How to Beat MMA DFS (Master UFC DFS Strategy Session)

If you haven't tried MMA DFS, you are missing out. Not only is it one of the best DFS sweats, watching each fight play out, rooting for the fighter that you're more exposed to to get that early knockout or submission, but it's also one of the most profitable DFS sports in the lobby in 2024. Our contest sims show that profitable UFC lineups have some of the highest ROI of any DFS sport out there, but this is only true if you know the right strategy to beat it.

I recently sat down with our data scientist Will, one of the top MMA DFS players in the world and the mastermind behind our fight simulations, to learn what makes his strategy so successful. In this video, I'll share Will's top three keys to beat MMA DFS in 2024. But make sure to stick around to the end of the video, where I'll put it all together and show you how you can build 150 profitable MMA DFS lineups in just a few minutes using SaberSim.

Key number one is to understand the range of outcomes of each fight. Just like any DFS sport, to take down big UFC DFS tournaments, you need upside. To win a tournament, you're going to need to put together a lineup of six fighters that all win their fight, but it often takes even more than that. Fighters hit tournament winning scores when they dominate their fights.

I'm talking first round knock ups, racking up a ton of control time, and beating their opponents down with significant strikes. A slow paced fight that ends in a decision is often not going to be enough points for a top finish. The problem here is that the average projections that most people use to build their lineups don't correctly account for the upside potential of each individual fighter.

Underdogs that fight fast and loose and have the massive potential for a huge score if they can manage to win their fight have upside that isn't captured in an average projection, and big favorites that fight conservatively and are very likely to win a decision will be overvalued by these numbers.

Average projections are heavily weighted by the Vegas win odds, but it's equally important to understand what a fighter scores when they win their fight as it is to know how likely they are to win. Here's Will giving the perfect example of the bias that average projections have in MMA DFS, discussing Renato Moikano in UFC 300.

I mean, Renato Moikano would be a fun dog in the sense that like, if he wins, he's gonna win in like the first round or two. And sure enough, win this card actually played out, what happened? Moikano had a huge early win and ended up with a massive point total, landing himself in the top lineups for that slate.

But this goes one step further when you start to think about how two fighters interact together in the octagon. These fights don't take place in a vacuum, and the strategy used by a fighter depends on what their opponent is doing. What does the upside potential look like for a fast paced striker when they match up against Grappler seeking a submission?

How much more should you be willing to play the heavy underdog that is likely to knockout the opponent if they win versus the heavy favorite in a five round fight but that is most likely to win by decision? You need a simulation to answer these complicated questions. A simulation goes beyond the average projections to play the fight out thousands of times taking into account not only the vagus odds for the fight but also the pace and style of each fighter.

Let's hear again from Will about how SaberSim's robust MMA simulations take this data into account. So, we're doing all of the, the line based stuff, uh, in terms of like the win probability, the probability by knockout, that stuff is all straight from Vegas. Um, we're just taking a bunch of books lines, I weight them by how sharp I think the book is, um, and what lines they have available.

Um, and those all get input. So indirectly, like, you know, if they're, you know, when my knockout comes out to 30%, then we're saying we're, we're designating 30 percent of those Sims, they win by knockout. And so that is like, that's the core, those projected, like those points are going to be uniform across the industry in terms of like the finishing and decision, win points.

Anybody can just take Vegas Lines and do that. Uh, the rest of it comes from our fighter stats, which we take an approach similar to what we do for other sports. Yeah. Like, uh, you know, like we look at the historical data, we apply, you know, decay to it, uh, we do, you know, a lot of advanced statistics stuff to basically filter out the noise, get the signal.

Um, so it's taking that, um, and applying it to like their takedown rates and everything like that. Um, and then we, we train a model based on the fighter, um, their, what their opponent is like in that fight. So, you know, if you have two grapplers facing each other, They're more likely to not grapple because they both can grapple.

Whereas if it's a grappler versus a striker, the grappler is much more likely to go higher on than his average grappling type thing. Um, so that it's basically like taking those combinations of things. and the model then updates to like, here's what the projected mean, um, you know, strike strike attempts per minute is for this fight.

Uh, and then we're just taking a distribution of that in each. SaberSim simulations incorporate the odds for each fighter to win and win inside the distance, but it goes beyond that in our simulations. And when you build lineups on SaberSim, each lineup you build is the optimal lineup for a given way that the entire card could play out, accounting for this hidden upside that averages ignore.

