Log in Register

Login to your account

Username *
Password *
Remember Me

Create an account

Fields marked with an asterisk (*) are required.
Name *
Username *
Password *
Verify password *
Email *
Verify email *

fb mb tw mb

Tuesday 17th Oct 2017

Today we’re going to continue looking at the repercussions of APE, the ADP Principles of Equivalence. If you are new to the site and need to catch up, I introduced the concept HERE and talked about it in terms of position scarcity HERE. The next focus will be on ADP or average draft position.

The Internet is now overflowing with essays hammering the concept of ADP. I’ve got a few of them out there myself, which is ironic since I was producing ADP’s from National Fantasy Baseball Championship satellite leagues for our Platinum customers before they were even produced by the NFBC. At the time, they were helpful. As with any strategy or tool to gain an edge, it is most effective when no one else is doing the same thing or using the information. ADP’s are everywhere so they are no longer the drafting tool they were five years ago.

A few of the conventional reasons cited when making the case why ADP’s are no longer the cat’s meow are they often are a mishmash of different league formats and rules. Some are biased by the default ranking of the site administering the mock draft or even the inclusion of computer-drafted teams. Some include mocks previous to decision altering information being made public such as the recent PED scandal, the pause around Felix Hernandez signing his extension and even Michael Bourn signing with Cleveland. In short, the credibility of the ADP is questionable.

All of this is well and good, but my main beef with ADP’s is the perception of what they are and what they should be used for. Let’s say we generated an ADP using the same league format, within a short time frame so the information was all the same. The result is not a guide to help you rank players like many appear to believe, but merely a quantification of how the market ranks the players. I realize this may seem to be one and the same, but here’s the difference. If you want an opinion on something, who do you ask? I hope it is someone well edified on the subject, so their answer is credible. Now think about an ADP. With due respect to those generating it, and I don’t care if this is an NFBC ADP, do you really trust each and every person’s opinion that goes into producing the ranking? Think about it – only one person wins a league. There are 14 losers. To be blunt, an ADP aggregates the opinions of more losers than winners. If you use an ADP to help rank a player’s potential production, you are misusing it. You should be coming up with your expected production independent of the ADP, then perhaps gauging market value using an ADP that best resembles the league of interest. To base your picks on the ADP is a big mistake.

I and others have been preaching this and similar arguments for a couple of years but I still hear things like ”that was a great pick, you got him at pick 90 and his ADP is 72” or “you took him way too early with the 52nd pick, his ADP is 67.” In fact, a feature of the NFBC draft room is to grade your draft relative to the current NFBC ADP. That drives me nuts.

Much like the scarcity argument posed last week, it’s one thing to continually make a point anecdotally to a populace that likes numbers while it’s another to actually use numbers – so let’s use numbers to help make the point sink in.

Reviewing the concept of APE, you determine the dollar value of each player like you would an auction and then line them up highest to lowest. Given that this is not the same as a properly constructed draft list, if you assign a round number to each value corresponding to where they fall (in a 15 team draft, the 15 highest values would be round one, the next 15 round two, etc) and multiply that by three, the resulting number represents the amount of players above and below that pick player are fundamentally the same, or at least they have the same value. Same value is defined as anything within two dollars. In other words, a $16 player and an $18 player are basically the same guy. Expected performance, thus the corresponding value is best thought of as a range of outcomes. Outcomes with $2 worth of value are the same. Going back to the rankings, a sixth round player would have eighteen players above and below that can be considered to be equal in value.

Let’s use an example. In a 15-team league, pick 80 is the fifth pick of the sixth round, so the players ranked between 62 and 98 are all the same as player 80, assuming you agree a player +/- $2 is the same player. What would happen if at that pick, you chose a guy with an ADP of 95? You’d be chided for taking him too early, but did you? According to APE, you took a guy within the two dollar range, so no, it was not too early. Now think if this were the tenth round, there would be thirty players above and below. If you took a player with an ADP 27 picks later, hysterical laughter would ensue and insults would fly – all unwarranted.

Going back to pick 80, what if player 65 was still on the board and you drafted him, what would the reaction be? Everyone would be lauding what a great value pick you just made – or was it? He’s also within that two dollar boundary, so was it really that great a pick? Not according to APE it wasn’t.

