Some thoughts:
First, you absolutely should actively manage your lineups to play the more favorable matchups. But I would suggest possibly tweaking how you are assessing which matchups are most favorable. But it’s most important to hit the GS and IP caps, so be careful about choosing when to bench players who are in their teams’ starting lineups. Once you get off pace to hit those caps, it can be difficult to catch up after a certain point. So be very selective about when you bench a player who could start for you.
With that said, I honestly don’t think that L15 or whatever is any sort of meaningful measure if you’re trying to predict short-term future performance. To put it in statistical terms, reversion to the mean dominates recency bias both empirically and theoretically. For example, if you have a hitter with a .350 wOBA ROS projection who has a .200 wOBA L15, then his expected production for next week is .350 not .200 (i.e., you don’t expect the “slump” to continue). Unless there’s an injury or some other reason to expect performance to decline, I just treat L15 fluctuations as statistical noise.
I often refer to this article in terms of how large a sample needs to be before different baseball statistics “stabilize.” For example, K% stabilizes relatively quickly for hitters (60 PA) whereas BABIP requires a much larger sample (820 BIP–i.e., 2+ seasons worth of data). If a hitter averages 4.25 PA/G and plays 7 games a week, then he’s averaging a little less than 30 PA/week, which is below those thresholds.
Because of all that, with respect to L/R splits, I look at career wOBA splits, not 2021 wOBA splits because the former are far more reliable than the latter. For example, IIRC Hosmer has great wOBA vs LHP in 2021 whereas he’s been less effective against them over the course of his career. Whenever I can, I still bench him against LHP because I’m basing the lineup decision on what I expect him to do in upcoming games not what he’s recently done (and the latter is not predictive the former).
I don’t think ESPN’s pitching forecaster is all that helpful mostly because it seems to be heavily subjective by staff who aren’t that knowledgeable (full disclosure: I think all of ESPN fantasy products are complete crap based on the last time I sampled them, which would have been 2018). However, Pitcher List does a weekly 0-10 pitcher matchup rating, which is a similar idea and better executed. I sometimes will use that to decide when to start a SP. Also, I play a lot of 5x5 where SP wins are a big priority and so one of the other things that I look at is the SaberSim likelihood of a win (available of FG player pages). You could also perhaps extrapolate SaberSim projections to projected FGPts, although that would probably be a lot of data entry on your end as the projections aren’t CSV exportable.
So tldr version of my recommendations:
- Continue to actively manage your lineups;
- Be sure to hit your GS and IP caps;
- Don’t use ESPN for any purpose;
- Use ROS projections, not L15; and
- Use career L/R splits, not YTD.