Looking For Basic Advice/Direction on Backtesting

[BeatingBaseball]

I have a theory that I would like to backtest. I've of course used a ton of statistical data in my capping of MLB over the years - but have no experience in the area of backtesting.

Is there a talented modeler out there who would be willing to give me some basic direction on how to set up a relatively simple computer analysis of historical data that would involve both 5 inning and full game opening and closing line Totals and 5 innning and full game Totals outcomes over a full season?

If you could even direct me to a good book or primer it would be greatly appreciated. Hopefully I just need a starting point as I am reasonably computer and math literate - but nowhere near many of you on this forum.

Thanks.

[Data]

how to set up a relatively simple computer analysis of historical data that would involve both 5 inning and full game opening and closing line Totals and 5 innning and full game Totals outcomes over a full season?

You need to collect/obtain that data. Then, use Microsoft Excel which is a simple yet powerful tool.

[BeatingBaseball]

Thanks Data. I wasn't sure if Excel was the appropriate tool for this.

As to the data collection, do you think there are databases available that contain both the the 5 inning and full game lines and outcomes which can be somehow downloaded into the spreadsheet or is it necessary to collect it manually and then do the data entry by hand? This may be a stupid question in this day and age - if so I apologize for my ignorance.

Thanks again for your advice.

[Data]

There are websites and individuals that sell this data in the form of Excel spreadsheets, ready for use. Google it. Also, I saw an ad type of post in the forum's promo section a few weeks ago, they were offering data for any major league and college as well. If you know programmimg you can obtain this data from donbest.com. Note, you are looking for 5 inning lines and those are not widely available as full game lines.

[BeatingBaseball]

Thanks again, Data. Appreciate your help.

[suicidekings]

When you do get going on this project, you need to set aside some of the data for use in the testing portion of your model, but not use it for the creation of the model. Do a quick read on data mining before you start.

[LLXC]

When you do get going on this project, you need to set aside some of the data for use in the testing portion of your model, but not use it for the creation of the model. Do a quick read on data mining before you start.

Yup, typically about 1/3 for cross validation.

[BeatingBaseball]

Thank you, gentlemen. I can see I have a few things to learn - knew that I would - if I'm going to get valid results on this.