I realize there are fallacies in this argument. The ranking by dollar values does not necessarily mesh with the projected dollar values if you order them via ADP, but they’ll be pretty darned close. At minimum, they are close enough so the notion of APE can still be applied.

The bottom line is I no longer need to use straw man arguments why I feel ADP’s being misused, I can demonstrate it mathematically. Even if the ADP is a perfect reflection of the market and is generated by a group that really knows their stuff, the application of the ADP is faulty. It’s not a judge of the quality of the pick unless the choice was so egregious it falls out of the limits set by APE. And even then, the intrinsic value of the pick could make it viable.

This is a good place to stop as the next installment of this series is going to focus on intrinsic value and how the proper use of ADP helps you acquire the most intrinsic value. Please feel free to comment below and I’ll do my best to respond.

Comments   

0 #25 Joseph Thelen 2013-02-25 01:57
Todd,

ADP is immensely valuable in one way, and one way only. It represents the market for players at a point in time, that is, it alerts you to when a player will be taken in a snake draft.

If you are the type who targets players, and I am, it permits you to reach, with confidence, one or two rounds ahead to secure the player.

I don't look at the ADP as a source for accurate player valuation. Rather, it's a means to assess, from one's personal perspective, where the market is undervaluing a player. And a means to make relatively certain your ability to obtain that player.
Quote
0 #24 OaktownSteve 2013-02-14 04:29
Hey Todd. I don't want to mess up your day job!

There's still a lot of meat on the bone for this topic and I'd love to pick up the conversation when the dust settles. I'll check back...I'm enjoying the content on the site anyway. You can always hit me up with a message.
Quote
0 #23 Todd Zola 2013-02-14 02:50
One more comment before I go back to profiling...

An argument can be made the only pick/player with any value is B14. If I don't get the second pick, I don't want to play.

Unless of course the person with the first pick takes A9 -- then I want the third pick :P
Quote
0 #22 Brandon Gavett 2013-02-14 02:48
The easy answer to Steve's hypothetical is extremely simple. Just rank players and assign tied values to the lowest in the group. For example...

Position A would be ranked as
14 14 14 14 14 14 14 14 14 14 14 14 14 14 15

Position B would be ranked as
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

That's a very straightforward and intuitive approach to drafting. Player values can be any derivative of such a method (e.g., percentile ranks). The problem is that the numbers above are just projections, and each player will deviate from their projections by some unknown amount. The winner is the one who can get the guy - projected for 9 - who produces 10. Projections need to account for individual variability in players (see my blog linked by my name).
Quote
0 #21 Todd Zola 2013-02-14 02:36
Quoting OaktownSteve:
Let me give you a counter example:
15 drafters need to take one player each of player A and player B (call them catcher and first base)

The distribution of home runs for A and B is as follows

A: 9 9 9 9 9 9 9 9 9 9 9 9 9 9 1
B: 15 14 13 10 9 8 7 6 5 4 3 2 1

Obviously you want to pick from column B all the way until the last pick even thought the VAR is higher in column A on half the picks. It's a little less obvious if you use Value Over Average or a standard deviation calculation, but even still the draft formula doesn't help you understand who to draft because it doesn't tell you how the values are distributed within the ranges. And now you multiply this complexity by many orders of magnitude for 23 draft slots across multiple postitions and you see why making positional adjustments leads you down the wrong path. I might have time to say a bit more later to further clarify. Hope that made sense.


Just my luck -- I've been waiting 10 years for someone to come up with the counter and I finally get it during my busiest week of the year :sigh:

I see what you're saying, I understand what you're saying, my initial gut reaction is two fold.

1. at the end of the day, performance variance is going to trump any game theory - akin to what Bill James calls getting lost in the fog

2. Not quite ready to give in yet, I still think the VAR method will work in a practical sense since in most fantasy formats things like utility and multiple position eligibility will all but eliminate the instances of an example like presented.

