Before the 2023 BestBall fantasy football draft season for Underdog, I set up a spreadsheet to determine if I drafted the top quarterback, with one or two other QBs. The goal was to see which combinations worked best and to see what gaps emerged between two and three quarterbacks.
2022 Underdog Fantasy BestBall Mania research - 2 QBs or 3 QBs?
I developed my spreadsheet a step further for the decision between drafting two or three TEs. I'll add that article next.
I updated this data/format to its own web page on Fantasy Football Paydirt.
Here’s what I asked Claude AI about the quarterbacks using my Excel spreadsheet:
Me: In 2023, I ran some data trying to determine if it was better to roster two or three quarterbacks in Underdogs bestball Mania contest. This contest used slightly lower fantasy points for passing yards (.04 per yard compared to .05 in the data I used). Can you put context to my research?
Here’s what I found digging into your workbook — and it actually tells a pretty clean story once you factor in Underdog’s real Bestball Mania scoring.
What your spreadsheet is actually simulating
Your “2 QBs” and “3 QBs” tabs aren’t just adding QB point totals — they’re taking the weekly max across your rostered QBs and summing that (i.e., simulating best-ball’s auto-start-your-best-scorer engine each week). That’s the right way to model this, since roster fit (Image 1: 1 active QB slot, 18 total spots) means bench QBs only matter for which one gets auto-plugged in each week.
The numbers that came out of it (Mahomes as the anchor QB1):
QuestionResultAvg gain from adding a 2nd QB (any partner)+29.2 pts/season over Mahomes soloAvg gain from adding a 3rd QB (on top of the best 2-QB pairs)+14.9 pts/season (median +10.5)Best 2nd-QB partnersJalen Hurts (+93.75), Josh Allen (+82.2), Lamar Jackson (+76.2) despite Jackson’s low season total (267.5 pts)Best 3rd-QB addsAlmost always Lamar Jackson (+50 to +65 pts), even paired with mediocre QB2s
The takeaway from your own data: the 3rd QB’s marginal value is roughly half the 2nd QB’s, and it comes almost entirely from adding a low-floor/high-ceiling rusher (Jackson) whose big weeks don’t line up with your other QBs’ big weeks — not from adding “more points on paper.”
Now, how the .04 vs .05 passing-yard scoring changes this
This is the useful correction to layer on. That 0.01/yard gap isn’t trivial — over a season it disproportionately hits pure-volume passers:
Mahomes (~5,250 yds): loses ~52 pts
Herbert (~4,739 yds): loses ~47 pts
Burrow (~4,475 yds): loses ~45 pts
Hurts (~3,701 yds): loses ~37 pts
Lamar Jackson (~2,242 yds, missed games): loses only ~22 pts
Since rushing yards/TDs are scored identically in both settings (0.1/yd, 6/TD), dual-threat/rushing-driven QBs lose proportionally less value than drop-back passers under Underdog’s actual scoring. Practically, that means:
The gap between “elite arm” QBs (Mahomes, Burrow, Herbert) and rushing QBs (Hurts, Jackson, Fields) is a bit smaller in real Bestball Mania than your spreadsheet shows.
Your core finding — that the 3rd QB’s value comes from adding a volatile rusher for de-correlation, not raw points — actually gets reinforced, not undermined, by the real scoring. Rushing floor becomes relatively more valuable, not less.
The overall size of the marginal 3rd-QB gain (that ~15 pt/season number) shrinks slightly further under .04/yd, since it’s largely made up of high-volume passers’ weeks getting capped — so if anything, the case for stopping at 2 QBs and spending that 18th roster spot on a bench RB/WR/TE gets a hair stronger, not weaker.
So: context-adjusted, your conclusion holds up — arguably more cleanly than it did under the .05/yard assumption you originally modeled with.
Me: So in the end, by drafting the top quarterback (if delivering expected repeated success), it makes more sense to roster only two QBs while adding an extra RB, WR, or TE. In addition, there is a draft cost to pairing the number-one-rated QB and drafting another early QB.
Weighed marginal QB gains against draft capital costs while acknowledging data limitations
That’s the right read, and it lines up with what the numbers show. A few things worth adding to sharpen it:
On the draft-cost point — this is actually the stronger half of the argument. Your spreadsheet measures points, but bestball drafts are a scarcity game. QB is the deepest, most redundant position in fantasy — a late-round QB (Mac Jones, Derek Carr, whoever falls to round 15+) still gets you a real chance at 15-20 pts most weeks, and your own data shows even weak Mahomes pairings (Trubisky, Darnold, Mayfield) still floor out at 482.3 — i.e., zero marginal loss, because Mahomes alone covers those weeks. RB and WR, by contrast, don’t have that redundancy; hitting a league-winning bench WR/RB in rounds 10-15 is a much bigger swing than the ~15 pts/season a 3rd QB nets you. So spending an early-to-mid pick on QB2 (let alone QB3) is capital that’s usually better spent chasing spike weeks at a scarcer position.
One caveat on the conclusion, to be honest: your data only tested this with Mahomes as the fixed QB1 — the “elite, low-variance workhorse” archetype. The logic (2 QBs > 3 QBs, prioritize a rushing/correlation-breaking QB2) likely holds broadly, but the specific marginal-gain numbers (29.2 / 14.9 pts) are Mahomes-anchored. If you ever wanted to stress-test it, re-running the same weekly-max approach anchored on a QB2-style Rank-8-12 arm (say, Herbert or Jared Goff) would tell you whether “2 QB > 3 QB” still holds when your QB1 isn’t a locked-in top-3 arm — that’s the one place your current dataset doesn’t fully answer the question.
But as stated: draft one true difference-maker at QB, pair him with a rushing/spike-week QB2 for de-correlation, and put everything else into RB/WR/TE depth. That’s a well-supported takeaway.
Me: In today’s AI realm, I could easily take this question another step further for the top combos for the top 32 quarterbacks from the previous season. It would be much harder to use for the current draft season based on projections, because we have no idea what the outcomes will be for each player.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.