Building a Model/Database

[The Peoples Champ]

If some of the older users have seen this thread repeatedly, I apologize in advance, but a poster asked about this in another thread, so I figured a new discussion might entail.

Some posters (TheSportsMarket among others) have a database and model system in place that clearly is working for them, and I'm sure there are others, too. It is a timely, potentially costly adventure, but if done properly, it can pay dividends.

I'll put my thoughts out there, and feel free to add, pick at, or post your own thoughts about if and how you've done it, whether you feel it's worth it, etc.

I once attempted to construct an MLB model, with the idea to get all of the relevant information in, focusing on ERA, WHIP, and AVG, among other things, to get me started. I created a scraper, and began learning some programming, and got about a quarter of the inputting done before I hit a time wall, and was never able to finish. However, when I was able to utilize it, at hit at a 56% clip over 105 plays, which isn't an adequate sample size, but it was certainly helpful. I'd love to get back and finish it, but I just don't see the time opening up to do so.

I think this is probably what happens with most potential database/model builders; unless you have enough time to consistency build it up at the start, you're never going to finish it.

Does anyone have any quick starters or way to more efficiently gather information so you can begin queries faster?

Thanks for any input!

[TheSportsMarket]

There are some databases that are already out there. Either by networking and sharing (like we said were willing to share our raw data), or to buy or for free on the internet. Here's a link for football stuff (obviously a little late): http://thenationalchampionshipissue.blog...-data.html

For basketball utilizing sites like Ken Pom and 82Games and then linking them to excel has been a huge time saver for us.

Covers has a great database of lines, but other than copy and pasting them into excel we haven't found a more efficient way to deal with this issue.

[Beyond the Bets]

If you are into this sort of thing, I'd recommend checking out the Handicappers' Think Tank at SBR Forum. Click here.

That place is infested with math-heavy people who are always scraping data, building dB's, etc.

[The Peoples Champ]

(12-23-2011 10:01 AM)Beyond the Bets Wrote:  If you are into this sort of thing, I'd recommend checking out the Handicappers' Think Tank at SBR Forum. Click here.

That place is infested with math-heavy people who are always scraping data, building dB's, etc.

Hey BTB, what's your policy with linking to other places? I was going to link to the same forum over at SBR you did, but didn't want to step on anyones toes, as SBR gets all pissy when you link off their site.

Anyways, yes, it's an awesome little niche to take a peak at if you're interested.

---

(12-23-2011 09:34 AM)TheSportsMarket Wrote:  There are some databases that are already out there. Either by networking and sharing (like we said were willing to share our raw data), or to buy or for free on the internet. Here's a link for football stuff (obviously a little late): http://thenationalchampionshipissue.blog...-data.html

For basketball utilizing sites like Ken Pom and 82Games and then linking them to excel has been a huge time saver for us.

Covers has a great database of lines, but other than copy and pasting them into excel we haven't found a more efficient way to deal with this issue.

Here's a question for ya, TSM:

As I said, I believe the first issue database builders come in to is time, but the second one has got to be once all the raw data is input, where do you go from there?

Without giving away any of your doings and methods, do you or anyone have a recommendation for how to start going about the process of picking what you want to use, properly weighing it and adjusting it, etc.?

[TheSportsMarket]

Yes, it takes alot of time to put together a database (great thing to do over the summer when there's not as much to bet on). As far as what to do with it. We just came up with ideas of how to model data and then backtest it. Backtesting was the key for us. We had ideas for how to model thigns and would just play them and were pretty much a breakeven/losing bettor. When we started incorporating backtesting to various algorithms to figure out which ones worked that swhat we really started to turn the corner - hope that helps!

[Beyond the Bets]

TPC,

I don't have a linking policy as long as it's useful and not promotional in nature.

[The Peoples Champ]

(12-23-2011 11:59 AM)TheSportsMarket Wrote:  Yes, it takes alot of time to put together a database (great thing to do over the summer when there's not as much to bet on). As far as what to do with it. We just came up with ideas of how to model data and then backtest it. Backtesting was the key for us. We had ideas for how to model thigns and would just play them and were pretty much a breakeven/losing bettor. When we started incorporating backtesting to various algorithms to figure out which ones worked that swhat we really started to turn the corner - hope that helps!

It does, thanks!

Another FAQ while we're at it:

How did you go about the backtesting process? I've seen a bunch of ideas thrown around, curious how you guys did it.

[TheSportsMarket]

(12-23-2011 12:17 PM)The Peoples Champ Wrote:  

(12-23-2011 11:59 AM)TheSportsMarket Wrote:  Yes, it takes alot of time to put together a database (great thing to do over the summer when there's not as much to bet on). As far as what to do with it. We just came up with ideas of how to model data and then backtest it. Backtesting was the key for us. We had ideas for how to model thigns and would just play them and were pretty much a breakeven/losing bettor. When we started incorporating backtesting to various algorithms to figure out which ones worked that swhat we really started to turn the corner - hope that helps!

It does, thanks!

Another FAQ while we're at it:

How did you go about the backtesting process? I've seen a bunch of ideas thrown around, curious how you guys did it.

Took historical lines, compared them to our projections for those games and counted the record. We did it all in Excel. Utilizing the "If" function alot, along with data tables. The key is utilizing the averages or data AT THAT SPECIFIC POINT IN TIME - not how a teams average finished or what not.

Thats why we use raw game by game data, because in game #7 in 2010, you can calculate a teams average whatever going into that game. That will compare when you're handicapping a game, because you have their data up to that game to project on, not after.