New Updates & Improvements to the SaberSim 2024 NBA Model

Jordan: All right. What's going on, everybody? It is opening night of MBA here for the 2024 season. So we are back with another video talking about the MBA model for this year. Um, how it works, how our Sims work, some of the big improvements we've made for this season to help out with that. We've got Eric and Will joining today.

How are you guys doing?

Cool. Well, let's go ahead and get right into it here. I think as a good place to get started just to set the stage to either of you guys kind of just want to give like a general overview of how the NFL, the NBA model works, what some of the main inputs are, like what's going on behind the scenes when somebody opens up SaberSim, looks at the projections for the first time.

Wil: Yeah, I mean, uh, I I can sort of give like a broad overview and then I think, uh, Eric and I, there's like two main pillars and, and Eric and I worked on like the, the, each side of the pillar. So I think we can each dive into like our side of it more sweet. Um, but basically, uh, especially this year, we're a very Vegas heavy model.

Mm-hmm. Um, and so we have, um, basically all of the, the props that are on the board for each of the players. Um, and these are setting up. Um, a lot of what our, like, input rates are, um, and as those go into the SIM. So it's, um, in previous years, when we were running an MBA SIM, we would have, um, all of our own, like, projections for these rates.

We'd have, okay, like, what is the percent chance that this player has the ball? Uh, when this player has the ball, what is the percent chance they're taking a shot? Um, and so on and so forth for all those things. Um, and this year, we're leaning a lot more into Vegas for a lot of those rates. Um, so you can intuit or you can solve for what the probability that the player has the ball based on, you know, okay, well, if they have this many, this is their points.

And this is their teammates points prop. And, um, you know, you can sort of intuit out, you know, all of the different rates that we use. Um, and so we've made a pretty big switch into using those, uh, in the MBA sim. Um, and so it runs, um, sorry. The, yeah, we're using the props as the input now, um, is like a major switch that we made.

And then we've also upgraded our minutes model. Um, so it's, it, Eric, can you take that over?

Eric: Yeah. Yeah. Um, so the, the minutes model we just like developed internally, um, essentially like the, the big challenge with minutes in the NBA and like why it's important is obviously you have to be on the floor to score fantasy points.

Um, so it's important to understand like the types of lineups that are going to be on the floor. Um, and then, you know, the other thing is obviously with news, uh, you want to make sure that you're getting the right players, uh, playing the right amount of time. Um, and so, yeah, what that looks like is, you know, just looking back at, uh, previous games, uh, where players have started or come off the bench.

Um, there's Sort of like a positional type adjustment. Uh, that's one of the trickier parts of NBA now because the positions are just so intertwined and, you know, it's sort of like abstract. But, um, you know, obviously you don't want to put, like, 5, 7 footers on the floor, uh, unless you're Minnesota, maybe, but, um, that's like, you know, a part of the, I guess, just like the overall, uh, Model that, um, that we worked on this year.

Jordan: So, yeah, well, you mentioned 2 pillars. So if I understand it correctly, we're, we kind of have the internal model is setting who's going to be on the court, how many minutes they're going to play. And then we're relying on the Vegas markets to kind of say, what are they going to do when they're on the court?

What are the, what is the rebounding rates rebounding share? Um, all that kind of stuff.

Wil: Yeah, so I see the 2 pillars as minutes, which is the side that Eric worked on and then converting the props and all of that into the rates that the SIM uses itself. Um, and so, like, that's the, the side that I've spent the most of my time working on.

Um, so, yeah, it's, it's just the, the way that we take, okay, like, this player has a 28. 5 points prop, um, and it's slightly juiced to the over, um, okay, so that means, you know, 28. 7 points or something on average, um, and then how does that convert into all of the rates that we need for the sim?

Jordan: Cool. So the last time we had talked about this kind of Vegas model approach for NFL, um, we had foreshadowed some of the stuff for NBA talking about, like, you know, there's going to be sometimes less props, sometimes less liquidity in those markets, um, sometimes just less players on a given team that are even going to have props posted.

How did we end up thinking about and overcoming some of those problems as it relates to NBA?

Wil: Yeah, so we, we lean pretty heavily into our priors, um, for players that aren't going to have props. Um, we have a combination of priors. We have it from a couple different sources. And so we have, you know, per minute priors, uh, and then we also have, uh, sort of to keep something from our previous sim.

We have those as their per possession priors. Um, so we have, like, baseline rates. So, you know, we could run the sim exactly as we did last year without these, these prop inputs. Um, and so we're using that sort of as a prior. So, you know, if this player is missing props, we can fall back and no one know, like, okay, this is a pretty decent projection for them.

Um, and obviously those update throughout the year. So, like, rookies, as we get data, like, those will, like, match their play styles, et cetera. Um, but we're, we're basically, we have 2 to 3 sources of priors and we use that, uh, to accommodate any missing props.

Jordan: Gotcha. Does the Vegas stuff kind of rest on top of the priors, or does it almost just override it?

Like, are we, if, if our priors do kind of disagree with what the lines look like, are we going to say, no, like, we're going to trust the wisdom of the crowd here a little bit and go with props here? Yeah, we're always going to

Wil: go props. Gotcha. It's just, that's the methodology that we want to go with. And especially, um, because it, you know, It would be different if, you know, we just agreed on, like, a full team or something like that.

