Here are some things that were not always obvious to me, but that I believe improved my ability to analyze information. (thread)

1) Many standard stats can (and should) be adjusted to have more value if you have a small sample size (i.e. DVOA).

but...

2) A large sample size of unadjusted stats is always more valuable than a small sample size of adjusted stats.

It's easy to get overly confident in what adjusted stats tell you because they just "make sense". Keep your confidence level relative to the stat's predictive value.

3) Never use fantasy points to project... fantasy points... in fantasy football.

You could do with for NBA on a per minute level, but there's too much noise in football stats. The small sample size and high % of pts from TDs are the biggest issues.

Use market share and rates.

4) When looking at stuff like the % of times Lamar Jackson was on league-winning teams or how often stacking works in DFS, that information on its own has literally almost no forward-looking value. Context is critical.

Here are some made up examples...

Without context: 65% of "stack" lineups win GPPs
Standard takeaway: Stacking is +EV!
Context: 80% of lineups use stacks

Well that changes things, huh? That means that stacking is actually bad if you add just one piece of context, but you could take it even further.

Without context: Damien Williams was on 2% more league-winning teams above average
Standard takeaway: Drafting DW was +EV!
Context: Damien Williams was on 2% more Christian McCaffrey teams than league average due to him being commonly available for those owners' 2nd pick.

Oh.