We’ve gotten a lot of questions about our methodology and we are happy to describe our process to provide some transparency to how we do our accuracy rankings.
We collect our initial data on Thursday from each of the sites. That way we can write a trends article to be posted on Fridays describing potential boom/bust candidates and some “safer” bets based on the data we are seeing at that point.
On Saturday we do a quick scan to see if any sites have adjusted their rankings and then adjust our database accordingly
On Tuesday we enter in the actual results. For the actuals we use standard performance scoring (the same method the sites we rank use to determine their rankings).
- We also use decimal point scoring, which is to say that a player wouldn’t just score 2 points, he’d score 2.44. This allows us to more finely identify the actual rank order of players. We recognize that many of you do not yet use decimal point scoring, but the trend is headed in that direction, and we have not yet heard of a league that has moved from more complex to simpler rules and want to stay on the cutting edge of Fantasy Football trends.
- When there are ties, both players are given the same ranking and we then continue to calculate from there (ie LT and Shonn Green both are ranked 25th, then Jamaal Charles receives a 27 to account for the fact there were two 25s there is no 26)
- Players that are injured during a game and do not finish the game are removed so as not to unfairly penalize sites/experts that were high on those players. Our rule is that if a player plays a down in the 4th quarter and then are injured they go into the rankings, if they are hurt and unable to play at any point before that they are removed
- Our algorithm gives us a score for the overall accuracy of each position for each site/expert and overall. That becomes how we force rank the sites/experts for our scorecard
- To determine the levels of accuracy of each site we apply a tiered structure that is customized to each position. We customize this to each position because the sample sizes of each position are different. We assess more RBs and WRs because the sites provide us with more data to do so, so the overall accuracy of those positions is lower than TE where the sample size is smaller.
Our scorecard and assessment is then posted late Tuesday night or early Wednesday morning.
We will start tracking a cumulative set of leaders at each position and overall after week 4 once we have enough data to deem our findings statistically significant.
I hope this answers folks questions but if you have more feel free to email us or head over to our forums and ask there!