nba dec 19
https://pastebin.com/sPUqjfYR
it depends. Some examples:
I think every day they're playing I've bet Shia Gilgeous-Alexander and Steven Adams over rebounds. Might seem weird betting two people on the same team over every day, but I feel both are really undervalued by the market so I'm okay betting them (and it's been very profitable). And some are being overrated like Harden and Westbrook I feel ok clicking under on those all the time.
Versus someone like Will Barton where he projects well but for some reason his usage rate is super inconsistent. I adjust usage rate for things like turnovers, but why some nights is he only in the low teens for usage and others mid-20's? He's getting roughly the same amount of minutes and there hasn't been any role change due to someone coming back from an injury.
I tend to never bet at GSW or Brooklyn Nets just because their teams are a disaster. Usually D'Angelo Russell and Draymond Green pop up as over bets, but either they're injured and not saying or they've both really given up for the season because they consistently underperform. Dinwiddle on Brooklyn is the only person I trust will meet his projections.
There's also injury scenarios. Yesterday based on all the news about Gordon Hayward I felt okay betting over on every Celtic, but not for Dallas spread and it worked out great. Also, I really thought Towns would sit so I felt confident on Minnesota overs and NO spread but only Wiggins did anything. If the injury could go either way and it's a key player I'll just wait until the news is announced, re-run my model for the game without the person if they're out, and then bet if there's still anything available.
Long term injuries are important too. Like Heilds role was super increased the last two weeks or so due to injuries and I felt fine about all the overs my model suggested (and he dominated).
So I have some "consistent" metric I use and someone like Will Barton I might only bet 2-3 times, versus people I think are more consistent I'll bet as many times as I can find.