July was a great month for the Razzball daily projections. Based on our ongoing testing vs actual stats (aka our Ombostman), July 2015 was our most accurate month ever (tracking since May 2014) for:
- Projected starting pitcher values – both Roto $ and DFS Points
- ERA and WHIP
- Hitter Plate Appearances, At Bats, Hits, Runs (tied with June 2015), SLG %
July also represented our 2nd best performance for RBIs (with tops being last month) so I am feeling very positive about the enhancements chronicled in our July update.
I guess our competitors have been wise not to take us up on our MLB daily projections challenge. And it is only going to get harder based on our latest update….
While the July update might have been a little tough to grasp, the value of our August update should be as simple as playing baseball according to Colin Cowherd. We have incorporated two critical pieces of information that are unique to ‘same day’ projections (and many of you have been clamoring for):
Lineups – Thanks to our friends at LineupLab, we now have a direct feed to daily lineups. We will update the projections a couple times a day until Roto/DFS lineups finalize (usually around 7PM EST). Please note that we cannot commit at this time to real-time updates whenever a lineup posts but we will update multiple times in the afternoon on most days.
It will be clear which lineups have been incorporated into the projections as hitter playing time will move to 100% for starting hitters and benched hitters will no longer appear in Hittertron and DFSBot. In addition, the link in the Lineup (LU) column will be green vs red (in both cases, they will be linked to the team’s lineup page on Baseball Press). And, as some of you have already seen, we now have a text box at the top of our tools pages where I am time-stamping updates and providing relevant notes.
Lineup Lab has added a green check mark for starting lineup hitters in our co-branded DFS lineup optimizers. The green checkmark should synch up closer to real-time lineup posts than on the Hittertron/DFSBot pages. Players not in starting lineups will be removed near in concert with DFSBot (so if DFSBot has incorporated the lineup and a player is not in it, he should disappear shortly from the lineup optimizer).
The lineup data should lead to the following quality improvements:
- Plate Appearances/At-Bats – While my latest model for projecting PA/AB is doing very well, having the actual batting order removes all of the noise/uncertainty inherent in predicting the decisions of an MLB manager. The PA/AB improvements will disproportionately help the handful of players (roughly 5%) that bounce between leadoff/2nd and bottom of lineup. But with some cellar dweller teams punting the season and the unpredictability that comes with expanded September rosters, the value of this enhancement should only increase as this seasons progresses. My PA/AB model will still be used earlier in the day before lineups have posted and for West Coast games where lineups are not available in time.
- Runs/RBIs – Actual lineup spot will be incorporated into the Run/RBI weights driven by lineup slot and strength. So a hitter moving from 6th to 4th will see an uptick in RBIs (both from more PAs as well as an increased RBI rate per hit/BB/HBP/out) and vice versa.
- All Other Hitter Counting Stats – Since all our projections are ‘rate’ stats (e.g., for hits, I’m projecting H/PA and then multiplying by the projected PA), an improvement in Plate Apperances and At Bats leads to across-the-board improvements (a key reason why July simultaneously marked our best PA/AB/Hits/Runs month)
- Pitching Stats – Opposing lineup strength is a key factor for projecting pitcher performance. Actual lineup data improves the accuracy of those strength factors – especially when a middle-of-the-lineup hitter is benched or on the DL.
Weather – I circled back to this after a failed test in June when I was able to get my hands on past-season stats overlayed with gametime weather. I found the combination of weather + park factors to be an improvement on park factors alone but the overall impact on the projections was milder than I think most daily players would believe them to be. I am not going to divulge exactly what weather inputs and how I am using them but the only weather input that subscribers need to worry about is rain as I do not use threat of rain to hedge projections. I have considered displaying rain probability into the projection tables but I think several sites are doing a better job than I can do for weather. My go-to is DailyBaseballData as it shows hour by hour weather.
Weather factors are incorporated into all of today’s pitcher, hitter, and team (Teamonator) projections. The gametime weather is grabbed early in the morning so I will make a special note going forward if today’s projections do not incorporate weather.
One side effect of these enhancements will be greater shifts in the projected stats of ‘same day’ stats vs the projections from 2-7 days prior. There will also be some shifts (generally minor) between the morning projections (which will incorporate weather) and the afternoon when lineups becomes incorporated.