Projection Accuracy

#49107  4/20/2016 7:08 PM

Some guy on Reddit's DFS board has been tracking the projection accuracy of various MLB sources. Last night was first night BM was included and he only tracked hitters for some reason, but the results were poor. I've been using the daily projections here off an on since the season started and they seem decent.

How much 'worse' (or better?) is this product for DFS relative to the basketball product which I think is great? Anyone else have any insights into how well the projections here have been doing this first month? They use Steamer projections as the base, right?

Link to accuracy analysis for last night: https://www.reddit.com/r/dfsports/comments/4fothq/last_nights_mlb_projection_accuracy_41916/

ken_admin  4/20/2016 7:20 PM

I assume they're analyzing the accuracy of the total projected value which seems like a good estimate.  It'll be good to follow the results over a few weeks.

For projections, we generate our own, and the references to Steamer on the website are for users who imported theirs since we did not generate full-season projections this year (only short-term).

#49107  4/20/2016 7:40 PM

Thanks Ken. Any rough insight into how you create your projections? I know it's proprietary and you can't explain everything. Would help my confidence in the product though to have a better understanding. Basketball and baseball are such different sports and thus very different to model, and so I'm curious.

Anyone else compared Baseball Monster last year or so to other projection sources?

ken_admin  4/20/2016 7:53 PM

We look at a past # of plate appearances (or innings) for each player which differs based on each stat.  We also adjust for park factors for those PA's. For newer players, we rely on minor league stats and generic adjustments based on their league. This establishes a baseline that we then apply our daily factors to (such as batting order, opposition, and park).

#49107  4/21/2016 9:54 AM

I know in basketball it's vital the projected minutes for each team's players equates to 240. Is there any similar standard for baseball? It would be a good idea if the projected runs scored for each player mirrored their team's Vegas team total.

Exnihilation  4/21/2016 11:53 AM

We look at a past # of plate appearances (or innings) for each player which differs based on each stat.

I am using the "Comparison Columns" feature on the DFS projections page and want to compare players' full season performance vs recent performance to gage if a player is on a hot streak or in a slump, but I am unsure of how best to define "recent". Should I only look at the past week? Past month? Half season? Quarter season? You get the picture. 

So my question is: Based on how you formulate your projections, in your opinion what timeframe would best reflect a player's recent performance? I can understand wanting to keep the specifics about how you produce your projections a secret, so if you don't want to disclose this information I will understand. 

ken_admin  4/21/2016 12:13 PM

For the base projections, we use a fairly long time frame, and then a short-term adjustment is made as part of the factors.

If you're looking to see if a player is "hot" I would not go very long, days up to a week.