Um, but, like, I think it could be sort of mess up, like, the usage calculations, et cetera. If we were just like, uh, off on, like, 2 players or something, the rest of the team the same. Um, so it makes sense to standardize where we can, um, because, like, the prior stuff is mostly going to be filling in, like, the. 9, 10, 11 guys on the rotation, like the 9th, 10th and 11th guys, um, that aren't getting much significant play time because, you know, the, the bulk of starters have say, all the starters have props from those part.

Jordan: Gotcha.

Eric: Yeah, and just kind of to add to that, like, I, I was looking through props, um, this morning and like, it's. Like eight deep on every team so far in terms of like points, rebounds, assists. Um, so, you know, I think a lot of what Will is talking about, we'll see in the late news where all of a sudden, like, you know, 20 minutes before lock, someone gets ruled out and then.

The books are, like, scared maybe to put up some of these, like, uh, I guess, like, lesser known players or, you know, props in the situations that haven't occurred before. Uh, but I think the priors do a pretty good job at, uh, You know, sort of picking up that slack and, uh, yeah, I think

Wil: they'll always keep us reasonable for those things.

And I think that's really with, like, NBA and breaking news. It's obviously there's a ton of value. You can add if you do no ball in those situations. Um, and I think the best thing that we can aim for is making sure that those. Uh, weird cases are reasonable, you know, we have a strong baseline. Okay. Like this, this, you know, spot starters up to 25 minutes now from his usual 5 to 10.

um, and, you know, accordingly, like, you know, this is what his stats look like. Um, so I think that's, uh, I think a pretty good strong point, but definitely, you know. We don't have the correct answer for, uh, what usage is going to look like in those weird scenarios.

Jordan: I would imagine too, and maybe I'm oversimplifying it, but I, I figure you could probably do more with less in NBA.

Like in, in NFL, there are some of those kind of more niche props for lack of a better word are probably pretty useful, like longest reception kind of stuff, just because can happen on a given play is. So diverse, but the, what can really happen on a given possession of NBA, there's, there's really only a handful of things.

And like, you're going to start to have a pretty good idea of what's going on with PRA points, rebounds, props, um, for, for the top eight players. I mean, that's, that's like a pretty solid image of, of what at least the market thinks are the means for each player's performance.

Wil: Right. And I think, I mean, that's sort of what we see with just NBA being more.

Projectable in a sense, like, then other sports, because it is, it's that nice simplified outcomes. Um. And it's, you know, MLB is similar, but you know, you have, you have potentially like unlimited extra innings, whereas like NBA, it's just super defined. Um, so I think that, that obviously makes our job a lot easier, uh, with keeping everything good.

Jordan: I did want to ask a little bit more about the minutes model, um, because that is internal. So what are some of the, the things that we're really looking for there? I think two things that really jumped out to me, obviously is the, the injuries. Like when we have players that get ruled out, how we're going to navigate that.

Especially for situations where it is a little tougher, like early in the season, like the first time a big player sits this year, we might know how it looked last year. We're not going to know what it looks like this year necessarily, but also I'm curious if there's anything else that's going into that, like if we're doing anything, you know, matchup related or, um, just like the variety of different things we might look at for figuring out what a rotation is going to look like.

Eric: Yeah. So matchup related at this point, we're not doing, um, or, you know, we're not accounting for like, Oh, Minnesota again, they like play bigger. So maybe this team will also play bigger. Uh, it's something like that. Um, a lot of the like purely automated side of it is just what we've seen in the past. Team.

Um, and so of course, like at the start of the year, we're going to be a little bit less confident, uh, in some of, you know, those projections, I guess. Uh, with that said, like we are, we sort of have like that automated side. And then we're also going to be obviously taking a look at. The minutes projections, um, you know, trying to best apply, you know, places where we see that there's just a blind spot in terms of, I don't know, a player or something that like a coach said, like the automated model can't pick up that, like, Oh, uh, You know, J.

J. Reddick said that Max Christie was going to play more in this next game or something like that. Uh, and so for situations like that, we will be going in and making adjustments, uh, for things like that. Um, but I mean, really the, the advantage of the automated model is in the scenarios where someone gets ruled out close to lock, uh, where we can be like pretty confident we're on the right track.

Right. And, you know, that we're getting like the right starters. And, um, you know, if like a, if a wing is out, then obviously we're putting a wing into the starting lineup and. They like bench wings, they're going to be the most impacted by that. So, you know, stuff like that, where we're getting close to the right answer, uh, very quickly, which I think is like, that's sort of paramount in NBA, just with how, how the news works.

Jordan: Gotcha. What are some of the things that we found that we were getting wrong often last year that you guys are the most excited to see how we improve upon for this year?

Wil: I'm not sure if it is specifically last year, because I feel like we did improve with it last year, but just in general, like, we're usually under on star players, but we've just had that problem because it's hard, um, you know, with the, the level of, like, regression and stuff that we do, we usually have a trouble, like, er, we usually have a problem getting people to high enough usage that, like, a Luka is.

