THE BLITZ Ownership Projections (BETA) FAQ

by Derek Carty (DerekCarty) Part of the product THE BLITZ Projection System by Derek Carty

For the first time ever, I’m excited to announce that in 2022, a subscription to THE BLITZ gets you access to Ownership Projections! With the help of Deuce of Diamonds Analytics, I’ve been working on these for a while, and I think they’re really fantastic. THE BLITZ Ownership Projections are being released in a beta phase. While they have been thoroughly back-tested and monitored ahead of time, unexpected things can always pop up once a project sees the light of day. As always, all feedback is welcome. Best of luck out there! — Derek Carty

Why do Ownership Projections matter?
In order to be truly successful at DFS, there are three key components:
1) great performance projections (like THE BLITZ!)
2) great ownership projections (like the ones you’re reading about now!)
3) smart game theory to combine the two

While many DFS players think their knowledge of the sport is the most important thing, it isn’t. DFS is a game of math and probability, and all the most successful players embrace this. Even a bad football team may, over a full-season, have a decent chance of having one or two high-scoring games. Put them in a fast-paced game in a dome, with a condensed target tree, and they could win someone a GPP. Winning a GPP isn’t about playing the best teams, it’s about playing the teams that have the best chance to do well relative to how many of your competitors are going to own them. If the that bad team has a 5% chance of being the best team today, but less than 1% of the field is going to own them, creating lineups with them would be a +EV move (i.e. a move that would be expected to be profitable long-term). If you continually make +EV lineups, over the long run, you will very likely end up profitable. Successful GPP players do lots of losing. Lose, lose, lose, lose, BIG WIN. Those big wins, which may only happen 5% of the days you play or less, make up for all the losses and more. If you have that mindset and you build good lineups using smart game theory, you will be profitable long-term. And in order to build those good lineups, you need good performance projections and good ownership projections. THE BLITZ now offers you both.

For more on the game theory aspect of this, be sure to check out Blender’s audiobooks, The Theory of DFS. I can’t recommend it enough.

How were THE BLITZ Ownership Projections created and what goes into them?
THE BLITZ Ownership Projections have been developed over the past several months using highly sophisticated machine learning techniques, such as XGBoost. You don’t have to know what that means to use them, though — just know that they are highly advanced! These are the types of techniques that actual sports teams use to develop their own internal processes.

On a more practical level, the more important thing to know is that they have been created using historical ownership data. Our algorithms looked at the actual DraftKings ownership rates of players throughout previous seasons along with several hundred different variables to determine what actually drove ownership. To the best of my knowledge, no other industry ownership projections are built this way. From what I understand, other ownership projections are built based on what the creator thinks drives ownership, but THE BLITZ has actually tested to see the reality. The projections have been tailored specifically for the DraftKings Millionaire Maker multi-entry tournaments for the main slate each Sunday. That means they will be most accurate in projecting ownership for those types of contests, but they will be similar for other types of GPPs (particularly multi-entry ones) and ought to be at least directionally accurate for Single Entry GPPs and cash games.

Will you have Single Entry or Cash Game Ownership?
They aren’t part of the initial launch, but it’s a possibility. I need to work out some logistics for it. For now, though, Ownership Projections are geared only towards Multi-Entry GPPs.

What sites do they work for? Do they work for every slate?
THE BLITZ Ownership Projections currently work for DraftKings, FanDuel, and Yahoo!

It should be noted, however, that the DraftKings version of the projections will be the most accurate. DraftKings provides users a way to see and save the actual ownership rates in any given contest. This has allowed my team and me to base our projections on this data and what actually drove ownership historically. FanDuel, however, does not allow users to collect this data. Therefore, the FanDuel and Yahoo! projections are built in a more theoretical way using the knowledge we gained from what drives DraftKings ownership. I’m confident that these projections are still very good. But by necessity, they can’t be as good as the DraftKings ones, and we haven’t been able to back-test them the way we could the DK ones.

While THE BLITZ Ownership Projections were built for the main slate on any given day, they will also work for most other slates. These same caveats apply apply here, though. The projections weren’t specifically designed for smaller slates, but the model should do a good job of projecting them regardless.

