General Recap of my Statistics/Tables

The people asked for it!

Hey Everyone,

After some people asking for it, I figured it would be a good idea to put together a list of common terms/stats and stuff that i use in case you are curious. I have been doing this a LONG time and have made a ton of different models over the years. I am a degen and type like a madman so sometimes i just use shorthand for player names like DAdams or Mandrews or MHJ or DMONT, those you might just have to use context clues lol, but for the other stuff you can always refer to this substack!

BOSS Models

• Old school. One of the first models i made with any kind of popularity.

• The basic idea is that using realized opportunity (targets, airyards, etc) and situation based opportunity (filed position, down, Redzone, etc) we can predict how well a player SHOULD have performed

• We can then compare this to their actual FP and see who a “BOSS” is and if we should be expecting any positive regression for said player

• I usually break this down by DK salary “High Cost Boss” (HCB > 6K) “Low Cost Boss” (LCB <= 6K) and “Keep an eye on” which is an extra set of 5 players that are salary agnostic

• I provide an Excel file of these bad boys for $1 a month on my Patreon, but will generally tell you if a player is a BOSS in my write up each week.

XTD

• play by play model created by Tan Ho (your favorite coders favorite coder)

• Goal is to simply predict the probability of Touchdown on every play

• I aggregated this into a nice little table looking at the past 3 weeks. I then compare who has the most expected TDs vs their actual. If there is a big difference, consider me interested.

Single Coverage Monsters (SCM)

• Using PFF ultimate data, we can look at which players see the biggest increase in Target Share when they end up in Single Coverage situations. (players like AJB are great examples)

• Then we can look at what defenses have been allowing the highest % of pass attempts into single coverage situations

• Lastly, I filter on the top 10 defenses that allow Single Coverage situations and sort by what players get the biggest boost from single coverage

First Read Beasts (FRB) — (very similar to SCM)

• Using PFF ultimate data, we can look at which players see the biggest increase in Target Share when they end up being the first read on a play.

• Then we can look at what defenses have been allowing the highest % of pass attempts to first reads

• Lastly, I filter on the top 10 defenses that allow the most first read pass attempts and sort by what players get the biggest boost from being the first read

Coach, I was Open (CIWO)

• I have written lots of articles on this for PFF (here is my first one), you can also search “Coach I was open” in PFF and find all of them

• Basic idea is we can predict who should have been targeted on a given play using route level PFF data. Then compare this to who actually got targeted.

Predicted WOPR (PWOPR)

• to take CIWO a step further, we can use Predicted Targets and Predicted Air Yards to create a new metric “Predicted WOPR”

• PWOPR outperforms basically every single receiving metric you have ever heard of (not kidding) for predictability and stability

BOOM Percentile

• BOOM is just an extension of PWOPR on yards

• basically its players that need more YARDS to match their PWOPR and the percentile is a representation of how bad they need those yards to match

Route Based Heroes (RBH)

• RBH is similar to BOOM but meant for Fantasy Points

• I also write an article each week for Route Based Heroes for PFF

• It is a regression to the mean model on FP given PWOPR

Quantile Regression Forest Probability Models

• These are my attempt at “creating a game environment” given data

• I use stats from the season to predict variables we care about in a QRF. I predict the 0th-100th percentiles in lots of different models.

• This probably needs an entire substack of its own to explain lol

• this basically creates a simulation of our game environment. From light testing, i think the model does a very good job at identifying edge against Vegas. Definitely not going to win 100% of the time but it can identify edges.

Thats about all of the major ideas i can think of! hopefully you find this helpful in the future!

you may have noticed one of the things my models really care about is “how well SHOULD you be performing relative to some metric” I dont focus too much on players who are performing above expectation. This is a weak spot of mine.

See you tomorrow with the next week of thoughts notes and statistics!