And I think that this approach is going to be a lot, this approach isn't going to have that issue because it's just going to calculate it from the props and say, okay, yeah, it's, you know, 37 percent or whatever. And so I think that's a pretty exciting. Improvement, at least for me. Um, I don't know if Eric, you have like a different one.

Eric: Um, it's, it's like related. I think a lot of times when there were spot starters, we would see their rates in times where they played off the bench or, you know, played in more of like a bench heavy lineup. Uh, and then they would go into the starting lineup and we just kind of would assume that they would keep, um, you know, their like rebounding rates or their usage rates.

And I think. Anecdotally, and this is like in the data as well, but, uh, guys like that tend to, uh, like shrink almost when they get into the starting lineup because the other starters are like, they're the usual starters and, you know, this guy doesn't just want to, you know, Take random shots away from better players.

Uh, so I think that sort of was a place where we could improve from last year. Uh, and I think with props, especially, like, we will be accounting for that. Um, definitely, I think. There's like a piece where if the news is like, too, too late, we might still kind of have that issue. Um, but again, I think it's just more something to be aware of that, like, if a bench guy goes into the starting lineup, he might not necessarily like, hit his season averages, uh, which, you know, I think you can just adjust if you think that it's wrong, I guess.

Wil: Yeah, I'd definitely say that. It's going to have a similar problem if we're missing props for that player. I think it's mitigated, I think it's less so than it's been in previous years. I think we're going to be more closer to a midline for them, but that's definitely something to be aware of.

Eric: Yeah. And, like, just to add to that, too, we, like, we took steps last year to try and mitigate that and try to get, um, you know, essentially, like, giving usage to.

The Lucas and like the Kyrie's sort of like a head of these other guys. Uh, again, I, I don't know that we got a hundred percent of the way there, but, um, you know, it's like, it's not like when we put this scrub into the starting lineup, like, He's getting his lead average. Like, we are taking steps, uh, but sometimes I think it's just not quite quite enough.

Um, and with no props for those players, like, we'll still kind of be at that. Maybe not quite enough

Jordan: level. Yeah, I mean, I definitely noticed, like, especially with some of the star player stuff. I know there were, like, a few players that we typically had biases on, um, that the other thing that came to mind to me is, like, I think relying on Vegas probably does help us pick up on even a little bit more of some of the matchup y kind of stuff.

Like it just, it's going to, there's like all these little micro things that might affect the way a player could be projected for that given game, even if it's just by like a point or so. And pulling in that data, I think is just going to do a much better job of picking up on as much as possible, rather than it feeling like we have to account for all of these things.

Little things that are that are coming in. So, um, the 1 of the thing that came to my 2 is rookies. I imagine, like, with I was just going to say less data. Yeah. Do you want to just speak on that a little bit?

Eric: Yeah, just like, um, you know, I remember. So last year at the start of the year, uh, I don't think we really had any, like, props in place.

And, you know, midway through the year, we sort of started applying this approach or, you know, something close to this approach. But the start of the year, uh, when we were like, sort of all over the map on him. Yeah. And, uh, I think. Like with props, we're just going to better like know kind of where he's going to be at in terms of like a fantasy basketball player, uh, which is like a hard problem because.

You know, you get these guys in like college basketball, if they have data that are like the like alpha basically of their college basketball team and then they're hopping into the NBA and it's like, well, this guy went from 30 percent usage to like 10. Yeah. So it's. Like, it's just, that's difficult without manually going through and like, analyzing every rookie situation.

Uh, but that's what people are doing when they're betting props.

Uh,

Eric: and so we can take advantage of that. Um, and yeah, just basically get a better sense of how these rookies are going to play as. Like, a fancy basketball player.

Wil: Yeah, I also want to add, I think that we have better priors for rookies this year.

Um, just with the way that we set up the prop script to fall back to sort of the permanent and per possession priors. Yeah, I think that, uh, we've, we've always solved, you know, for priors based on like, you know, okay, they're a 1st round draft pick and they're 18 years old or, you know, using some of those things that we could use to interpret interpret priors last year.

Um, but I think that this year will be, you know, obviously it's not perfect because we don't know a lot of what, like, the usage and stuff looks like, but I think that it will be better than what our baseline has been in previous years.

Jordan: And I think one of the most valuable parts about this, and I mentioned it in the NFL video we did as well, is, like, the, the, especially the closing props are going to be pretty good.

Not to say they can't be beaten, but they're, they're going to be good. And it, it lets people have a really strong understanding of where we're coming from. It makes the projections really easy to understand. Like, why is this guy projected for this? Well, like that is the combination of how many minutes we think he's going to play.

And essentially the markets take on what his usage and rates look like in the game. And if you disagree, it's just becomes very easy to say like, no, I think people are wrong on this and take that stand rather than it kind of being black boxed and being this, like, this is what our, You know, hidden thing believes about this player, like, what we've decided this rookie who hasn't played a game yet looks like on the court.

Um, I think 1 thing that would be kind of interesting to go through is basically to just step through, like, I guess, like, a 12 hour ish period from when the projections would show up to lock to post lock news. So we can, like, Kind of explain what people are looking at at different parts of the day and how certain pieces of news start to be incorporated.