I'll show you more about how that works in just a moment, but first it's time for my key number two. Key number two is to consider the cost benefit analysis of each fighter. Now my first key here was all about measuring the upside potential of fighters correctly, but that's not the only consideration in your lineups.

Fighters have other costs that we need to consider before sticking them in our lineups. They of course have salaries that it costs to roster them that are based on the opening odds to win for each fighter, and we also have the cost of the ownership of each fighter we roster in our lineups. Because we're competing against other opponents lineups in our contest, the more owned a given fighter is, the less relative advantage we gain against other lineups when that fighter scores points.

A fighter having a huge upside outcome at 10 percent ownership is a significant more boon to our lineups. than a fighter scoring the same amount of points at 50 percent ownership. Now, I know I'm not reinventing the wheel here, and you probably already know that ownership and upside versus salary is important in DFS, but in MMA it's particularly interesting because these different factors influence each other so much.

A given fighter that might be unrosterable if you took only his projection and salary into account becomes a great play once you account for the low ownership of that fighter. And vice versa, a slam dunk play based on upside versus salary alone might carry so much ownership that it becomes hard to build great lineups that include that fighter.

Will provided two excellent examples of this at play for UFC 300. First, Justin Gaethje, who Will expressed some concern about given that his salary was so high relative to his win probability, but who was also still garnering a ton of ownership on the slate. Like the thing, like when you're playing Gaethje, you're playing Gaethje at a price higher than what his money line is currently.

So I like, you know. In the purest sense, like if you removed everybody else in the field, you wouldn't want to do that. Right. You have to be okay that that's a higher variance play. And another great example in the opposite direction is Kayla Harrison, a fighter actually underpriced relative to her win probability, but with an extremely low ownership projection.

because of a perceived low upside when she wins. An alternative kind of similar thing might be like, um, on this, on this site with Kayla Harrison is making her UFC debut, uh, against Holly Holm. And she is probably going to be like the lowest owned of the big favorites, um, because she's expensive. There's a lot of other people you want to get to, um, and her upside really isn't the same as the other people at her price tier.

Um, But, like, that's a, a type of play where, like, you're still taking the risk in that, like, her upside is lower, um, but she's low owned and she is very likely to win. Like, she's as likely to win as, if not more likely than her salary suggests, um. So it's, it's not, it's at least a different type of risk.

You're, you're, instead of taking win risk, you're now taking how well do they score in a win risk. Um, but that's, uh, at least a, a similar, but different kind of stand ultimately a good MMA DFS lineup is a lineup where the total upside of the lineup as a whole outweighs the combined costs that you're paying for that lineup in the form of fighter salary and ownership.

But even a fantastic lineup on paper by these metrics can end up being a terrible lineup once the contest locks, which brings me to my third key. Key number three is avoid duplication. Duplication is unequivocally bad for the expected value of your DFS lineups. And to be totally honest, it bothers me that there is this ongoing debate about whether duplication matters in DFS.

Because we can show very quickly with math that it does. Now, to be clear, before I do that, I'm not saying that any lineup that is duplicated in a contest is immediately unprofitable. But it is a fact that each time a lineup you played is played by somebody else in the contest, The expected value of that lineup goes down.

Here's why. To calculate the expected value of a DFS lineup, you take the probability of that lineup finishing in every possible position in the contest. Multiply that probability by what the lineup wins or loses when it finishes in that spot, and then take the sum of that entire set. But in top heavy GPPs, the expected value is heavily weighted by the lineup's probability of coming in the top 10, top 5, or first place.

For simplicity, let's assume that you're playing a 10 person, win or take all contest, and you have a 20 percent chance to win the contest. You're going to have a very high expected value in this contest. With no rake, you're going to be winning that contest twice as often as your average expectation is.

and 10x'ing your money each time you win. But now let's assume that one other person in the contest plays the same lineup as you do. Now, when you win, you split the pot with that other player. You can probably already see how this has cratered your profitability in this hypothetical contest. While you are now slightly more likely to win, because you're competing against 8 other lineups instead of 9, the upside potential for you when you win is cut in half.

