Transcript

Hey guys. This is Max Steinberg, @SaberSim. I am in Miami in my hotel room and I’m coming out with you, with part two of my series of The Road to the DraftKings Fantasy Football World Championships. Really excited to be here, really excited about my potential lineup. I’ve already been thinking about it a lot and how to adjust people. And obviously I’m going to eventually build up to building my lineup with SaberSim, which hopefully it’s going to do really well. So I want to start out and this video is going to be a lot about research and process for me personally. And so, one thing that some of you are probably aware of is I’m a data scientist. I code in Python. I like using my Python background to actually use machine learning, do some research, do some analytics and create my own stuff to help me with my daily fantasy football.

And I think this is actually kind of unique to me. Obviously there’s great daily fantasy players who are also data scientists like myself. There are also a lot of daily fantasy players who don’t do this stuff at all and they’re quite good. You do not need to do this stuff to win. The reason I’m showing this to you is I really wanted to be transparent and show you what my full process really is. So that’s why I am doing this video and showing you all this stuff. So I am excited to show you this. I think it’s really cool. I have created some really interesting things that I think are helpful and I’ll just get right into it.

So I’ve created a few tables and I would say that my crown jewel table that I’ve created is something that’s called Expected Fantasy Points. So you might have heard of this somewhere. And with all my tools, and especially this one, Expected Fantasy Points are a way of me evaluating what has happened in past NFL games, in a way that’s a little better, that takes away some of the variants. So what I did is I created a machine learning model using what’s called a random forest. Which is a decision tree model. You don’t really need to understand this.

But I use a machine learning model and I looked at play-by-play data and instead of a result of a given play being what actually happened, i.e., Saquon Barkley have doing a 50 yard run and getting five fantasy points, I calculated using this machine learning model, or I didn’t do it, the computer did this. Calculated how much Saquon Barkley was expected to score in fantasy points given a running play. So where he might’ve scored five fantasy points for running for 50 yards, my model’s probably only going to give him about a half a fantasy point. And obviously if a running backs’ closer to the goal line, he’s expected to score a touchdown more often, then that’s going to raise the expected fantasy points. Catching a pass has a different expected fantasy points based on the depth of target.

So it basically combines a lot of factors and says, “Okay, given the opportunity this player had this week, i.e., a rushing attempt, a target, or something like that.” I think those are the only two things. “How much was he expected to score, fantasy point wise?” And I think this is a really good way to look back at what’s happened and make some predictions about the future. I’ll just get into a couple other things. I do the same thing, again using a lot of different play-by-play data. I’m just using NFL play-by-play data, which is free. You can get it from a cycled NFL scraper, you can get it a lot of different sites.

Then I use that play-by-play to do other things. And one thing I really love is something that I call Pass Above Expectation. And so what this is, is something again, using play-by-play data. It looks at every play and whether team ran it or passed it, but it adjusts for it given the context of the situation. What was the down? What was the distance? What was the difference in score? Obviously if you’re trailing a lot, you’re going to have incentive to pass more. If you’re up a lot, you’re going to have incentive to run more. Everyone knows this. Also as well, the quarter. How much time is left in the game? That’s going to have an effect as well.

So I took all of these factors and I made a machine learning model that would say, what is the expected pass percentage or run percentage for an average team in this situation? And then weighted that against how often they actually passed. And these give a lot different of a picture then you might be used to. For example, Washington has been expected to pass over the last four games 60% of the time. However, they only passed 54% of the time, which means, over the last four games they were the most run heavy team in football, adjusted for context. In the contrary, Kansas City… Kansas City is a pass heavy team. I think everyone knows this, but if you’re just looking at their pass share the last three weeks, they passed it 58% of the time. Not a lot. However, they’re only expected to pass it 50% of the time.

So they actually, according to my model, are a very pass heavy team. And this just weeds out some of these random variant factors, or flawed games, or just luck with down and distance, that you might not be taking into account. And it really shows which teams are really pass heavy. It usually really came in handy for me when I qualified for the DraftKings Fantasy Football World Championship because I noticed that Chicago is actually a really pass heavy team. They have a huge pass tendency. It allowed me to play a double stack of Mitchell Trubisky, Allen Robinson and Anthony Miller and was key for me winning the seat in the first place. So this is something that’s really helpful for me.

