focusing on 1B/2B/SS/3B very early and then filling out my outfield with some solid mid level guys. Plus with only 3 OF, there should be a plethora of depth on the waiver readily available. Thoughts?
10 teams with no MI/CI changes things a lot. Shortstop needs a decent sized boost, and catcher needs a huge boost. Second base doesn’t have any scarcity. You can wait until the end and take a guy like Kendrick or Wong and you’d be in fine shape there since they’re comparable in value to the OFs that will be taken with the final picks. If you have a pick near the end of the first round I would really try to grab Posey. If you miss out on him, try to get Lucroy or Gattis in round 3.
My problem with using “player production minus replacement level production” as a means of comparing players between positions is that it fails to account for how talent is distributed across the full range of players at each position. Here’s a hypo:
You’re contemplating drafting one of two players:
Joe, a 1B whose raw stats are worth 20 “units” of production.
Bob, a SS whose raw stats are worth 18 “units” of production.
Your analysis indicates that a replacement level 1B is worth 10 “units” and a replacement level SS is worth 4 “units”. So Joe’s delta over replacement level is 10 “units” vs. 14 “units” for Bob. Clearly Bob is the better choice, right?
Probably. But suppose the pools of remaining 1B and SS are extremely pathological. The 1B pool consists of Joe and then a bunch of guys projected to produce at a level of 11 units. That is to say, barely above replacement level. The SS pool consists of Bob, plus another guy projected to produce at a 14 unit level, plus a guy projected to produce at a 12 unit level, plus a guy projected to produce at an 8 unit level.
Now Joe looks like the better choice, since not drafting him ensures a production dropoff of at least 9 units, whereas if you opt not to draft Bob you can pick up a lesser SS later in the draft and still get a guy projected to produce significantly above replacement level.
Do valuation systems take this into account, or do they calculate “value” using a simple “projected production – replacement level production at position” formula?
In this example, you omit the fact that I can simply draft other positions for a while until the end of the draft at which point I take my $1 1B. In the meantime I can draft all the other positions. If I want that SS who is projected to produce at a 14 unit level I would need to draft him much earlier than I would need to draft any of those remaining 1Bs who all produce at an 11 unit level and are thus last round picks.
I was coming at it from a snake draft perspective where, with my pick, Joe and Bob are the two most appealing players and I’m trying to choose between the two. The situation I describe is one in which using the metric “delta between a player’s projected production and the replacement level production at his position” would lead to the wrong decision.
That metric would suggest picking Bob, but the different compositions of the player pools for their two positions strongly suggests Joe might be the better choice. Take him and I might wind up with 20 + 14 production vs. 18 + 11.
@Buddy – This was the issue I noted in Brad’s prior post. I’d be happy to get feedback on this, but a solution I’ve come up with is to incorporate a position-specific Z-score into a player’s valuation. Doing so accounts for differences in distribution within a position.
After making the usual calculations and adjustments so that I have a $0 player at each position, I add to each player’s value a positional Z-score, calculated as:
([Player’s unadjusted z-score sum]-[Average unadjusted z-score sum among players at that position who project to be drafted])/[Standard deviation of unadjusted z-score sums among players at that position who project to be drafted]
The result is that players who are below average at their position (among rostered players) are penalized and above-average players receive a bonus. By magnitude, players who are far above average for their position receive a larger bump than players who are closer to average, even if they have the same value otherwise. Put differently, these adjusted values now favor a player who provides a stronger advantage at his position (relative to both replacement level and average level) rather than relying on replacement level to sort things out.
Would welcome any thoughts…
That’s true if the team only has two positions, 1B and SS. But since there are many other positions it’s not a binary choice. I can take the 18 SS and then take a 16 3B and then a 15 Pitcher and then the whole rest of my team, and then with my last pick I’m guaranteed the 11 1B. Whereas if I took the 20 1B first, and then the 14 SS, I would be missing out on the 16 3B. Replacement level theory holds up extremely well.
And really, it shouldn’t be replacement player theory so much as the last player picked at each position.
All this makes my head spin but, it strikes me that one useful thought experiment might be to consider the endgame of a deep mixed league season.
Even if it’s some sort of keeper/dynasty league, as October looms and the risk of being stuck *the whole year* without Tulo or whomever your injury-prone SS1 starts diminishing, at some point one feels free to drop the lame-but-hey-maybe-a-breakout-is-coming backup one has rostered in order to, say, stream another starter or grab a hot prospect …
I think it would help solve the problem here to determine when that “go ahead and drop lame backups” moment is, and to what extent quasi-prospect status (as exemplified by a commenter above in Semien) factors into these decisions
The question of positional depth at draft time, and more broadly the distribution of talent (skewness), is interesting and one I have pondered as well. However, it is probably not all that relevant.
To answer the question of whether to draft the high priced OF or SS, the answer probably comes from the question: 1) From which position can you likely extract excess $ value from your lower round/priced picks? It is then the other position(s) that you should draft (pay) for early (more). Or, 2) From which position are free agent gems most likely to emerge and be available to me (supply & league-wide demand)?
