I pulled this discussion about wrestlestats.com out of the depths of the #4 OK State Vs. #5 Missouri thread... and wanted to follow-up on it separately.
It is directed to andegre to ask what the plans are for the wrestlestats website? Is it intended to be a revenue generating site eventually? My interest is driven by the fact that your (or whomever's) basic idea of allowing the user to adjust lineups and instantly do comparisons between teams is truly innovative and very helpful.
You have always been open to requests for improving the site. While I don't have any specific requests for tweaking the site, I do have a few long-term suggestions for how to more thoroughly mine the database you already maintain to produce more valuable content for viewers (or hopefully future subscribers).
1. score all duals upfront using your rankings and create predicted team outcomes. Viewers will undoubtedly want to change the lineup and you can automate the predictions to reflect lineup changes. I have read some of your discussion of using algorithms to do this so am assuming you have a comfort level with building probabilistic models. I would think it is time for someone to begin leveraging data analytics to a greater degree in wrestling (as it is in most other sports) and building predictive models for upfront scoring of dual meets. I think this would be of tremendous interest to wrestling fans if they could see a predicted outcome for each dual and make their own adjustments to the lineup.
2. produce a wrestler on wrestler comparison. Much of the debate on the boards centers around hypothetical head to head matches. I would think the historical database you maintain would produce enough data for building models to predict outcomes. As always, there is a margin of error, but a model based prediction using a well defined statistical model would be very interesting. You could begin by building models that explain the influence of a number of factors on the likellihood of winning a match (home vs away, years wrestling, past success, current year record, high school rankings, team success, normal weight class, starter vs reserve, coming off injury, etc...) and then match this with head to head and common opponent data. There is a lot of discussion about how useful these metrics are (or are not), but I have never seen any serious analysis of the data to reach a conclusion. I would think this would be of value to coaches as well.
I could offer other suggestions but wanted to start the conversation and see if it is of any value to andegre or just suggesting more work for him.... : ))
andegre
It is directed to andegre to ask what the plans are for the wrestlestats website? Is it intended to be a revenue generating site eventually? My interest is driven by the fact that your (or whomever's) basic idea of allowing the user to adjust lineups and instantly do comparisons between teams is truly innovative and very helpful.
You have always been open to requests for improving the site. While I don't have any specific requests for tweaking the site, I do have a few long-term suggestions for how to more thoroughly mine the database you already maintain to produce more valuable content for viewers (or hopefully future subscribers).
1. score all duals upfront using your rankings and create predicted team outcomes. Viewers will undoubtedly want to change the lineup and you can automate the predictions to reflect lineup changes. I have read some of your discussion of using algorithms to do this so am assuming you have a comfort level with building probabilistic models. I would think it is time for someone to begin leveraging data analytics to a greater degree in wrestling (as it is in most other sports) and building predictive models for upfront scoring of dual meets. I think this would be of tremendous interest to wrestling fans if they could see a predicted outcome for each dual and make their own adjustments to the lineup.
2. produce a wrestler on wrestler comparison. Much of the debate on the boards centers around hypothetical head to head matches. I would think the historical database you maintain would produce enough data for building models to predict outcomes. As always, there is a margin of error, but a model based prediction using a well defined statistical model would be very interesting. You could begin by building models that explain the influence of a number of factors on the likellihood of winning a match (home vs away, years wrestling, past success, current year record, high school rankings, team success, normal weight class, starter vs reserve, coming off injury, etc...) and then match this with head to head and common opponent data. There is a lot of discussion about how useful these metrics are (or are not), but I have never seen any serious analysis of the data to reach a conclusion. I would think this would be of value to coaches as well.
I could offer other suggestions but wanted to start the conversation and see if it is of any value to andegre or just suggesting more work for him.... : ))
andegre