Confidence Percentages 101
Go ahead: get confident about your wagers versus the Vegas line.
Purchasers of the STAT 2010 handicapping program and subscribers to my weekly High Percentage Picks service get the benefit of trend and game confidence projections that have been uncannily accurate over the past few seasons.
This level of accuracy would not be possible without the use of truly effective NFL trend analysis combined with the following 7 different calculations, starting with:
T-Values
The first step towards creating an accurate confidence percentage for an upcoming game is to look at the past winning percentage of each active trend in combination with the number of games that have been involved.
This calculation is called a 'T-value' and is worked out, in this case, by subtracting losses from wins and dividing this figure by the square of the total number of games. A trend with a 33-3 ATS record would therefore have a T-value of 5 ((33 - 3) / Sqr(36)) as would a situation with a record of 75-25 ATS ((75 - 25) / Sqr(100) = 5).
The first trend may look more impressive because it has only lost 3 times in 36 games; however, the 2nd trend is equally impressive (and of equal value) because of its success over a hundred games as opposed to only 36.
Once a T-value has been calculated, there are a total of 6 different adjustments that need to be applied and they are:
Condition Ratio (CRAT)
2 different situation's with identical records of 60-20 ATS might have the same 'T-value', but, their chance of success in future games could be very different depending on the number of conditions that make up the premise of each one.
As more and more conditions are added to a situation, its record will typically improve. The flip side of this; however, is a corresponding decrease in confidence due to the more complicated logic involved. To put things simply: a situation with a record of 60-20 ATS that was built on only 2 conditions, will probably produce far better results down the road than a 60-20 trend that has 6 different stipulations.
CRAT basically compares T-values with the exact number of conditions involved. A situation with a T-value of 6, as an example, will normally use 6 conditions (the average ratio is actually 1 : 1.1). Situations that produce good results based on a lower-than-normal number of conditions will receive a boost, while those that are more complicated may be penalized.
Average Spread Margin Ratings (ASMR)
Analyzing ATS records alone will only tell us if a team either covers, or does not cover the spread. Looking at the amount of points that teams normally cover (or do not cover) the spread by can give is further insight into the actual strength of the situation in question. Situations that have an ASMR below zero are penalized while those with an ASMR above zero receive a bonus.
Performance in Recent Games (RGAM)
After studying patterns related to NFL situational analysis for the better part of 16 years, it's become fairly clear that recently played games (i.e., from the current or last 3 seasons) do offer a better indication of which way a situation is headed than results from before this time period. Statistical regression to the mean, an effect of the 'law of averages', must also be taken into consideration and based on these 2 factors, STAT weights games from the current season higher than last season, and games from the last 3 seasons, higher than those played before this time period.
As a result, a situation with a historical record of 26-0 that started the season 0-3 would see an accelerated reduction in its T-value well beyond a re-calculation based on overall record alone.
Previous history for this Team (TREC)
Has the current team been involved in this situation before? And if so, what were the results? As with Recent Games, extra weight is applied to past games involving the current team (as long as they occurred within the L3S).
Trend Distribution (TDIS)
Does the situation usually select the same 2-3 teams or is it spread out across the league? Situations that are spread out evenly across the league are less likely to be affected by team personnel and game-plan changes and are more stable as a result.
The very last adjustment applies only to games where more than one trend is present:
Situation Overlap
Some situations and systems can overlap with each other's results. Certain situations in the Playoffs and on Monday Night, for instance, can have similar premises while other situations may act as a 'subset' of a larger one. In such cases, STAT adjusts the T-values for competing situations down, in order to give a more accurate assessment of their total effect on a game.
In Summary
Calculating confidence percentages is as much of an 'art' as a hard science and it's a sobering fact that even the strongest of situations will rarely produce a CP that exceeds 57% ATS, although games with multiple situations working in the favour of one side can help push this threshhold to 65% and beyond.