So I guess, first of all, I know last year, a lot of times we would have this very initial projection run that often popped up like the night before or sometimes overnight. Is that still the plan? Like, our first kind of wave of projections will show up. Then, or like, when, when's the first time somebody would actually see projections in the app?

Wil: I'm not sure what the, the schedule we've gone with. Uh, yeah. I have a feeling it's probably 5am Eastern. Okay. Um, I think that's what we've like sort of standardized now. Uh, I could be wrong.

Eric: Yeah, yeah, first run should be like early morning. Um, you know, on the day of the games, uh, I feel like at some point we had like the day before, but those don't really make a lot of sense.

Um, especially with like all the back to backs and everything. So. First run will be, yeah, in the morning, like

Jordan: early on the day of the games. Is that run looking at, is that basically just like almost our raw internal or are we already picking up on props? So I think it'll have

Wil: internal minutes and then any props if they're posted, but my, maybe early on in the season, there will be, my gut is like by like the first or second week, they won't be posted that early.

Yeah. So I'm guessing that's going to lean like heavily into our priors for stuff. And

Jordan: okay. Yeah. So like, as lines move throughout the day, when people are seeing Sims, like, I know even today we're recording this in the morning here, we've seen like a handful of Sims kickoff, is there a kind of a schedule to how often we're collecting and scraping props and saying like, okay, now this is what we think, or is it just lines move a little bit and we're going to readjust, like, what does that look like as kind of information is updating throughout the day?

Wil: Yeah. I mean, I, I think, uh, I don't think we've like settled on a final schedule cause it's obviously like. You know, we want to checking it we want it to be checking very frequently close to lock. Right? But we don't need it running every minute now. Uh, so I think that we're still, you know, figuring out exactly, like, what schedule we want to put it on.

Um, but I, I feel like it's going to check, you know. Periodically early in the day, and then as we get closer ramp up. And yeah, so it'll be like, as lines get posted, um, like, you know, if a player now has a prop and, uh, just doing like a little, uh, validation check. So it's like, you know, okay, do we need to run a sim?

Or is this like, basically the same? Like, if somebody goes from like, um, their line is like, Over under 25. 5, and it's like, minus 1, 10, minus 1, 10, and it moves to, like, minus 1, 15, minus 1, 0, 5. That's probably not enough to run a SIM, like, on its own. So we'll be checking for, like, how much it should move and use that judgment to kick off a SIM or not.

Eric: Yeah, and like I think, I think about an hour to kind of the first game of a slate,

like

Eric: we're pretty, I don't want to say like slow, but we just don't have to be as like meticulous around incorporating, uh, like prop movement like that, or, you know, even like a player being out. We might wait 15 minutes to run a sim if the game is in seven hours or whatever.

Um, but once we kind of get into that, like, like crunch time window, uh, Then like we, we sort of will have like incorporated all the props. They'll be all up to date to that point. Um, and from then on, it's more so just like the player news. Um, as opposed to like, Oh, this prop is just moving a ton. Well, obviously still run a SIM if that happens, but I think a lot of that prop movement happens before that.

Uh, unless there's news obviously, which, um, yeah, we're going to pick up on and run a SIM.

Jordan: Yeah. So let's talk about the new stuff a little bit. I guess first we get injury reports in the afternoon. Sometimes we'll have that the day before for, for some games, but, um, the main injury reports for that night are coming out in the afternoon.

Are we. Above and beyond like what Vegas is kind of doing. Are we doing anything that hedges questionable players at all? Or if, if that player is in the projections and in the sim, it's Essentially, assuming they're they're playing.

Wil: Yeah, I mean, I'm assuming because I can't think of a time that, like, Vegas would like their median of if they're playing isn't going to shift.

So I don't think Vegas is like, having the line if they're 5050 to play. Yeah. So I think that if they're in the same, we're assuming they're playing. Maybe not like normal, like, we could be like, the prop line could be accounting for, you know, them playing coach that they're going to play 20 minutes or something.

Um, but it's definitely going to assume that they're playing and not that there's like, sometimes that they're not.

Jordan: Right. Yeah. I was just thinking like, I don't know, maybe they, they have like a limited Potential or like less effective in their kind of the way that the market's reacting, but that might not be a real factor where you're going to say, yeah, I think.

Eric: Yeah, just, I think, like, more likely kind of, like, to what will said if someone is questionable like that, obviously, it depends, like, if they're. normally a starter or not, but like for the starters, I guess, like if a book hangs a line on LeBron James points and he's questionable, he's for the most part, he's either going to play and be like full go, or he's not going to play and the prop will just be void.

So like, I don't know. I'll like have to kind of see what they're doing with questionable players. Um, my thought would be there, you know, especially for like big questionables like that, they're just not going to hang lines. Probably. Yeah. And like, you know, even for the other players on the team. Right.

That's the big one. Obviously, that's the impact. Uh, and so for that, we're like, we'll mostly just be relying on our internal stuff.

Mm hmm.

Eric: Until there's sort of some kind of news on if he's in or out, um, and then props are posted. We'll pull those

Jordan: in. Uh, but yeah, so, yeah, let's, let's talk about that. Then.

