The ScoreMetrics Method is based on incorporating multiple proven, back-tested, profitable systems on sporting events to create a collective pool of trade alerts.  It combines the systems with a technique for risk management and unit allocation, and even builds-in signals for when to abandon a system that is no longer valid (yes this happens occasionally).  To understand the method we first need to break down how we find winning systems.

If you have followed sports (which is by no means a requirement for participating in ScoreMetrics – see the section about widgets) then you may be familiar with what I like to refer to as Old Gambler’s Wives Tales.  Something like warm weather football teams can’t cover on the road in cold weather.

The reality is there is no system derived from a single repeating pattern.  That is a trend and a trend can and will change. It is actually a naturally self-correcting issue because as a repeating pattern becomes more evident, money shifts to trade the pattern, which in turn changes the odds or line to eventually account for the pattern.  Or in some cases the pattern simply normalizes via a ‘regression to the mean,’ which essentially means that data can get out of whack, but ultimately re-centers itself over time, otherwise known as the correction of an anomaly. 

A great example of this is simply analyzing the flip of a penny.  If you flip a penny 10 times it should produce tails 5 times and heads 5 times, right?  Reality doesn’t work this way, however, as any given 10 flips can produce a wide range of results.  Perhaps you flip 8 heads and 2 tails. This appears to be a pattern of heads, and that data could actually continue to widen on your next 10 flips, but over the course of a million flips of the penny the number will come very, very close to 50% heads and 50% tails, as over time the laws of statistical probability will ‘regress’ the data towards its natural mean.  In simpler terms, the law of averages will play out over time.   

So when you hear one of those Old Gambler’s Wives Tales, we must keep this law of averages in mind and realize this type of tip is really just catching a pattern deviating outside of its expected long term average, and over time it will likely normalize either because the odds/lines will adjust for the tendency or due to the underlying action regressing to its mean performance.

ScoreMetrics approaches systems quite a bit differently.  It all starts with an underlying principle or logic, like a certain umpire in baseball has a ‘tight’ strike zone and therefore creates more walks and less strikeouts.  Therefore, one could expect that this would increase the on-base percentage in a game which leads to more runs scored, which leads to higher scoring games and the tendency for these games to go over a particular over/under total.  

On the surface this might hold up for a while, although one could argue that oddsmakers will eventually identify this pattern and adjust accordingly.  However, we start with this core logic and (always) back-test results. We find all umpires that call games tight and check their track record with over/unders, not just recently, but over time.  In this example, let’s say this produces a 55% success rate taking the ‘over’ in those games.

This, in and of itself, is not groundbreaking.  However, ScoreMetrics uses a ‘layered’ approach to systems development.  By this I mean that it takes that original logic and adds other logical layers to the rules of the system.  For example, the next layer might be to only include games in which the wind is blowing out of the outfield side of the stadium in excess of 8mph.  One could argue this will increase the odds of homeruns and thus increase the score total for the game. Now let’s say, for example, that of the games with those ‘tight’ umpires that also has wind over 8mph blowing in the ‘right’ direction, the over hits 59% of the time.  We continue to add additional layers until the system hits the sweet spot. For a system to be in the sweet spot it should have enough layers to optimize the ROI, but not so many layers that the system is diluted by nonsensical inclusions. This is typically 4 or 5 layers, with some instances of 3 or 6 being appropriate.

Always remember that it is not enough to just find a pattern that produces a positive ROI for the last week, month or year.  With ScoreMetrics I mandate that each system must be tested for a minimum of ten years, with at least 8 winning years, in order for it to be accepted into our portfolio.