Each win you earn now pays half as much as it would have if that other player did not play the same lineup as you. This trend continues into large GPP tournaments, and the profitability of a given lineup goes down each additional time you're duplicated in a contest, because you're sharing winnings with each player that duped you, without making it really any more likely that you win.

And this is exactly why most of the top MMA DFS players in our contest flashback tool will have so few dupes in their lineup portfolios. These top players know that duplication is bad for their expected value, so they intentionally take steps to avoid it in their lineup construction. And duplication is particularly important in a sport like MMA.

With a smaller player pool, six man lineups, and big contests, Some of the most duplicated lineups in a contest might be duped hundreds of times. But how can we predict what lineups are most likely to be duplicated? Well, one of the best ways is with the salary used in a lineup. Most DFS players using optimizers or building their lineups by hand are extremely likely to use the maximum amount of salary available or very close to it when building their lineups.

And while salary is generally correlated to fighter upside because the salaries are based on the opening win odds for fighters, there are plenty of profitable lineups available that leave salary on the table. And this is once again especially true for UFC DFS. We've already seen that underdogs in these fights sometimes have just as much upside as some of the favorites if they can manage to pull off a win, so you can absolutely build winning lineups that leave salary on the board.

I recommend leaving upwards of 500 salary on the table in large field GPPs, and even in smaller contests, avoiding lineups that use the maximum amount of salary. This is a great, low cost way of avoiding dupes in your lineups. Now, we've covered a lot here, and it can be hard to figure out exactly how to piece all of this together.

Take care. How do you go beyond average projections, weighing the styles and pace of different fights to measure fighter upside? Compare that to salary and ownership, all while building lineups with a low probability of being duplicated? The easiest way is to use an MMA DFS simulator. Building lineups based on fight simulations and scoring your lineups using a DFS contest simulation to see how they're expected to perform against the lineups your opponents are most likely to play and with the payout structures the contest you're playing have.

Let me show you exactly what my process looks like for getting this done in SaberSim. When you pull up SaberSim, you'll immediately see we have the same average projections as other tools do, but we have a ton of other fight data available here on this screen. And, uh, All coming from our round by round fight simulations for each fight on the card.

The first thing we want to do here is just build some lineups, which is going to allow us to use the data from these simulations and build out an optimal lineup for 5, 000 different ways that the slate could play out. All I'm going to do here is make two changes to abide by the DraftKings rules, change the number of lineups I'm requesting to 150, and change the maximum salary down to 49, 500, which is leaving 500 salary on the table to try to get lineups that are a little bit more unique in the contest that I'm playing.

The next step here is to just click build lineups here, which is essentially going to simulate out this card 5, 000 times, building the optimal lineup for each contest. each of those 5, 000 simulations. Because these lineups will be built using the individual fight simulations, they go beyond what averages can do alone, taking into account the upside that big underdogs have when they win the fight, and appropriately weighing the upside of big favorites based on whether they're winning by decision, landing a huge knockout, and putting up the big scores you need to be successful.

Now that our lineup pool is built out, we can next run a contest simulation. When we run a contest simulation, we're taking each lineup in our pool and competing it in a hypothetical version of the contest that you're playing on that slate. We're playing that out using the fight simulations a hundred thousand times to directly calculate the ROI of each of each lineup in the pool.

This is essentially solving the cost benefit analysis puzzle and the problem of duplication all at once. By taking a lineup and seeing how it would perform in the contest that you're planning on playing, we can weigh the salary, the ownership, and the upside of each fighter in that lineup against all of the other lineups in the contest and see which ones are making the most money.

After the contest sim is complete, all it takes is sorting by the ROI for the contest that you're playing, which will immediately identify the most profitable lineups from the pool, taking everything into account. After we've sorted by ROI, if you're an MMA expert or are spending a little bit of additional time doing some research on your own, You can make adjustments to the min or max exposures of different fighters to take stands, but for me, I'm happy with these top 150 and I'm ready to enter them into my contests.

All it takes is clicking the save to my contests button, saving these lineups to the contests that I'm playing, downloading the entries, and heading over to DraftKings to get these uploaded for the A simulation is like a cheat code to solving the complicated problems in MMA DFS, but if you want to maximize your profit when using them, you need to have a deep understanding of how they work.

Next up, check out my comprehensive guide to DFS simulations, and become a sim expert for the next MMA card. I'll see you in the next video.

‍