For the past nine years, I have been the manager for the Deadwood Ambulators fantasy baseball team. In that time, I have learned frighteningly little about how to construct a fantasy baseball team, judging by my usual place at the bottom of the standings. So this year for the draft, I decided to try a new strategy, using Claude. My rationale was that my team was at a bit of a turning point, with many of my top performers either retired or too expensive to keep, given declining performance. So I needed a pretty decent rebuild.

I tried having Claude run in the browser while I was doing the draft, but Arc didn’t like the extension, and I couldn’t get it to load. So I ended up copy-pasting the rounds and available players into Claude every round. I used the Sonnet 3.6 model and didn’t have it do online research, since that would take too long.

The experience:

Claude frequently forgot who was on my team. I fed my existing roster into it at the start of the draft, figuring this would stay in its memory. It did not. I don’t think this was to my detriment; I have no depth on the team to speak of. But after the first two picks, it was obvious it forgot I didn’t have a catcher or second baseman on the roster. So I got in the habit of including the current roster, past draft picks, and available free agents at the start of every pick.

Claude also definitely did not know which players were expected to start the season in the minors. I was all for a youth movement on my roster, so going after prospects was a fine approach. But it recommended one pick where it was clear it was going to be a stretch if they got called up in 2026, but expected them to fill a starting role in my lineup. This is a factor of it only being able to look at the projections from CBS Sports. If I had it do research on minor league fantasy prospects before the draft, I assume that would have been better factored in.

It definitely helped with strategy, balancing my approach to scoring. That’s my biggest weakness in fantasy baseball: I forget where I need to focus for scoring, so I end up with a bunch of lumbering utility players that hit 20-30 homers but don’t walk and can’t make contact. Claude made it easier to see what I was trading off, especially as I was selecting from a rather thin class of second basemen.

Other random takeaways:

It hated that I had Lars Nootbaar and Victor Scott on my roster. Like it constantly talked trash about them being dead weight. Like most LLMs, it was quick to tell me that my draft and my team were going GREAT! Based on all experience, I have a feeling I won’t improve much in the standings. I could have just told it to be more critical or to hate on my lineup, but by default, it was way too congratulatory.

Did it give me an edge?

Probably not. It helped me be more strategic in my approach to building a better balanced team. But a little bit of discipline and I probably could have gotten similar results. I liked that Claude gave me a range of options. It definitely would present one as the obvious choice, but it never got so bossy as to assume I didn’t want to know the reasonable possibilities.

If I really wanted to go all in on this, I would have fed it chats from the fantasy baseball subreddit and had it do research across a bunch of different blogs. But as a tool to help me prioritize scoring, I’m hoping it worked well.