So let’s just get into it. I’ve just put stars next to some players that stood out to me, for one reason or another. And I just want to go through them to show you what I’m looking for. So the first star I have is Larry Fitzgerald actually, which stood out to me. So let’s just go to my DraftKings account real fast. And while I’m doing this, I’ll just talk about this a little bit. Larry Fitzgerald is only 4,300 in this fantasy football world championship contest. He had a real hot start of the season and had a real lull in the middle. And now in his last four games, he’s averaged about 13 expected fantasy points. Again, this is the opportunity he had given his targets, the depth of targets, the yards and fields, etc. How much he was expected to score given that over his last four games. And 13 is really not bad, actually. I’m pleasantly surprised by this number. He is only 4,300 on DraftKings, as I said. He’s a guy that I think I certainly will be considering and let’s just wait for this lineup.

So he’s someone that you could consider, and I’m sorry I’ve repeated myself, but just going here. With someone like Larry Fitzgerald 4,300, you could stack him with Kyler Murray. I think this makes a lot of sense. He’s only 5,600 so that makes some sense here. So another player that I put a star next to is Ian Thomas. So it’s Friday, Greg Olsen’s ruled out. This is a real impressive number for a tight end. I think in general they don’t have this high of number and this is a game script. Carolina is a six point underdog where they probably are going to have to pass the ball a lot and even if they don’t, I don’t think they’re going to crush Seattle. It’s certainly going to be a game. He is a guy who’d played a lot of slot snaps according to Adam Levitan, who I followed on Twitter.

He clearly seems like a pretty good player. He’s a second year player and if you look at him you see, okay 3,100 for Ian Thomas. That’s a real good price for a tight end who is capable obviously of having a really big game. But the expected fantasy points as well, follow. And you can see he had 10 targets last week. I think it makes a lot of sense to go somewhere with Ian Thomas. He’s definitely someone I’m going to project highly and we’ll see if I get him in my lineup. Another guy, let’s see. Okay, yeah so I wanted to make a couple stars. Well I actually missed this, Denver situated. So Denver I think is real interesting because they’re 10 point underdogs and I think that there’s one player here or a couple players here that do stand out to me.

Noah Fant actually stands out to me for a different reason which is, he obviously have had some real good production in a couple of games. But as you can see he, in terms of expected fantasy points, he really has not been popping off that much, save for one game. Jeff Heuerman is back. DaeSean Hamilton is back. Tim Patrick is back. I think this takes a lot of opportunity away from a guy like Fant and I personally think he’s someone that I’m going to probably not even consider, at the tight end position. Phillip Lindsay on the other hand, he was named the starter in recent weeks and it’s really shown. You can see with Royce Freeman, his production has not been as good as the recent past. And Phillip Lindsay has really gotten a lot better. He is, let’s say 5,600. It’s probably a little too high priced for my taste, but I think he is someone who might be in consideration. I’m definitely going to project him highly.

But the real guy that I’m really into here is Courtland Sutton. He has many games of over 19 fantasy points. He averages 17 expected fantasy points of the year and 17.67 in his last four games and his price does not match that at all. It sees only 5,900 in a game script where I’m pretty certain his team is going to pass it a lot. You can say that with certainty when they’re 10 point underdogs. And I think honestly Kansas City should be more of a favorite. And so I think I like him as well and I think it’s actually good to compare with SaberSim, what we’re doing here. So Sutton, they have him 14.22. I’m almost certainly going to boost him. Obviously expected fantasy points says that he should be closer to 17. I’m not going to follow exactly, but he’s certainly worthy of a boost.

Phillip Lindsay, 13. Again, I think he’s someone that I could raise as well. I’m going down to Danny Amendola. So we have TJ Hockenson, is out. Marvin Jones is on the higher, both are the higher. So that leaves a lot of opportunity for a couple of receivers here. Kenny Golladay and Danny Amendola. Danny Amendola is specifically 4,100. Really cheap, he’s at home. Tampa Bay, by the way, one of the worst pass defenses. I’m sure you know this. And teams actually passed against them a lot. So he’s someone that is definitely going to be on my radar. I remember looking at this before, I think we protect him very highly. So I don’t know if I’m going to take this strong of a stance. I think this is a real strong stance and I’m going to lower Detroit’s passing game in general because it seems like these numbers are really high. And I feel a little weird about that given it is David Blough who is starting at QB. But obviously there’s some numbers that support that even without these injuries.

