Tiering up

A hot piece of advice these days is to focus on ranking and drafting players using a tiering system. The reasoning behind the system is that the best talent (in the tail of the distribution) is thinly distributed—the difference in expected value between the top two players is usually larger than the difference between the 99th and 100th players. Moreover, it is sometimes possible to see a noticeable step down in value between two consecutively ranked players; for example, there could be a larger than normal gap in value between the third and fourth ranked players at a position.

The most cited example of this is at shortstop, with Hanley Ramirez, Jose Reyes and Jimmy Rollins significantly better than Alexei Ramirez, Stephen Drew and Troy Tulowitzki.

A tiering system has two aspects: a ranking/forecasting part and a drafting strategy.

The ranking part of the tier system recommends paying more time and attention to slotting players into tiers, both overall and by position, rather than worrying too much about rankings within tiers (i.e. getting more exact dollar values or rankings). In general, I think this is useful. I would rather have system that put all of the players into their actual, true tiers but then got the ordering within the tiers somewhat wrong, than a system that got many of the exact rankings rights but messed up on some players’ tiering. In other words, many small mistakes are usually better than several big ones.

This comes with an important caveat though: the exact location of tiers gets harder to find the further down you go. The deeper the position, the easier it will be to find a third and perhaps fourth tiers (like at starting pitcher or outfielder). Still, meaningful differences between merely tiering and exact slotting get harder to notice as you go deeper into the ranks. It is pretty easy to get tied up in logical knots. I fooled around with tiering the outfielders and I quickly ended up with a quandary: I don’t think Nick Markakis is equal to Manny Ramirez, but it is equally hard to find a way to divide the outfielders ranked between them into two distinct tiers.

The drafting in tiers strategy basically boils down to the following type of advice: it is your turn in a draft and you’re trying to choose between Rollins and Chase Utley. Hanley and Reyes have already been drafted, so Rollins is the last of his tier left while Ian Kinsler, Dustin Pedroia and Brandon Phillips are all still on the board, but you have Utley ranked higher than Rollins. Whom should you draft?

The tiering strategy, if you’re a purist, says that you should draft from the bottom of the tiers whenever possible. In this case, that would be Rollins over Utley. Of course, it is rarely this easy.

A much better tiering strategy should also take other things into account. For instance, is someone from the alternate tier (like Phillips) likely going to be still available when you next draft? How much worse is the difference between Rollins and Alexei Ramirez than the difference between Utley and Phillips and between Utley and Brian Roberts (assuming he’s the best of the next tier)? How far is drop to the next respective tiers?

If you’re drafting at the tail ends (picks one and 12 in a 12-team league) the tiers have to have rather more players in them for tiering to be useful. If you’re not going to pick again for 22 picks, what’s the likelihood that any players in a three-player tier are left? And if you have to start worrying about the relative differences between players, then you’re going to need a ranking system that makes finer distinctions between players than mere tiering would do, which kind of obviates the whole point of tiering in the first place.

Print Friendly
 Share on Facebook0Tweet about this on Twitter0Share on Google+0Share on Reddit0Email this to someone
« Previous: Using Gameday to build a fielding metric (Part 2)
Next: Today at THT »

Comments

  1. GBS said...

    I’m certainly not advocating tiering in place of dollar values, but as a supplemental visual cue for when things get a bit hectic.

  2. bpasinko said...

    This isn’t basic economics or common sense for a beginner fantasy player.  If it’s my first draft, using a tiering strategy will help me and it’s not difficult to incorporate.

    You don’t need to make tiers to draft Rollins over Utley, but it certainly helps right?

    I’m a fan of tiers.  If you do dollar values to every person than yea you’ve done more work, but if you don’t want to do that tiering can achieve a similar goal.

  3. Mike Podhorzer said...

    GBS- Color coding each position when sorting your dollar values in descending order is a good way to visualize the drop off.

    bpasinko- I still don’t get how you could value a stat line though without calculating a dollar value. Don’t you still need to calculate values if you want to compare two players at the same position? I’ve been calculating values for years, yet I still can’t always look at a stat line and know which one is more valuable. So if you’re creating tiers without dollar values, it’s sometimes complete guesswork as to which stat line falls into which tier.

  4. Jonathan said...

    I think everyone is more or less on target here.  Tiering is certainly a sub-optimal forecasting and drafting strategy in a perfect world.  If you have lots of time both before and during a draft than you might be “leaving money on the table” by strictly tiering.  If you’re going quicker and dirtier, then tiering is a (but not the only) way to go.
    And I definitely agree that tiering is a rule of thumb for dynamically observing supply and demand.

  5. Millsy said...

    “GBS- Color coding each position when sorting your dollar values in descending order is a good way to visualize the drop off.”