So, using the LeBron example, let's say we're heading into the slate. He's questionable and we get the news. We get to confirm starting lineup, um, and assume there's like. Let's say there's 30 minutes before lock, like, it's not like fire drill type situation. Um, what will people kind of see in the sim from there?

Like, I would, I would imagine there's kind of first, like, maybe some, like, LeBron needs to get out of the sim, sim that runs. And then presumably as we start to get more information about what the market thinks about this game now, that's updating. But like, what does that kind of look like when the news first hits?

Wil: Yes, I mean, like, we definitely don't want, you know, we've had like a problem with like a yo yo Sims before, you know, where it's like, okay, he's out. And then this changes and this changes and it moves back and forth. So we're definitely trying to limit the impact of that this year. Um, I'd say, yeah, you definitely like, as soon as we get the news, I believe we're just going to run a SIM, you know, we'll fall back, um, completely internal stuff if the props are pulled.

Um, but just, you know. Run a sim so that LeBron is out because it could be a fire drill, you know, like, right. Right. Yeah. So it's, it's, we want to just get that. So we have the most accurate information. Um, and then I think that we'll probably wait to see what if props get reposted, if, you know, these things update, et cetera.

Um, and then that will, you know, kick off the next sim to make sure that it's, you know, a better baseline. But I think we definitely don't want to get into a yo yo type situation.

Jordan: Right. That makes sense. Um, Eric, you would have mentioned to like, especially when we were talking about matchups before and just some of the other things that maybe the internal won't pick up on perfectly a manual review kind of process.

Um, what, what are you looking at there and what are the kinds of things that maybe would would catch your eye of like, I'm going to keep an eye on this situation. Um, what are the, what does that process look like?

Eric: Yeah. I mean, I think a lot of it is just kind of, uh, like trying to be quicker than the like automated model, uh, which, you know, is going to be important at the start of the year here.

Uh, because we don't, you know, the, it'd be a lot more confident in the last 10 games when it's like within season as opposed to, uh, the new season here, um, but yeah, there's just like, I think. There's just going to be situations where, uh, the automated model isn't going to be able to look back at like, Oh, uh, you know, LeBron and AD were out this game and like this guy played a lot in LeBron spot, but for this game, it's like only LeBron is out.

Um, it, like it, it just can't, there's like nuance that it misses that I think, like, We can pick up on and sort of like recognize that pattern, uh, which, you know, it might only be like a two game sample or something like that. Uh, so again, it's hard for something automated to like pick up on that. And be confident that like the two game sample is correct.

Uh, but yeah, it'll, it'll mostly just be looking for like patterns that the automated model is missing. Um, and then picking up on, you know, just like quicker, like, Oh, the rotation change, this guy's now just like completely out,

you know,

Eric: he hasn't played the last two games, um, and this other guy has stepped up in his spot.

We likely will be a little bit. More conservative with like, well, you know, maybe this guy gets 5 minutes and this guy gets 15 as opposed to it might just be like the second guy should just get 20. Yeah, so just, you know, situations like that, where there's kind of been like a material change in how a player is being used, um, that's kind of going to be the stuff that we're looking for in the manual review.

Jordan: And it sounds like a little more focused on tweaks to the minutes. Rather than underlying rates. That's like kind of probably mostly what we're looking into there.

Eric: Yeah, for sure. Um, again, it's just like, that's That's really the easiest, not the easiest, but like, like, we want to make sure that the right players are on the court for the right amount of time.

Jordan: Yeah.

Eric: Um, we're like pretty confident in the priors and in the props and everything. Um, but minutes, it's like if you're five minutes off on a player, that can be five fantasy points, which that can be the difference between 100 percent of a guy and like 20%. So, um, just, yeah, it's important to like get that right and dial that in, um, as opposed to like the rates, which there just isn't that, like, we won't be that far off on the rates.

To, you know, have a player 5. 2 high, um, as opposed to minutes. It's kind of like, right. I think

Wil: minutes is like. Definitely the most impactful lever, um, for DFS lineups, which is what we care about. So,

Jordan: yeah, that makes sense. How do we kind of. Incorporate data as the season gets going. Um, like right now, obviously we have nothing, I assume, like some of the first games that get played are pretty valuable in terms of like, oh, this team's actually going to do this.

But also it's only, you know, been a handful of games early in the season. Like, what does the process look like for the way we incorporate that what's coming in for the, for the 2024 season?

Wil: So from a prior's perspective, I'm going to tell you something I've never said before. All right. Decay rates and my initial regression values.

Okay. It's the same stuff that, you know, I've worked on for almost every model and for a while now. Yeah. Um, that's, you know, updating our priors, um, as the season goes along. So as we get more data, uh, you know, we're looking at, okay, we're going to decay at this amount. We're going to, you know, regress at this amount.

Um, and that's going to incorporate what our priors are. And so those will update after every game for every player. Um, and so, you know, as the season goes along, they'll get more accurate, um, with their role, their playing time, et cetera. Um, and then like, Eric can talk more about the minutes as well, because I think that is also incorporating some of that.

Eric: Yeah, so, I mean, the minutes, it's, it is just a mix of, like, what a player did last year, and then, you know, obviously, we're just giving more weight to, like, the current season. Um, so, like, the first two games. We'll wipe out a bunch of data in like February from last year, essentially. Yeah.