The projections won’t be available initially for the very small, single-game Showdown slates or tiers formats. I am in the process of developing Showdown projections, but because the format is different than the main slate (they include kickers, an MVP/Captain spot, etc), it’s not as simple as just scaling the Main Slate model to Showdown the way I can do with, say, the Thursday-to-Monday slate. There are some fundamental assumptions that will need to be changed, which will create a bit more uncertainty, but I’m confident the end result will still be quite good.

How accurate are THE BLITZ Ownership Projections?
As part of our quality control process (i.e. in order to make sure our projections are actually good), we tested them against two of the more popular existing sets of ownership projections from around the industry. While the results won’t be made public (and the two comparison sites won’t be named), if you’re willing to take my word for it, I can say that THE BLITZ’s ownership projections were the most accurate in our internal tests. I’m very confident that THE BLITZ will be delivering the best ownership projections available anywhere.

Is there a manual component to the projections?
Because the projections are generated automatically using machine learning, there is no direct manual component to them. It’s possible that certain nuances get missed as a result, but regardless, the back-tests have shown them to be very good. And on the other side of that coin, a strictly automated approach means that we’ll be avoiding human biases.

Sooooo, what types of things actually drive ownership?
While I can’t divulge too much, I can say that basic things that you would expect to matter are being considered. Player quality. Matchup quality. Opposing defense vs. position stats. Vegas totals. Recent performance. Industry perception. (And, of course, lots more).

Is there anything THE BLITZ Ownership Projections don’t account for?
Obviously it’s impossible to account for everything, but I believe THE BLITZ Ownership is pretty thorough. One thing that I can think of that is not in the model is the likelihood that Thursday night players will be higher owned on the Thursday-to-Monday slate than they would be if that game were part of the Sunday Main slate. The game is on an island, people watching the game want to own a piece of it, and so those players will likely be a bit higher owned than they should be. But because the model was built specifically for the Main slate (and is then scaled to the Thursday-to-Monday), that wasn’t a variable that was able to go into it.

Where can I view THE BLITZ Ownership Projections?
For now, you can view them inside LineupHQ. Soon, I’ll be adding them (on a team-level) to THE BLITZ Stacks page and (on a player-level) to the Advanced GPP Metrics Page, perhaps with some additional information to help you get the most out of them.

How often do THE BLITZ Ownership Projections update?
They update every time THE BLITZ player projections update. Due to the nature of football projections, there isn’t an exact time or schedule for these, but you can find the rough schedule on THE BLITZ FAQ page. Changes may sometimes be small or non-existant, but the same as THE BLITZ does, the Ownership Projections are constantly taking into account the latest information. And the same as the THE BLITZ performance projections, the ownership projections will become more accurate as the week goes on and more information becomes available.

Will you be creating special tools for them?
Yes, that’s definitely the plan. I’d like to create tools that let you compare a player’s projected ownership with what THE BLITZ thinks their ownership ought to be based on their performance projections. I’d like to create tools that lets you see who the best leverage plays are, which players and stacks project to be the most over/under-owned, and that sort of thing. Step 1 was creating the projections themselves, but now that they’re completed, these types of tools will likely begin development soon to help you gain additional edge from the ownership projections.

If THE BLITZ Ownership Projections are so good, and people start using them, won’t they start to influence the ultimate ownership?
This is a high-level concept, and the answer is unfortunately “yes, probably.” Many DFS players know, at least on a basic level, that in order to win GPPs you need to be “contrarian”. You need to roster players that other people aren’t going to have. And how do you know which players and teams other aren’t going to have? By looking at ownership projections. As THE BLITZ Ownership Projections begin to gain influence, it’s possible that players who would have been high-owned before now become less high-owned simply because people are looking at THE BLITZ’s projections and trying to be contrarian to who they think will be owned. Our back-tests were based on a world where THE BLITZ Ownership Projections didn’t exist. As soon as they do exist, the assumptions the model is making are no longer exactly the same. This is absolutely something we’re aware of and will be monitoring and adjusting for as much as is possible. But it’s also something for you to keep in mind as well as you use these. The good news is that we don’t have to be perfect. Perfection doesn’t exist when we’re predicting the future. Lots and lots and lots of people play DFS. They won’t all be looking at these projections. And so even if they hold some influence, it’s unlikely they completely reshape the game.