Roster Construction in Ottoneu: How to Build a Winning Roster

As we move deeper into the season, teams are starting to ask themselves if it is time to buy or sell and, if so, what buying or selling looks like. How aggressively do I want to ship out potential keepers for a shot at a title? How far do I have to get in rebuilding my roster before the deadline so I am ready for 2027? In at attempt to dig into some of this, I have been looking at 2025 roster and results data. I have a long series of questions I want to look into – how much is too much to spend on top players? Stars and scrubs or balance? – and today I am going to start with this one:

If you are trying to build a strong 2027 team, is it best to do that via keepers you held through the off-season? Players you trade for during the off-season? Or the auction?

I think this is a pretty useful question to answer right now because it impacts sellers and buyers. Sellers are already thinking about 2027 directly; but buyers are trying to determine how willing they should be to break apart their core to win right now. So they are also wondering how they can/will build a 2027 roster if they are giving up keepers today.

To do this analysis, I constructed post-auction 2025 rosters by combining a snapshot of all rosters on February 3, 2025 (post-deadline, late enough that all off-season trades have processed but before the first auctions of the 2025 season) and a spreadsheet of all draft results (basically merging together this data from all leagues). There were multiple rounds of cleanup – first to merge the two data sources effectively, then to filter down to one format (FanGraphs Points) and to standard 12-team leagues. I also discovered some challenges with the data and ended up using three data sets to do my analysis and compared the results from each.

1. A set of 100 leagues that are all 12-team, FanGraphs Points leagues that had between 450 and 490 players rostered post-auction and were not new in 2025.
2. A 62-league subset of that data that filtered out any league that had teams with >40 players or >$400 in cap post-auction. Because this is all before in-season trades, this 38 leagues that were filtered out here are leagues that had one or more teams that missed the keeper deadline and made cuts in February or March, before the auction. There isn’t a great way to filter this out of the data.
3. A 35-league subset of that data that filtered out any league that had one or more team leave the auction with less than 35 players rostered. This was to try to account for league quality/completeness. Yes, there are cases where a team legitimately stops the auction with 34 players, but a number of these leagues had teams with <30 players or under $250 spent. I suspect most of these are teams that missed the auction, and that can skew results.

I used all three datasets because it was possible the impact of those two additional filters was small, and I wanted to use the more robust 102- and 64-league samples, if possible. Basically, what I was left with was something of a sensitivity analysis on sample size – the largest sample is, well, the largest, but it is also the one with the highest likelihood of weird results due to atypical leagues or teams. The smallest set is the smallest, but it is also the cleanest, and the one I am most confident represents a typical league.

With that data, the first thing I set out to answer was if better teams build more through keepers, through off-season trades, or through the auctions. It is worth keeping in mind that this is just about your 40-player, $400 roster immediately post-auction. There are in-season trades, in-season auctions, lineup management, etc. that go into winning a league. But in terms of what you can do today to prep for 2027, all you can right now is get (or keep) as many 2027 keepers as you can. So does having that core in place matter?

Each cell in that table is a correlation between the points a team scored and the percent of the teams post-auction roster that came from each source. For example, the top right cell tells us that in the 35-league sample, there was a -.105 correlation between points scored and the percent of the roster that was kept from 2024.

The thing that jumps off the table, from the start, is that these are all really low correlations! One of the things I hear regularly in the Ottoneu Slack during the off-season is that you win by having a strong core and not relying on the auction. But that doesn’t appear to be true, at least with regards to how much of your roster comes from the auction. How you build your team – by collecting keepers the year before, making off-season trades, or leaning on the auction – explains very little of the variation on how many points you score over the season.

The limited signal that might exist in this data suggests that teams that are more auction-reliant, in general, score more points. The impact, however is small. And if we look at the data from another angle, you can see how small.

The table below breaks up teams in the 35-league sample into buckets based on their final ranking – 1-3, 4-6, 7-9, and 10-12 – and then looks at how those buckets constructed their rosters. I chose the 35-league sample because the other two gave similar results and this is the cleanest data.