Jordan: That's cool.

Eric: Uh, that all of a sudden like, it's like, well, you know, we have two games with like this new head coach or whatever it is.

Uh, we probably don't need to be looking back at February from last year. Yeah. Uh, so it. I mean, it's essentially just like a, like back tested, uh, you know, model on how much to wait, like the current season, the last season, you know, how far away a game was, um, and that's kind of informing our initial minutes.

Jordan: Yeah.

Eric: For a player.

Jordan: Is that consistent across like all players and teams, or do we like the, the Timberwolves make a big trade? Are we like a little bit less willing to trust how they looked with Towns on the roster last year because he's now not on the roster? Or do we kind of just assume that last year is, is as valuable for everybody until we start getting new information for this season?

Eric: Yeah, so, like, that's another thing that we're kind of trying to catch in, like, the manual review to start. Um, because, I mean, I think, at least, like, Randall with Gobert and then Nazreed off the bench is different than, like, Townes, Gobert, and then Nazreed. Um, and so, yeah, again, that's something that we'll catch in the manual review of, like, oh, Randall played 33 minutes as a starter, like, with the Knicks, um, does that drop a little bit?

Like, what does that kind of look like with Minnesota? Uh, and that's something that the, It's just, it's, I think it's, it is like a case by case basis of how that's going to impact it. And so, uh, we've just kind of left that to, to the manual review portion.

Jordan: Yeah. And I, I realized I asked that question probably in a confusing way.

We're obviously aware of who is on the teams and assigning minutes accordingly and the usage accordingly. I like just think like the idea of like, I mostly was asking just like about like decay rates, um, as kind of like how that information changes or how valuable we think. Past season information is, but, um, that's kind of, yeah, I think anyway.

Eric: Yeah, I mean, I think ideally you kind of would have like almost like fast regress teams where yeah, you know, you like, um, I don't know, the Lakers like didn't change that much last year this year in terms of their starters. And so, you know, they, we might be more reliant. It is kind of just a one size fits all.

And then, uh, our like fast regress is just the, the, the review that we're doing and the adjustments that we're making manually.

Jordan: Um, any changes to ownership and field lineups and things like that for this year, or how are you guys feeling about that side of things overall?

Eric: Yeah, so no, like, official changes, I think, uh, with the prop stuff, kind of, what we've seen when we've applied it to other sports is Our ownership has gotten better when like the props are parts of the projections.

And then the projections are going into the ownership model. That's like building the field lineups. So I guess from that perspective, like, I think our ownership and our field lineups will be better. Um, the process itself hasn't changed. Uh, maybe it's a little bit confusing, but yeah. Um, nothing like too, too big.

In terms of ownership. Cool.

Jordan: Um, overall, I, I, I am pretty excited about the season. I think, like, even just the, the opening night projections, like, definitely look pretty good. Obviously two games, um, early on, it's like, I think pretty easy to do a pretty good job at it tomorrow. Yeah. Yes. Tomorrow will be a fun one.

I am curious, like, just. Yeah. How are you guys are feeling about where the model is at overall? I, it sounds like you've put a ton of work into it here this, this fall and summer. Um, I'm pretty excited. Things like look good. I'm really excited to, to just like see how the season starts, but like, how are you feeling heading into the season?

Wil: I feel great. I, I. You know, I'm excited. It's obviously like I talked about on the NFL, you know, it's hard to say goodbye to, like, some of the stuff that we were doing with the old model that, you know, but we also put a lot of effort into. But, uh, this approach just is so I feel like safe is the right word.

Like, it's going to get really good baselines. Um, and I think that that's really valuable to have, especially in, like, time crunch scenarios. Like, you know, I am obviously stressed about tomorrow with the massive slate and all that. Yeah. Um, but less stress to than I was last year with, you know, in terms of, you know, points of failure, I feel like it's, it's way reduced and I feel like we're a lot better handle or a lot better prepared to handle stuff.

Um, so I feel like, very optimistic and really good about this season. So I'm, I'm excited.

Eric: Yeah, like, I'm the same, I think. Uh, again, we sort of like started this approach kind of midway through last season, uh, and then really refined it through MLB, um, through NHL and then obviously through NFL and like all of those sports, I've been like extremely happy with how projections have worked, how they've backtested.

Um, I think that those sports are in like a great spot and we're taking what we've learned from that. We're applying it to the NBA. Uh, and so I, I feel really good about it. Obviously there's still, you know, like NBA is still a late news sport. And like, that's the important part. But, uh, you know, I think looking at how we've applied this to other sports.

How projections have worked, how people have done, um, I'm, I'm really happy with, with how it's turned out.

Jordan: Yeah. One, one thing too, that, um, I had heard mentioned by Jay fresh who won the, the Millie with the saber logo for the bills and dolphins game, slightly unrelated to NBA, but I promise you it's related.

Um, he had mentioned that part of what was very interesting to him about us doing the Vegas model update for NFL was that he felt like it made. Post contest flashback Sims way more valuable because he had a much better sense of like what we thought about the games, knowing that they kind of matched up to what the market thought about the games and put a little more stock into those numbers of like, yes, you're profitable here because it's not like did Sabres and black box.

