Get Weekly Projections – The GLSP Projection App for Running Backs
Get a free NFL subscription for 3 days.
A few years ago I came up with GLSP (or GILLESPIE) projections when I was trying to think about a way to forecast for weekly lineup decisions. A common refrain to hear in weekly lineup columns is something like “DeMarco Murray ran for 130 yards the last time he faced the Giants defense.” That seems like reasonable information to consider when setting your lineup, but it also has a fatal flaw. That’s just one game. It’s a small sample. So in order to address this problem I came up with GLSP projections (and GLSP actually stands for Game Level Similarity Projections, so the word “projections” is redundant – you may think that’s stupid and yet how many times have you said “ATM machine”?). The idea behind GLSP is that we take the search for games against an opponent and make two changes to widen the search. We widen the search to include similar players and also similar defenses. So instead of just looking at DeMarco Murray against the Giants defense we would also look at Frank Gore’s game against the Giants defense. Gore and Murray touch the ball a similar number of times per game and they are roughly the same size. But then we would also look at Murray’s game against the Eagles defense. The Giants defense and Eagles defense allow roughly similar numbers to opposing RBs. All of this searching is done with a simple algorithm that scores each matchup by similarity. After 25 similar matchups have been found we can look at the fantasy scoring in those games in order to get some expectation of what might happen when Murray faces the Giants defense. As an added bit of improvement this year I’ve also added a point expectation for the offense (based on using the over/under and the point spread from betting lines) to the similarity search. Now you know that you’re not only looking at games between similar players and similar defenses, you’re also looking at games that oddsmakers think will result in similar amounts of points for the offense. The GLSP RB app below is one of the apps that we’ll have on the site in order to help you with weekly lineup decisions. You can select an RB and the app will do most of the heavy lifting in terms of projecting for the upcoming week’s matchup. It also has a control that will allow you to adjust usage for your subject player. For instance, if I want to forecast David Wilson it probably doesn’t make sense to use his 2012 numbers as he is in for a usage bump. I can adjust the slider and get to a closer estimate for Wilson and that new estimate will be used for the similarity search. As a caveat given that we’re entering week one of the season, the error rates are likely to be large as we collect new data about how teams are using their players. So the app will give you some expectations during the first few weeks, but after we have more data the estimates will be more reliable. Also, once we have the WR and TE version of the GLSP apps up you’ll be able to use them together to make flex lineup decisions. Here’s a quick breakdown of the tabs: Tab 1 – Projection This is essentially the results tab. It includes the stats that resulted from the similar games matchups. It gives you low, median and high projection for three scoring formats. Tab 2 – Player Game Log This tab will give you the stats that are going into the projection on the player side. If you adjust the game touches slider at the top of the app, games will appear and disappear from this tab and the averages will change. Tab 3 – Matchup Summaries This tab just shows you the averages for the season of the players/defenses that are included in the similarity search. The averages are net of the actual subject game, so the averages give a picture as to what the player/defense did outside of the game being studied.