One challenge to streaming hitters vs. streaming pitchers – or playing daily fantasy games – is that teams do not publish ‘Probable Hitters’ a few days in advance. The closest thing is Jim Leyland who publishes the positions a couple days in advance and then fills in the name the day of the game (Miguel Cabrera is getting an off day, Don Kelly you’re hitting 3rd).
So you might find the perfect hitter to stream only to find out on game day that he isn’t in the lineup (aka the dreaded ‘!’ in ESPN or ‘x’ in Yahoo. )
BTW, isn’t it odd that the site with the exclamation point in its name uses an ‘x’ and the site that promotes the X-Games uses an exclamation point? And what’s the deal with…..nope, I got nothing else here.
There is a related challenge with weekly leagues – particularly deeper leagues – when you have to choose between hitters on your team and need to account for their projected playing time in the coming week.
I have tried to address both of the above challenges via the %Start column in Hitter-Tron Daily and in the adjusted projections for Hitter-Tron Weekly. The %Start ranges from 5-95 and predicts the likelihood that a player will start in a specific game. When streaming hitters, you can either ignore or filter out any players who are unlikely to get the start.
For the Daily Hitter-tron, I still project their full stats as if there was a 100% chance they would start. That way, even if the estimate is wildly wrong (holy c**p, Jonathan Herrera is starting over Troy Tulowitzki !), you still have the correct $ value for your start/sit decision.
For the Weekly Hitter-tron, I have adjusted these totals based on the player’s likelihood to start each game. So the projected AB for Jonny Gomes will be highly dependent on the number of LHPs that the Red Sox face that week. And the back-up catcher will likely only have 1-2 Games’ worth of projections.
Below are some general strengths vs. weaknesses of the %Start estimation model. Please note that the impact of any weaknesses will be felt more on the Weekly projections since all the stats are being reduced based on the playing time estimates.
Strengths
- For players secure in their starting position, it will look at last 30 day data but will not underproject players coming back from injury (e.g., Mark Teixeira has a floor of 85% if healthy)
- For platoon/part-time hitters, it takes into account how often a player starts vs. RHP/LHP so, for this coming week, Brandon Moss is at 90% likelihood to start vs. RHPs and 18% vs. LHPs. Gomes is at 6% for RHP and 90% for LHP.
- All DL’d players are ignored by Hitter-tron.
Weaknesses
- The model does not know if a guy is ‘banged up’ or when a catcher is going to get his inevitable rest day. So a Catcher who plays 120 games a year with no real bias on LHP/RHP will just have a 75% in every game where it’s likely more like 85% and 25% on the day game after a night game.
- For players who lose their full-time role and are still on the major league team (e.g., Lyle Overbay with Teixeira back), their %Start will slowly decrease over the coming weeks vs. a one-time adjustment as his playing time is driven by ‘last 30 day’ data.
- For players who lose their full-time role and are sent down (e.g., Josh Rutledge), it takes a week to remove them from the Hitter-Tron universe (which is based on 1+ AB in the last 7 days)
- For players who have recently gained the full-time role but are not assured of keeping it full-time (e.g., Nick Franklin), his %Start will slowly increase over the next couple of weeks as he secures the starting role.
- Players coming off the DL or just called up that day will not be projected until the day after their first game back.
One last item regarding Hitter-Tron and Stream-o-nator. The current default sorting is based on $ value in descending order. So the top hitter/pitcher may be for a game several days in advance. I am considering changing the sort so that it is by Date and then $ so that the best hitter/pitcher of that day is at the top of the list. Please vote in the below poll if you have an opinion.
