This evening, Sunday, June 28, we will be making an important change to one of our most visible and important systems.
Our in-season projections are probably one of the features that sets BaseballHQ.com apart from most any other fantasy information service. It is a system that was developed over many years and maintained using elaborate spreadsheet modeling. Tonight we will launch a new system, one that has been automated and will allow for more frequent updating of player baselines.
For those interested in the process, here is how it all works:
Since we were moving to essentially a new platform, this transition has also allowed us to make some tweaks to the projections algorithms, something that has not been done in many years. As a result, many of the balance-of-season player projections are going to change. Most will change by a little; some by more than a little. (Note that the algorithm for calculating Rotisserie values will not be changing, only the raw data that feeds the calculation.)
If you see any player whose projection has changed significantly, please contact me. And by significant, I mean BY A LOT. If a player was projected for 21 HRs the rest of the way and is now being projected for 18, that is not significant. That is three errant gusts of wind.
Comments welcome.
Our in-season projections are probably one of the features that sets BaseballHQ.com apart from most any other fantasy information service. It is a system that was developed over many years and maintained using elaborate spreadsheet modeling. Tonight we will launch a new system, one that has been automated and will allow for more frequent updating of player baselines.
For those interested in the process, here is how it all works:
We start with a Primary Baselines file, which lists every player's statistical rates of performance, based on historical trends. These are the numbers you'd find in the Baseball Forecaster, adjusted as players change teams, roles, etc.
Then we have each player's current stats from 2009.
Then we create an Adjusted Baselines file, which combines the previous two files and creates new baselines for each player. How the data is combined is different for every player, based on playing time to date, age and change in BPIs. For instance, a player projected for a 4.50 ERA who is currently performing at a 2.50 level will be adjusted differently whether he's done it in 50 IP versus 100 IP, is 23 years old versus 37, and whether his BPV is currently 50 or 100 versus an historical level of 50. That is a rough generalization of the process.
The Adjusted Baselines file is then run up against our Playing Time file to produce each player's balance-of-season projection.
Then we have each player's current stats from 2009.
Then we create an Adjusted Baselines file, which combines the previous two files and creates new baselines for each player. How the data is combined is different for every player, based on playing time to date, age and change in BPIs. For instance, a player projected for a 4.50 ERA who is currently performing at a 2.50 level will be adjusted differently whether he's done it in 50 IP versus 100 IP, is 23 years old versus 37, and whether his BPV is currently 50 or 100 versus an historical level of 50. That is a rough generalization of the process.
The Adjusted Baselines file is then run up against our Playing Time file to produce each player's balance-of-season projection.
If you see any player whose projection has changed significantly, please contact me. And by significant, I mean BY A LOT. If a player was projected for 21 HRs the rest of the way and is now being projected for 18, that is not significant. That is three errant gusts of wind.
Comments welcome.
Comment