Danny Amendola, 10.9 expected fantasy points in his last five games. And so that number obviously is going to rise. So having a projection that’s 12 or 13, I think makes a lot of sense. And obviously at 4,100 that would put him at some of the top consideration in any lineup. So scrolling down a little more, Tyler Higbee. So at this moment, I don’t know if Gerald Everett’s ruled out, I think he’s going to. If you look at his numbers, it is just completely shocking how well he’s doing. He averaged 15.8 expected fantasy points in his last three games. And obviously we’re looking at Ian Thomas. Ian Thomas has done, better than him in a one game sample. But that’s going to be harder to trust, than this pretty massive three game sample for Tyler Higbee. He is only 3,900. He is someone who I’m almost certainly going to be high on. SaberSim has him at 10.3. I’m certainly going to be higher than that. He’s a candidate for a boost, absolutely. He’s someone who I think might almost certainly land in my DraftKing signup.

Patrick Laird, another guy. Going back to this sheet with the pass above expectation. One team that’s really high there is Miami, right? Miami actually passes it a lot and they’ve actually been that way all season. DeVante Parker and Albert Wilson are questionable. I think Parker plays. Albert Wilson, it’s more unclear. But if you look at this, Patrick Laird averaging about 16 expected fantasy points a game, last two games. If he can get 16 fantasy points at 4,500 that’s really good. And that’s not even considering a pretty cake matchup at the New York Giants. And not considering that his target share is historically in college been an unbelievable pass catcher.

If there’s wide receiver injuries, they’re probably going to split them out wide more. And what that means is he probably is going to get even more targets. I think he’s a great value at 4,500. I also think he’s going to be a very high owned play in this tournament and that’s something I’m going to touch on in another video, when I’m actually building my lineups, is how to adjust ownership projection. But he certainly is going to be someone who I’m going to project highly. So funny thing is, I started Jamison Crowder. I wish he was on the slate because he ended up going off for two TDs. But he was someone I liked and oh well.

Let’s see here, if we’re going down a little more. Ooh, last one, so this is a guy who I think is going to have some popularity, is A.J. Brown. The Titans are projected to score 27 points, I think by the sportsbooks and A.J. Brown had a big game. He’s very expensive. I think he’s almost 6,000. He is 6,000. Some people only think, “Wow, Titans are predicted to score a lot of points passing. This makes a lot of sense.” However, I’m real skeptical because the fact of the matter is before this, he really didn’t have any huge opportunity games that popped off like this and you can see has a lot of 9 expected fantasy points, ten, he had one with 13. That was a really good one. But I have a hard time believing he’s going to have another 16 expected fantasy point game.

And when I’m gauging this, I’m just really thinking about opportunity here because there’s a lot of variants in game-to-game fantasy football. So I’m probably not going to project him that high. I think SaberSim probably agrees with me, it’s an absurd take. 13.58, I think that’s reasonable. Maybe you could go a little higher. Yeah, I think in general someone who’s probably going to get some ownership and I’m just not really going to be on him.

So just wanted to show you that little behind the scenes thing with some of my tables. Again, I’m a data scientist. I don’t think a lot of people look at this stuff. And you can also come to the same conclusions by just literally looking on Twitter at sharp people. Doing your own research. You don’t have to get as advanced as this, but this is something that I have some skills in so I like doing it. So this is about to conclude part two of my video. Going to have a part three where I start talking about how I actually am adjusting projections, how am I adjusting ownership, how I use SaberSim. And this is going to be a lot more SaberSim happy. So I hope you’ll keep an eye on it and watch out for that video. Hopefully it’ll come out before Sunday.

But again, this is just my process. I hope that you enjoyed this video. SaberSim actually is offering a three day free trial, so if you want to try it out, three days you can try it out for free. We really are the best lineup builder in the business and the easiest one to use, and probably the most sophisticated as well. So try it out. I hope I inspire you to make some lineups this week. And good luck whether you’re playing in the FFWC or you’re just playing the Millionaire Maker or you’re playing a $1 tournament. Thanks guys.