Here we go streamers! Let's look back at our season's heroes.

As opposed to evaluating top holds (posted elsewhere), players who were best for streaming are not defined by consistent high scoring.

Two reasons for preferring streamers who sometimes do bad: (1) The consistently high-scoring players were likely held anyway (by your league-mate). (2) You probably dropped or avoided the player during his “bad matchup”… only to regret seeing him outperform expectations.

We don't want that with streaming. We want players who perform as predicted: they were safe to drop when predicted low, and they were beneficial to pick up when predicted high.

Therefore, for best streaming, we instead care about:

• who scored when we expected them to (predictability)

• who was not good to hold (availability)

• who was top-ranked more than just a few times (occurrence).

The top players for streaming

I evaluated top streamers like this: (1) Limit to players who were ranked as the week’s top 10 (10 is arbitrary) at least 4 times during the season (4 is arbitrary), (2) rank the players by Sharpe ratios of their fantasy scores, (3) combine this last step with a modified covariance calculation that extracts “beneficial predictability”, and (4) finally delete any remaining “best holds” from the list.

Anyway, here are the guys who most often produced for you when expected:

• D/ST:

  • Bears: 4 weeks, averaging 14

  • Chiefs: 5 weeks, averaging 13

  • Dolphins: 5 weeks, averaging 13.4

  • Eagles: 8 weeks, averaging 10.8

  • Rams: 6 weeks, averaging 9.2

  • Excluded for being best holds: Patriots and Bills

• Kicker:

  • Hopkins (with LAC): 5 weeks, averaging 10.8

  • Koo: 4 weeks, averaging 9.3

  • Crosby: 4 weeks, averaging 10.0

  • Excluded for being best holds: Folk, Gay, Tucker, Joseph, Boswell

• QB's:

  • Burrow, Wilson, and Stafford would have been better to stream than to hold.

  • Likely available streamers: Tannehill, Winston, Cousins, and Carr.

    • But these were not even that good. I want to point out that this was an especially tough year for QB streaming. (I have not previously seen such poor QB ranking accuracy by various sources.)

Biggest streaming spoilers: poor predictabilities

Some players 'ruined' streaming. They busted when they shouldn't have. Or they taunted us by booming when they shouldn't have. Maybe they were better to just hold and wait it out (rather than cycle through adding and dropping).

_{Predictability levels of 32 kickers. This year, a special peak emerged on the left, with kickers who defied expectations each week. Predictability values are taken as component correlation contributions (covariances taken with respect to league average instead of player averages).}

Anyway, here are the guys who killed not just predictive models like mine, but possibly your season as well:

D/ST: Ravens, Chargers, Steelers

• When recommended as a top option, none of these reached double digits even once: their average scores for those games were approximately: 1, 2, and 3.

Kicker: Gonzalez, Prater, Succop

• Succop must be the biggest streaming disappointment. He made the top 10 half the time, and also averaged just under 6 points in those games.

• Meanwhile Prater is the most surprising on this list. He ended the season at K12, and in fact I usually ranked him close to K12…. The biggest problem with streaming Prater was that all his impressive games came when least expected (by the model). The 4 times that he did make my top 8, he averaged just under 6 points.

QB: Matt Ryan, Daniel Jones, Bridgewater, Garoppolo

• Honorable mentions: Darnold fell fast early in the season. Wentz never boomed when hoped (and personally burned me in week 10 when I needed him).

TL;DR-- Thanks to Bears, Chiefs, Dolphins, Eagles, and (among kickers) especially Hopkins for supporting the streaming strategy this year. And thanks to the kickers who looked like streamers but turned out to be good holds: Folk, Joseph, Gay, Tucker, and Boswell. No promises, but I expect we can look forward to easier streaming next time, since the whole midseason of 2021 was particularly tough on prediction.

/Subvertadown