DT Deep Dive
***2023 Season Recap***
If you just want to dig into the data, full 2023 season (minus Week 18) and weekly results available here. If you are curious how I make all of these calls, I share details on my methodology are here. Lastly, if you are looking for an explanation of why I use the metrics I describe below, my write up of 2022 Results addresses the value of each.
At a high-level I am happy with the performance of the model this season. Compared to 2022, there were improvements in terms of accuracy against projections, performance of the solid and deeper tiers in most metrics, and my top end fell only .16 fantasy points per week behind this year's "Expert X" which I can now reveal is the #1 FantasyPros weekly ranker MasterIDP.
Still, there are some specifics that can be improved upon. Tier 1 still exceeded Tier 2 in all metrics but none were significantly different. The massive hit rate declined this year which could be attributable to better projections I am using as a benchmark or my missing some boom guys in Tier 4. Finally, while I didn't expect to surpass MasterIDP's rankings the gap between us in FantasyPros-style accuracy was larger than I hoped. The charitable interpretation here is he is just that good (rather than I did poorly)!
Tier 1 > Tier 2 > Tier 3 > Tier 4 across all metrics this season. This is the second straight year I've accomplished this feat which lends some nice support to the underlying data model.
Tier 1 still performed best but underperformed compared to 2022.
Tier 2 improved over 2022 in all metrics with large gaps to Tier 3.
Tier 3 declined from 2022 in average points, % over 10, and % at 0 but maintained an average over 10 (which is the 'startable' line in most of my leagues).
Tier 4 was much larger this year and continues to lag far behind all other tiers (as expected).
Using MFL Projections as my benchmark (and a margin of error of 1 assisted tackle) my hit+massive hit rate stood at just over 66% (59% if using strict criteria).
As a proportion of correct calls, my massive hit rate was down this year to 28% (one assist MOE) or 31% (strict).
Whiffs made up a bigger percentage of all calls this year but at 4% is still well under the rate across all DTs (10% in 2023 which is up from 8% lin 2022).
The biggest takeaways from the comparative approach (only exploring the top 12-17 rated DTs each week) are:
Predicting human performance is hard (as evidenced by the %'s correct on a week to week basis)
Whether you use me, Leo, or the FantasySharks projections on MFL, you are in good hands (some just slightly better in some areas than others)