What Are Good Backtesting Filters?
A good sdql filter is something like:
‘and 7<=total<=9’
totals 7 and 9 in MLB are the two most common (relatively frequent).
So we talk about logic in queries here a lot now. What does filtering out totals outside of that high frequency range mean?
It means that we’re going to focus on equilibrium. I’ve gone on about this in heavier detail in my own writings (journals, blog posts, forum discussions)
I project explicit outcomes of North American sports games using proprietary modeling systems. Some may argue that they are worthless. You can’t make a stats model that consistently predicts 55%+ ATS year in and year out. I respect that point of view, but don’t subscribe to it; however, I think that the naysayers might go out on a limb and agree with this:
If one can create a basic power model; again BASIC….almost similar to Accuscore on espn (which is a BIG TIME roi loser), you still actually use a non-roi yielding power model to project EQUILIBRIUM.
Many ways to skin the cat, but one way I’ve found (and frequently use) to consistently come through with my wagers is to set matchups into two groups:
1. Use the non-roi yielding power model to isolate normal matchups. Nothing ‘smells funny’. Let’s cap these differently than item 2…
2. Use the non-roi yielding power model to isolate matchups that ‘smell funny’ (meaning we’re predicting a tremendous edge for one side, and it is so profound that my gut feels that I’m missing something here (ie. an injury or some odd-ball weather conditions).
#1 = EQUILIBRIUM. This is a great time to comfortably (and with less capping work) use those stealthy angles to tilt the scale.
#2 = CHAOS. This is a great time to perhaps fade your big edge or look for contrary regression.
Some food for thought. I encourage everyone to learn to make a basic score projector model. This is how I cap my games. My record this season in MLB is 80-47-10 +28.71 units and this is precisely how I cap games.