This is made much easier to do by comparing the results of mock drafts, or as the compiled results are more commonly known: average draft positions (ADP). Most draft assistance programs (including FFToday's own Draft Buddy) present this information as an aid to improve your drafting.

But how much can you improve your draft by using ADP? Is there any way to quantify it? What, exactly, does ADP allow you to do? And finally, is there anything that might work better?

Before we can answer these questions, let's look at the details behind the acronym.

What is ADP?
A variety of different websites provide ADP data (e.g., Mockdrafts.com, AntSports). Regardless of the site, the composition of the data is similar: the averaged results of numerous mock drafts. Different sets of ADP data can be found, depending upon the nature of the source mock drafts used to generate the data. Typically, there are three types of variation in the source drafts: scoring systems, starting lineup requirements, and league size.

For a specific player, the ADP data includes three types of information:

• average draft position (round, pick within round)
• an indication of variation from the average
• the number of times that player was picked

For example, the ADP for the number one pick listed by Antsports is:

So, if this data is indeed accurate, you must have one the top three picks in your draft to have any kind of chance to draft Priest Holmes this year.

Uses of ADP
Most people use ADP data to estimate when a specific player is typically being drafted. But how does that translate into specific information for you to use in your draft?

When you compare ADP data to your rankings, you can use it to identify players who are bargains and target them early. You can also use it to identify players who are over-valued and hold off on drafting them, at least for a while.

And finally, you can use ADP together with dynamic drafting to assess the current demand for a position based on what's available when you draft next. This is a bit more complicated to do in a draft (especially without software), but it's really just an encapsulation of basic supply and demand.

Accuracy of ADP-based Predictions
The most accurate ADP data would technically be the results of your league's draft, after it happens. Obviously, that data is impossible to obtain before the draft. The next best type of ADP data would be from a series of mock drafts conducted by your league. After that, data from mock drafts with exactly the same rules as your league (scoring system, roster requirements, and league size). Worse yet, are data from mock drafts with deviations in scoring system and roster requirements. And finally, the worst data (having nothing other than random potential applicability) are from mock drafts that have different scoring systems, roster requirements, and league sizes than your league.

What type of ADP data is typically available? For me, it's typically the second to worst type, with the only similarity being the league size. I typically have to fudge the resulting data to get something that works for my league. I don't use the overall lists, but instead take the lists by position and then combine them to build something useful.

With all this uncertainty, is ADP data an accurate enough foundation upon which to base decisions? There's always going to be variation from averages. How accurate are the statistics generated from uncertain drafts?

The answer to that question depends in some part on your fellow league members. Are the players in your league representable by the average draft results? Or do you have guys that tend to pull serious left-field picks during the draft?

What we can do is to revisit the ADP data after completing your draft and see how accurately the ADP represented draft positions of players. Let's take a look at last year's ADP data and draft results from my league.

As you can see, 55% of the time ADP was accurate within twelve picks. This is a respectable number, but it still means that 45% of the time ADP was grossly inaccurate. That means if I had relied solely on ADP to determine when to pick a specific player, I'd have either missed out on them (13% of the time) or overpaid for them (29%). (This doesn't add to 45% as the missing 3% is due to players that were drafted but not included in the ADP data.)

An Alternative to ADP
Recall I mentioned above that the most accurate ADP you can get is that from your own league. Of course, we don't have access to that data. But if you're willing to give up knowing when a specific player is going to be drafted and instead concentrate on drafting trends as the source of your supply and demand decision making, you can extract that information from your league's historical draft data. If you don't have this data lying around, ask your commissioner for it, as they probably have it somewhere.

Looking at the last eight years of drafts for my twelve-team league, I can see how many players at each position are typically taken in each round. See the table below for this info.

As a point of comparison, we can generate the same table for ADP. A caveat: this data is only for last year, as I don't have access to historical ADP data beyond last year.

Now if we compare the ADP and historical drafting round-by-round data with the actual draft results from last year, we get the following tables. Green indicates an overprediction, and blue an underprediction from the actual draft results last year.

As you can see by a qualitative assessment of the number of black "0%", the historical drafting data gives a much closer approximation of the reality than ADP. That's to be expected, as the source data is specific to my league, whereas the ADP data is based on other people's rules and preferences. For the record, ADP overpredicted 32% of the time, underpredicted 26% of the time, and was accurate 42% of the time. The historical drafting data overpredicted 15%, underpredicted 18%, and was accurate on 66%. So quantitatively, historical drafting data was 24% more accurate than ADP at predicting the number of players by position selected by round in last year's draft.

Round by round historical analysis is fine and good for general comparison to ADP, but for true dynamic drafting, you must be able to compare arbitrary draft positions. This means you need to know what position will be drafted at each place in the draft. You can fudge it by assuming percentages of the round-by-round numbers based on the position of your next pick in the next round. The historical analysis of each pick is less precise, as the best it can do is give probabilities that a particular position will be selected at a particular spot.

To get around this element of chance, you must iterate to a valid checksum that ensures the total draft picture is an accurate one, and then use the individual position placements for the analysis. The checksum I use is the average distribution of players chosen by position over the period of time considered by the historical data. This, as you expect, is not trivial to do. But it works!

Final Thoughts
So with all this said, have we really answered any questions? Or have we just confused the matter further? It turns out that ADP data does have its uses, within reason.

If you want to get a read on when a particular player is going to go in the draft so you know when to draft them, ADP is your best bet. Be warned, however, that the "best bet" is most likely going to be less than reliable. Be prepared to miss out on players you've targeted if you get too greedy. To minimize your disappointment, make sure to use the ADP data that most accurately reflects your league's parameters.

For dynamic drafting, you can use ADP to estimate the number of players at each position that will be drafted between any two arbitrary draft spots by scaling the ADP numbers for each round based on the two spots.

If you want something more accurate than ADP, use historical draft data (if you have access to it).

Of course, historical data is only accurate when there is low (or no) owner turnover in league.

And if you want to go completely nuts, you can extend the historical drafting data to each owner in your league and use it to identify their drafting tendencies. Once you know your enemies' tendencies, you can exploit them. Mmmwwwahaaahahahaha!!!