    Mike,
    Aren’t you just creating tiers then?  It seems like the same concept to me.  While there aren’t groups, you’re essentially doing the same thing.  Why so vehement against this idea?  Assuming there is absolutely no subjectivity in the draft is a mistake, in my opinion.  Even when looking at your ‘dollar values’.  While you’re not consciously saying ‘Tier 1’, ‘Tier 2’, etc., you’re obviously taking into account the same type of dropoffs as this tiering system does.

  6. Mike Podhorzer said...

    I’m not actually creating tiers with the color coding, but I guess it does serve a similar purpose. However, I’m actually assigning dollar values to everyone so I know exactly what every projected stat line is worth, as opposed to a strict tiering system that just guesses what players are worth.

    Tiers just aren’t exact enough for me and also makes it impossible to compare one position to another. It assumes all Tier 1 players, Tier 2 players, etc across positions are worth the same, and that isn’t necessarily the case.

    Then how do you decide when a pitcher is now the most valuable player on the board? You would have no idea with just tiers and no overall list of dollar values.

  7. dan said...

    I’ve been using the colored valuation method for a few years now, simply because as someone mentioned earlier, in online drafts, there is some value to having a quick visual guide.  But also as has been said, it is simply that: a visual aide.  As you can see from an example here, the real key are the valuations (since this wasn’t an auction league, I didn’t bother converting to dollars).  It takes maybe five minutes to do and is nothing but helpful.

    http://homepage.mac.com/dls308/Sites/tiers.jpg

  8. KBO said...

    Mike,
      I’m interested in the “dollar amount” idea. I’m the type of person who would enjoy doing that kind of prep wor. How do you go about determining dollar amounts, as you said, it can be difficult to look at a stat line and instantly determine which is more valuable.

  9. Chad Burke said...

    I think Mike makes a good point in that tier 2 at two different positions are not necessarily equal.  One definitely has to look at which position has better value in tier 2.  The second tier of Roberts and Phillips might look better to you than the second tier of Drew, Tulo, and maybe Furcal but you also have to consider that the listed shortstops will be around a lot later in the draft.  Taking Utley allows you to stock up at other positions for the next several rounds and nab a second tier SS later than you’d have to spend to get a second tier 2B.  It might be that you can gain value at other positions in the meantime.

  10. chattanooga said...

    Tiering is a solid strategy as long as you follow it through to its completion—you have to “tier” your “tiers”.  When putting the list together, you have to establish a method for decisions made between positions (Everyone is dancing around this point, but no one’s articulated it).  for example, let’s say my master tier looks like this:
    3b-a
    1b-a
    ss-a
    2b-a
    of-a
    2b-b
    1b-b
    ss-b
    of-c

    to continue with the rollins/utley example, the decision becomes clear.  Because the margin between ss-a and ss-b is wider than 2b-a and 2b-b, I would know that Rollins is (according to this list) the correct choice.  The more effectively you have ordered your “master tier”, the more you will be able to take capture the comparative advantage across positions.

  11. KDaddy said...

    I’m not a fan of tiering, since I think it puts an artificial value bounds on players that exist more in a continuum.

    In your example, you take Rollins over Utley because Rollins is the “last of the first tier” while there are many others at Utley’s “level” available.

    I think that’s the right decision, but I think this is a false semantic distinction and you don’t need to invent “tiers” to make it correctly.

    Rollins is the right choice because at the moment of your pick, there is a scarcity of high value at shortstop, while there is a wealth at second base. You don’t need tiering to make this decision, just basic economics.

    Think of it as position scarcity (which most understand and use to some degree) that changes on a pick per pick basis. A draft is a rapidly evolving market, in which the supply-side of various items (shortstops worth having, second basemen worth having) goes up and down from turn to turn.

    Rollins might be worth more than Utley in some drafts, at some turns… but in other drafts that develop differently, Utley would be the right choice. This has much more to do with the dynamic supply than any tiering abstraction.

  12. GBS said...

    While not the only thing to focus on, I think tiering/position scarcity can be helpful, especially if you’re drafting online with a time limit.  It can be difficult to keep all your data straight sometimes, so having clean lines (literal or figurative) distinguishing castes of players at each position can be a nice visual clue if a position is starting to get thin.

  13. Mike Podhorzer said...

    I hate tiers. If you just calculate dollar values then you’ll see the drop off at every position more accurately. I feel like doing tiers is only half the work and you’re still making subjective cutoffs. Why only be 50% as accurate (using tiering) as you could be when you could be 100% as accurate (calculating dollar values) as you could be instead?

    Of course, we could only end up being about 70% accurate as that is the ceiling for projective accuracy at this point, but that doesn’t mean it’s not worth calculating values to be as accurate as possible despite the limitations of projections.

  14. Paul Sporer said...

    Mike,

    I think you’re viewing dollar valuations as inherently better than tiering because it’s a more laborious process that is usually done by only the hardest core of fantasy baseballers.  Though just because it’s more difficult doesn’t make it better.  You yourself acknowledge 70% accuracy as the *peak* for dollar value projections.  I’d trade a percentage of accuracy from tiering for the time needed to create an entire set of worthwhile projections.

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>