If you've played NASCAR DFS long enough, you know how to play things at Daytona and Talladega: you hunt out place-differential at all costs.

With variance maxed out and minimal laps to be led, drivers starting further back have both the best floor and the best ceiling. It's a rare combination and one we always want to exploit.

The question is how much that changes with the starting order set by an algorithm.

Once again this week, the Cup Series grid is determined by a combination of finishing position and fastest laps at Richmond plus owner points. Especially with Richmond having been a relatively clean race, this leads to a starting grid where the series' best drivers are at the front. If you want to get place-differential, you'll have to settle for lower-level drivers.

So, what's the proper balance here between getting place-differential without ignoring the drivers most likely to finish well? Luckily, we have some templates from last year's two Talladega races to lend us a guiding light.

The Back Still Has Value

The algorithm used to set this week's order is the same used for last year's playoff race in Talladega. The lone difference is that the 16 playoff drivers were locked into the top 16 spots in the starting order whereas that's not the case this weekend. As a result, we can lean heavily on that race to see what the optimal approach looks like.

For last year's race, none of the top 10 drivers in my finishing-position model were starting lower than 14th. The lowest starting spot for a driver in the top 15 of my model was William Byron in 21st. So, if you did want a top-end driver, you had to get them from the front.

This week's race is nearly identical. The top 10 drivers in my model are all starting inside the top 12 spots, and the lowest starting spot for a driver in the top 15 in my model is 21st. It's the exact same setup.

Even with that being true, three of the five drivers in the perfect FanDuel lineup started in the back half of the field, and four started 16th or lower.

We'll come back to Denny Hamlin in a bit, but even with the top guns all at the front, our tried and true strategy was still the optimal.

This is similar to what we saw in the spring race at Talladega, as well. That order was set by a combination of owner points and a draw, but the overall theme of "contenders up front, pretenders in the back" was still true. That perfect lineup peppered the middle of the starting order.

All five drivers started between 12th and 21st. So although we weren't stacking the back, there was still an emphasis on identifying place-differential options.

The big exception here has been race-winners. Hamlin won in the fall, and Ryan Blaney won in the spring. Those were the two guys who started inside the top 14 who made the perfect lineup. We should take note of those examples.

This lends itself to playing the "assumption game" for tournaments. The assumption game is when you identify a driver you think will win the race and plug them in. Even if they start on the pole -- like Hamlin did last year -- they're still very likely to make the perfect lineup thanks to the 43 points for a win. Then you lean on place-differential options for the other four slots.

The assumption game allows you to not completely ignore the drivers starting near the front. If you love Joey Logano, you can still use him; you just have to make sure you're getting proper place-differential upside in the other four slots.

When playing the assumption game, it's wise to lean on stacks. Because we see so much teamwork at the pack-racing tracks, you'll often see teams and manufacturers dominate the top of the leaderboard for individual races. If Logano wins, he probably got some help from other Team Penske drivers or Fords. You might want to bump up those options for your other four slots if you make the assumption that Logano wins.

The drivers at the front -- your assumed winners -- are not where your core should lie. If we take sportsbook odds to be gospel, then the most likely winner -- Hamlin -- gets the job done 12.5% of the time. The odds a driver starting further back winds up in the perfect lineup are higher than that. As a result, you should have the highest exposure to drivers starting in the middle of the pack or lower. They have more flexibility to not win the race and still pay off. The margin for error is much smaller for drivers like Hamlin and Logano. Your exposure levels when multi-entering for tournaments should account for that.

A helpful crutch for us is that we do have viable cars starting further back. Kaz Grala is starting 38th for Kaulig Racing, which isn't terrible equipment. Chris Buescher -- who ranks 13th in my model thanks to phis plus history at pack tracks -- is starting 21st. Bubba Walalce has always been competitive on these tracks and will start 24th. You're not using duds by looking for drivers in position to get place-differential, which should make us more comfortable when plugging them in.

As a result of the way the field is set, you're likely going to wind up leaving a bunch of salary on the table. The highest-salaried drivers are still likely to be the ones starting at the front. Thus, if we load up on those in the mid-range or lower, we're not going to use up all $50,000 in salary. Not only is that not bad, but it's actively good because it decreases the odds we duplicate someone else's lineup and have to split the pot if we wind up hitting the winning combination.

It can be unnerving to use drivers who aren't usually our key staples and leave a bunch of salary on the table. History tells us, though, that this is the optimal route for Talladega, even with the starting order set the way it is. Thus, we should stay the course and keep building around those drivers in position to get place-differential points outside of when we're picking an assumed winner.