Think you were profitable, but rather like, did the market think you were profitable? And I think there's some of that to be something to be said there for NBA to of like, you know, this is basically kind of like how the closing line market. Yeah. Thought about what was going to happen and that passed through a contest.

So

Wil: definitely, I do think that those two points that make it like a little bit less impactful for MBA, just with late swap. Um, obviously like when you're doing post contest flashback and you know, the, it depends, like, did you late swap to this person type of thing? Um,

Jordan: yeah.

Wil: And so I think that that knocks it a bit.

And then as well as, um, just like projections are. Like, you know, predictions are not going to be significantly different this year than last year. Um, just because projections and MBA are pretty similar, like solved. Um, and, you know, I don't think there are too many times where we were actually, like, pretty different from any other market sources.

This just makes sure that we're, you know, we're consistently at the market average, but, um, I think that it's less so than like NFL, where we definitely had more outliers that got, uh, regressed to the market

Jordan: for sure. Yeah, no, that makes sense. I think either way, though, it's easier for us to kind of basically say, this is a Vegas model than it is to say, like.

It's not right. Promise us. It's going to be close. Um, you know, for sure. The 1, the 1 other thing I wanted to ask you. Well, you had mentioned, um, I, I, I feel like the internal model. Kind of based on our priors probably still ends up getting. Some time to shine this year in the situations where we just don't get props back from breaking news.

Did I understand that correctly? Like, if we, if Janice gets like scratched, we were not expecting it. All the books weren't expecting it. Like he's gone. The props just never come back. We're going to kind of rely on the internal model. Plus the manual review stuff that Eric was talking about to get it done for that game.

Wil: Right. Yeah, and I think that so, yeah, that's definitely going to use, like, for the, the rates themselves, it's going to use our priors. Um, but it is a bit different than last year in terms of, like, the minutes, because of Eric's work on the minutes model, it's going to handle, like, that minutes getting redistributed.

Um, a lot better, which is obviously, I think that's the more impactful side of, um, especially when it comes to you guys, it's just more important to get those minutes right than it is to get the rates. Right.

Jordan: Right. Were you going to say something, Eric? Oh, well, okay. Okay. I saw your mic on mute. And I was like, he's cooking something up over there.

We, I think we, um, 1 of the last questions I wanted to ask here, and I think we've leaked some of these. Throughout the video here and there, but for people that want to spend a little bit of extra time. Diving into certain situations, maybe tweak a couple of projections, do a little bit of research. Where are the kinds of spots that you think that time is best spent?

If somebody has maybe a half hour or so before lock.

Wil: Yeah. So I think, uh, you know, Matt's touched on this in, in all of these videos. I was telling, which is if you're going to bet a prop, then it makes sense to boost the player. Um, like if you just think, you know, our, um, like our lines are, our player lines are the market lines.

So if you disagree with the market. Then change that I think that's, you know, that touches on what you said about, like, it not being a black box or that just being Vegas, you know, it's like, okay, I, you know, I disagree with Vegas. So I'm gonna do this. Um, and I think the other 1 is like, we've touched on, which is just in weird scenarios.

We're going to be a pretty like, decent baseline. But, you know, if you like, have seen beat reporters talking about it, or, or, you know, especially if it's something where props haven't been posted, uh, definitely add value there. Where do you think you have it?

Eric: Yeah, I just, uh, I'll echo like the, the uncertain scenario type situations.

I think those are really the places that you can add value and you can, you know, do research on maybe it's like a, when this guy goes into the starting lineup, where does the usage go or like something like that? Um, uh, yeah, I would say probably that, um,

Jordan: Um, I know NBA kind of has the. Luck of being like the last major sport to kick off at least up until like baseball in the spring. So it always feels like we have a little bit more time to continue to work on it throughout the season. Whereas like something like NFL, a lot of times it's like, well, what's next for NFL?

Well, NBA is next for NFL. So with that said, like, is there anything. That we haven't been able to do yet that you guys are planning on digging in a little bit more here as the season develops, or even just things you're already thinking about for 2025 of like, oh, you know, looking into that a little bit more, maybe trying to, to build out this might be interesting.

Wil: Yeah, I think for me, um, you know, 1 project, I'm not sure if this will get, you know, tabled for other priorities or not, but, um, is working on, like, the way that we, uh, hit the prop targets in the sim. Yeah, I talked about, like, that, the converting of the rates and stuff like that. Um, it definitely, like, I feel good about it.

It's in a good spot, but, um, there's still, like, more tweaks that I want to make. And, you know, if, if we're going to spend that time on NBA, I'll probably go in there and and do that. And the goal of that is just to make the, um, like. Get those rates quicker and get them to the right spot faster. Um, and so I think that's my, that's my main outstanding thing that I want to do.

Um, but like, all in all, I feel like it's one of like the most complete models that we have. Like, it's awesome. Really, really good feeling to have. So yeah.

Jordan: Is that I just want to ask a quick follow up question there. So when you say that, like, do you mean if we're like a guy supposed to have 9. 5 rebounds or something on average that we're like, getting to that number with like, faster?

