[terrortwylight]

I understand that this takes time and patience and lots of reading and trial and error..

But can anyone recommend some good building blocks so I can get started on my own?

I realize i'm going to need to build an excel table which I am proficient with, fortunately.

I just have no idea what data points I'm going to be pulling from and how I'm going to pull that data and ULTIMATELY how I'm going to incorporate all of those different factors in a formula.

Any help at all here is GREATLY appreciated.

[DOM_Toretto]

Adjusted field goal percentage is a big one for me.

Also, assist/turnover ratio is a good indicator that often gets overlooked.

[HeeluvaGuy]

One thing that may surprise you is how time consuming (and frustrating) it can be to get the team names uniform for college basketball. There are 350+ teams, and a variety of way teams are listed depending on the site you use for data. It can be a real nightmare.

FWIW, the model I use is based on opponent power ratings.

[incognitoh]

One thing that may surprise you is how time consuming (and frustrating) it can be to get the team names uniform for college basketball. There are 350+ teams, and a variety of way teams are listed depending on the site you use for data. It can be a real nightmare.

FWIW, the model I use is based on opponent power ratings.

You are right about the damn team names; lol. I built a lookup table using betonline, cbs sports (standard stats), kenpom (advanced stats) and ncaa.com (standings). It was a pain but now can copy in the spreads from betonline and everything populates automatically.

[PorkChop]

You are right about the damn team names; lol. I built a lookup table using betonline, cbs sports (standard stats), kenpom (advanced stats) and ncaa.com (standings). It was a pain but now can copy in the spreads from betonline and everything populates automatically.

Probably gonna have to manually enter each name

[incognitoh]

Probably gonna have to manually enter each name

Nope, not if you build a lookup table with one key site as the main input. I can copy and paste directly from betonline and my model automatically updates with no typing.

I typed in all of the games last year and it was a pain in the ass...

[*TheRisk*]

www.teamrankings.com is great. I copy and paste stats into Excel. Quick and easy.

[terrortwylight]

www.teamrankings.com is great. I copy and paste stats into Excel. Quick and easy.

So you do this daily then? And then that data directly effects your model predictions for the next day's matchups?

[terrortwylight]

One thing that may surprise you is how time consuming (and frustrating) it can be to get the team names uniform for college basketball. There are 350+ teams, and a variety of way teams are listed depending on the site you use for data. It can be a real nightmare.

FWIW, the model I use is based on opponent power ratings.

So I assume you have an Excel page with the names of all the D1 schools? like 300+ lines or so?

And what do you mean by opponent power rankings? Do you incorporate that into the same page with the team names?

[terrortwylight]

Nope, not if you build a lookup table with one key site as the main input. I can copy and paste directly from betonline and my model automatically updates with no typing.

I typed in all of the games last year and it was a pain in the ass...

I was thinking basketball-reference.. I just don't understand which stats are relevant (or which stats are most relevant for predicting outcomes).

[terrortwylight]

I realize this is asking A LOT..

but is these any way you guys would be willing to share ANYTHING with me? Even just a screen shot of a small part of what you guys are doing?

I'm more than willing to put the time in on this but I really am clueless right now.

[HeeluvaGuy]

I'm still learning and tweaking, and sharing what I have created over time probably wouldn't be very helpful unless you were inside my brain. But here's the basic concept I started with: http://sportsgambling.about.com/od/b...a/ncaaaopr.htm

From there I added multiple resources for power rankings, and for some sports I've been able to add home/away, and recency components. My "models" aren't super sophisticated, but it's remarkable how close my predicted lines and totals can get to actual lines.

Honestly, college hoops is the most difficult sport to do this with (outside of baseball for different reasons). Besides the shear number of teams, there's the problem of tournaments and neutral site games that make it really hard (1) to create season-long schedules (a must for automating things), and (2) to categorize stats (is a neutral site game home or away? How does site X treat neutral site games?).

But as far as stats go, all I use are power ratings (or opponent power ratings) and points for/against (home and away and in the last two weeks). You can create excel sheets to pull the raw data for you, but even that can take some time when there are literally hundreds of games per week.

When I identify potential plays from my model, I then look at matchups and stats from placed like kenpom and teamrankings. That's where the real time is spent...

[*TheRisk*]

So you do this daily then? And then that data directly effects your model predictions for the next day's matchups?

My approach is very unconventional, confusing, and time consuming. My opinion is that margins are the most important thing in comparing teams. Most will find this strange, but I only use three statistical margins and they are all based on scoring (Free Throws, 3pters, and 2pt FGs). Many will argue that rebounds and turnovers are important, but at the end of the day those things lead to more shots and thus more points and that is why I don't use them. All this said and years of using it, statistical analysis can only supplement your decision making and shouldn't be the only source of picking games.

[incognitoh]

I was thinking basketball-reference.. I just don't understand which stats are relevant (or which stats are most relevant for predicting outcomes).

College Basketball can be so random that I would suggest building a model based on Kenpom Stats; Adjusted Tempo, Offense, Defense and SOS; add in a factor for Home Court Advantage and measure the results. Then go to another site and copy in the basic stats; 2Pt FGs, 3Pt FGs and FTs and calculate an expected outcome.

When the two results both indicate the same result; FADE IT. That has been my experience...lol. I am still looking for the right "formula(s)" myself.

I understand regression to the mean; but I think that is much more relevant in Pro sports where you can compare the current season stats of a pitcher to their career stats, for example; but tough to do in College Basketball in my opinion.

Best of luck to you; but I wouldn't even bother trying to gamble in any sport without some type of quantitative model. But I, too, am still trying to get it right...

[HeeluvaGuy]

How's the model coming?

[terrortwylight]

How's the model coming?

I've been reading about Monte Carlo simulation and trying to figure out how to incorporate it into what I'm trying to accomplish. I'm still kind of baffled so I'm just reading as much as I can at this point.

I'm also trying to figure out how to develop my own power ratings. Once I have those, I think I understand where to go from there. But I still can't figure out how I'm supposed to calculate them to begin with.

I'm pretty math savvy and I'm proficient with Excel, but I'm having a really difficult time understanding how to actually get started.

[Louisvillekid1]

Hasn't Ken Pom done this for you already?

[terrortwylight]

Hasn't Ken Pom done this for you already?

Yeah, KenPom is an awesome resource and it's the first thing I look at when I cap the games. I just want to build an additional resource.