Basically, are you likely to be able to get a value premium from other positions?
For the first question, a personal example is that last year Corey Dickerson is a guy I tagged as someone getting no ADP love – often going undrafted. I knew I would value him much more (though, still not as highly as he performed) than other drafters in my league, given how I had him projected. While memory escapes me beyond this name, I believe I had a few other good OF value picks I felt comfortable with as well. This allowed me to focus my early-mid round attention on other positions, as I had a good chance of achieving excess value in later rounds from OF.
I tend to start with the 1st question, and try to see if there are positions where I feel likely to be able to draft/buy a good late round/low price bargain. If the 1st one does not seem to net me much, then I consider the 2nd question. I have not done any rigorous post-analysis of undrafted free agent impact by position – if someone has then this would be worthwhile to the discussion.
My 2 cents.
It gets riskier as your leaguemates get more clever. If you’re waiting til the 21st rd to take Dickerson, or Rendon, etc., and someone takes him in the 20th, the bottom can drop out of your strategy unless there is a similarly valuable player still available.
you “reach” for the guys you really want past the 10th-15th round. if you’re really depending on a guy like that, you make sure you end up with him
or you join an auction league
Totally agree wiht you Sir Larry. What I usually do is use a good set of recent ADPs (ie: the ADPs from March NFBC drafts), assign dollar values to every pick, and then find the difference between my value and the drafted value of each player. Not only does this allow me to find out what positions are being undervalued (typically SS is overpriced), but it allows me to see at what points in the auction certain positions go for the best bargains. A typical finding is that 1B/OF provide the most value at the beginning of the auction, and SS/2B provide the most value near the end. The SS’s taken at the beginning have a huge premium put on them to the point it’s not worth it. I’d rather grab a scrub SS late in the game for a buck. Whereas premium 1B’s typically go cheaper than they should because they’re perceived to be plentiful. I’ll fill my 1B/CI/UT slots with premium players.
Like Ryan says, it’s risky if you’re playing against smart people, but those pockets of value can usually be found if only because of variations in projections, even if valuation methods are the same.
So I have a novel view on this whole conundrum — you are on the right track when you talk about Replacement Level player relative to positional scarcity, but the piece you are missing is relative value for that particular position.
I play in roto and points leagues, but I feel like the score in my points league provides the best picture of value, but you can you use overall offensive rank for this exercise too. So using SS vs OF, here’s a comparison at #1 overall, last draftable (~150% of roster capacity) and middle of that range:
#1 Overall
SS – Jose Reyes 448 pts
OF — Mike Trout 595 pts
Last draftable:
#18 SS — Jean Segura 287pts
#50 OF — Jay Bruce 313pts
Middle range:
#9 SS — Asdrubal Cabrera 369pts
#25 OF — Nick Markakis 409 pts.
So what that tells me is that there is a 282pts spread between top and bottom OFs and a 161 pts spread between top and bottom SS, but 186 pts of the OF spread comes in the top half of OFs, where the SS spread seems pretty distributed.
So accordingly, a guy in the top half of OFs is more valuable than a guy in the top half of SS. Given that and assuming both the $17 SS and $20 OF are in the top half of their position, I wouldn’t trade the OF for the SS. No obviously, when I do the math, I break it down in more 2 tiers, but the logic still applies the same, marginal relative value is the only thing that matters.
If there are no SS worth drafting in the middle range, then don’t draft a SS in the middle range. Draft other positions. The fallacy people fall for here is that they assume the players are drafted in the same frequency for every round throughout the draft. If there are a couple great SS and then nothing but crap after that, but the OFs have a smooth dropoff, the draft might go as follows. For simplicity’s sake this is a 20 team league with a 5 round draft.
Round 1: 5 OF, 2 SS, 13 other positions
Round 2: 4 OF, 0 SS, 16 other positions
Round 3: 4 OF, 5 SS, 11 other positions
Round 4: 4 OF, 5 SS, 11 other positions
Round 5: 3 OF, 8 SS, 9 other positions
So there you go. If SS have a steep dropoff after the first two superstars, then they simply won’t be drafted til closer to the end. No need to artificially pump up the value of shortstops to push them into the middle rounds.
How is the number 50 outfielder the last draftable? You used 150% on SS (12 + 6 backups) but not with OF. In ottoneu we use 5 OF’s * 12 teams = 60. Plus 30 backups = 90 outfielders draftable.
Nevermind, you based it on 3 OF’s. I get it.
Blake Murphy tried to get at the question of distribution last year in FG+. I think you can see this chart, which graphs values by position
http://cdn.fangraphs.com/plus/wp-content/uploads/2014/02/auction-values-all1.jpg
unfortunately for (our?) intuition, it looks like shortstop has a ton of talent right around replacement level. Of course, the way the values are derived affects this graph as well.
Interesting, I like his pivot table using N>0 and avg>0 for each position. I might do some more research on how I approach position scarcity after reading that.
Are those positionally adjusted or non-positionally adjusted values? Makes a big difference in how you read the graph, I’m assuming they’re non-adjusted, is that right?