I was just like a little confused on kind of what that means

Wil: in the way that, you know, like, when the Sims running it and taking in those input rates, it's just about, um, you know, making sure. So, like, you know, we've set it to he's going to get. 25 percent of rebounds or something like that. Um, and then say, we're observing in the sim and he's getting 7 rebounds and his targets 9.

there's, you know, adjustments that we need to make in the same in order to like, reflect that there's something that's not, uh, working between the input rates and the outputs. Um, and so those are, that's, there's, you know, just some inefficiency there and I think we've done a good job to make it get there.

Um. And like, you know, it's, it's obviously like, it's, it's never going to be, I think, 7 and 9, like, something that dramatic, but, um, there's still just like, as a perfectionist, I want to like, I'd like to get it perfect, like, right from the start and figure out, you know, exactly how to, so that way we can just get the full Sims there.

Jordan: Gotcha. Eric, did you, uh, have anything that is on your NBA to do list?

Eric: Um, I think, like, so with the automated minutes, um, we, in theory, we can have, like, scenarios of this player is questionable. Uh, so, like, what do the minutes look like when this player is out? Uh, which I think would be really beneficial for, uh, When there's news in the late games and you're like, I kind of want to keep some swap options open.

Um, but as of right now, we don't have a way to sort of have like, Oh, in 50 percent of the Sims, you know, LeBron is out and the minutes are this. And then the other 50 percent LeBron is in a minutes or this, we don't have a way to do that right now. Um, we're like, I think we're getting there, uh, with the automated minutes and some other improvements.

Um, so I think like that would be kind of my main, I think it would be the biggest, uh, like move to the user is just like having, you know, not having to do like to consider all of that by yourself, whether you're using a group or, you know, just like boosting a 4, 000 player a bunch because like, Oh, he might be.

The best value of the slate, like when it hits 8 p. m. Um, so I think that would be like, super valuable.

Jordan: Uh, yeah, that sounds awesome. Yeah. I mean, both of those things, like, I mean, we'll what you mentioned, like, if our Sims are picking up. If I understood what you were saying correctly, like we get to the, what the correct output on the Sims faster, which would like literally make us quicker to more accurate projections when news breaks and then Eric, like potentially having a stronger, you would have set your portfolio of lineups up for more success to capitalize on news if it were to break the right way with that.

That seems. Like a very good one, two punch there of like the biggest edge in, in basketball. Awesome. Yeah, no, it'd be pretty sweet. Cool. No promises there, but looking forward to it. If we, uh, if we have the time. So, um, cool. Any other final thoughts? Anything we didn't get to, um, anything that you guys had wanted to mention that, um, we didn't get to talk about?

Eric: I don't think so. I think like the only, and this is more of like an MBA. Thing I guess uh is just like the the way that the game was officiated last year really changed at one point Yeah, and I think it'll be interesting to see Which nba we get early season. Uh, just because like I don't know, the ref started letting people play defense and obviously that like has an impact on DFS and, you know, stuff like game stacks and things like that.

Um, that's more of, yeah, again, like, uh, actual NBA thing that I'm interested in, in, uh, seeing what happens, but, uh, yeah,

Wil: my, my stance was much more, uh, much less serious. Uh, I would say boost all the Lakers by 15 points because the Lakers are going undefeated this year.

Jordan: Wow. Are you a Lakers fan? I had, I didn't even realize

Wil: I'm from

Jordan: LA.

Wanted to hear that. I am curious, uh, real quick, Eric, if you have like a take on that officiating thing, based on what you've seen in some of the opening numbers, like, do you think the, I guess a better question, do you think the market has a take? On that or nobody knows.

Eric: Yeah. It looks like the market is assuming that the end of season stays like that is the new NBA, as opposed to some of the like two 40 totals or like two 45 totals that we're getting early last year.

Um, it seems like that's. Like, I think the market expectation is kind of like tempered from that, but it's definitely, it's definitely incorporating a lot more of like the late season reffing as opposed to, uh, You know, assuming that like, Oh, the start of the season is just going to be like 300 points in every game.

Like it was last year. Like, so, um, yeah, I don't know. I think it's interesting to find like shifts like that. And, uh, yeah.

Jordan: Yeah. And it's like a good example too, of like another, I guess it's kind of a macro trend. We, in theory will be faster to like, not to say that's the kind of thing that we just like would have been completely unaware of last year.

Like, I'm sure we had ways of thinking about that. But. There's just like this nice backstop to that of like all of those kinds of things that that can add up over the course of the season, we're just going to be a little bit better on. I think I'm just picking up

Wil: definitely.

Jordan: Awesome. Cool. Um, last note from me, I will mention that for opening night here, we reset everyone's five day trials on our site, sabersim.

com. So if you've been hanging out with us here, uh, interested in checking out the NBA Sims, if you've tried it before, want to come back, give it another shot. That trial is also for all sports. So if you time it right, you get NFL Sunday in there, get Thursday night football in there. Everyone's trials will reset.

So you can come check it out for free for five days. Well, Eric, thanks for hanging on and. Talking with me here for the past hour or so. Um, and I'm super excited for this season. Yeah, me too. All right, cool. Thanks.

‍