Thanks guys, and thanks Winz for the file. I looked at defense for 2016-2019 because conveniently there were exactly 4 at the lowest price each year ($3 last year, $2 the years before) making my spreadsheet math easier. In 2015 there were 7 Ds @ $3 which makes for many more possible combinations.
So for the past 4 years, if you picked 1 of the cheapest Ds at random, you'd average 5.6 ppw. If you picked 2 at random, the average is 7.3 ppw (+1.7), picking 3 at random is 9.5 ppw (+2.2), and all 4 yields 10.6 ppw (+1.1 ppw).
Interesting that the 3rd D adds more expected return than the 2nd D does. Notably this appears to because you're raising your floor moreso than increasing your ceiling... there's always at least one terrible low-scoring cheap D (2019 CIN, 2018 OAK, 2017 CLE & SF, 2016 CLE), and if you pick a dud then you're very reliant on that second D... whereas if you pick 3 then odds are you've got at least 2 decent or strong performers to score each week (in addition your dud).
2019 Worst Duo = 7.5 ppw, Worst Trio = 9.9 ppw (+2.4 floor)
2019 Best Duo = 11.8 ppw, Best Trio = 12.4 ppw (+0.6 ceiling)
2018 Worst Duo = 6.6 ppw, Worst Trio = 8.1 ppw (+1.5 floor)
2018 Best Duo = 8.8 ppw, Best Trio = 9.5 ppw (+0.8 ceiling)
2017 Worst Duo = 5.4 ppw, Worst Trio = 8.3 ppw (+2.9 floor)
2017 Best Duo = 10.5 ppw, Best Trio = 11.0 ppw (+0.5 ceiling)
2016 Worst Duo = 5.3 ppw, Worst Trio = 6.8 ppw (+1.4 floor)
2016 Best Duo = 6.8 ppw, Best Trio = 7.9 ppw (+0.2 ceiling)
The question of 2 vs. 3 cheapie Ds then becomes if you think you can "avoid the dud(s)" with your first two Ds, in which case the marginal value of a 3rd D is reduced... and also weighing that against the opportunity cost of a few extra bucks elsewhere in your lineup.