All that said, it is making me rethink some of my argument against not raking a player early due to the impending drop off in talent at the position.
Quote
0 #20 OaktownSteve 2013-02-14 01:54
Let me give you a counter example:
15 drafters need to take one player each of player A and player B (call them catcher and first base)

The distribution of home runs for A and B is as follows

A: 9 9 9 9 9 9 9 9 9 9 9 9 9 9 1
B: 15 14 13 10 9 8 7 6 5 4 3 2 1

Obviously you want to pick from column B all the way until the last pick even thought the VAR is higher in column A on half the picks. It's a little less obvious if you use Value Over Average or a standard deviation calculation, but even still the draft formula doesn't help you understand who to draft because it doesn't tell you how the values are distributed within the ranges. And now you multiply this complexity by many orders of magnitude for 23 draft slots across multiple postitions and you see why making positional adjustments leads you down the wrong path. I might have time to say a bit more later to further clarify. Hope that made sense.
Quote
0 #19 Todd Zola 2013-02-13 22:34
Quoting OaktownSteve:
The reason it's steep is that there's a lot of standard deviation in catcher values and it skews the numbers. More importantly, you really don't need it because you are not trying win the catcher draft. You want to win the draft as a whole (you actually want to win your league, but that's a whole other argument). The value over replacement tells you nothing about the tipping point where a catcher has enough value over the next available on the draft board to justify taking a catcher with less unadjusted value over a player at another position with greater unadjusted value relative to the next player at that position. And that's just a simplified model where you are looking at jsut two positions. The math goes bonkers when you look at the draft as a whole. Game theory covers this but it's some PHD level stuff.


The implication is MORE steep. There is always an adjustment, just more this season for some reason that I'll figure out before the next update.

I disagree the bump is not needed, I think it is -- I only care about useful stats, those that everyone else does not have. I want my initial rank to be reflective of useful value. The incredible oversimplification is a two player HR derby league, we each get one from group A and one from group B

A GREEN - 45 HR BLUE - 40 HR

B RED - 20 HR YELLOW - 10 HR

First pick better take RED

The useful stats are

A GREEN - 5 HR BLUE - 0 HR

B RED - 10 HR YELLOW - 0 HR
Quote
0 #18 OaktownSteve 2013-02-13 22:20
The reason it's steep is that there's a lot of standard deviation in catcher values and it skews the numbers. More importantly, you really don't need it because you are not trying win the catcher draft. You want to win the draft as a whole (you actually want to win your league, but that's a whole other argument). The value over replacement tells you nothing about the tipping point where a catcher has enough value over the next available on the draft board to justify taking a catcher with less unadjusted value over a player at another position with greater unadjusted value relative to the next player at that position. And that's just a simplified model where you are looking at jsut two positions. The math goes bonkers when you look at the draft as a whole. Game theory covers this but it's some PHD level stuff.
Quote
0 #17 Todd Zola 2013-02-13 22:00
Quoting OaktownSteve:
I left a long response to Mike over there. I feel pretty strongly that positional adjustments are hightly inaccurate because the conflate the notion of value inside the draft context and outside the draft context. The only purpose of an adjustment is to inform the timing of a particular pick and that is not how the math is laid out. Ironically, I think doing away with positional adjustments dovetails in very nicely with where you are going with APE and the scarcity conversation. I've had a hard time expressing this concept to folks in the past. Wonder if any of my comments here are tracking with you.


But I'm not doing away with positional adjustments in terms of my initial valuation - I use a fantasy form of value over replacement. It turns out that only catcher needs that adjustment in today's landscape, but I feel it still needs it. As suggested, I am a bit surprised just how steep the adjustment is though.
Quote
0 #16 Todd Zola 2013-02-13 21:57
Quoting begavett:
No argument from me. I lean toward being more mathematically, inclined, so that part of me is interested in how you arrived at the 3xRound rule, apart from just your description. Obviously you've looked at data from drafts and what you're saying makes sense. Maybe you could follow up with a later post on how the 3xRound rule came about, and what happens if you are more strict (e.g., +/- $1) or relaxed (e.g., +/- $5) on your definitions of players being "essentially" equal.


The 3 x round is more empirical than anything.

Since the premise is +/- $2, I agree it needs more than the anecdotal handling I contend APE itself replaces (not saying that as well as I want) - sort of fantasy hypocritical

APE is a concrete means of proving something more abstract, but I am using something abstract to define APE.
Quote

Add comment


Security code
Refresh

Latest Tweets

 

LABRLOGO

xfl

toutwarslogo-new

Our Authors