How to build your own valuation picking system. For those who care

[otschamp]

I have received about a dozen emails asking for different forms of advice. Many of them asking about some tips on how to evaluate games and form a value system. So if you already have one great, no need to read this boring post, but if you don't have one and would like to create one for yourself for free read on. There are five steps in creating one for yourself...

1) Gather all (and I mean all) relevant and viable statistics

2) Come up with a score using a detailed formula

3) Get the lines and analyze them like its a stock portfolio

4) Pick the sides and totals that most deviate from the lines

5) Keep track of and analyze all your results like you were a record keeper for the third reich (bad analogy I know).

And poof you are done. I will reply to this thread detailing each step, and hopefully you will decide that you don't ever need anyones advice or a capper to give you a pick. I found the more threads I read on peoples advice the more doubt and fear I got in my own system and the worse I did. If you want to share your system with others feel free, but one piece of advice before I get to the details is don't read about other people's opinions unless you feel there is something you can learn from them. A guy I knew that was pretty darn wealthy told me once that my income will be the average of the 10 people I spend the most time with in my life, so I should be careful who I hang out with. That sounds a little cold, money is not everything granted, but if you take that philosophy into handicapping you will realize that you are trying to win money and not make friends, so maybe you should take that advice into this venture and not take advice from people who don't do better than you. Anywho here we go... For those interested still hehe. I know of 6 that are anyways.

 

[otschamp]

1) Gather info

Ok so what stats are viable and what stats are a waste of time. Well after sitting in front of a computer programming/testing/going crazy at times for 3 straight years I found it differs with every sport. You just need to take your expertise in the sports you love and figure that out for yourself. It will take a little time, thought, and trial and error but here are some steps to get you started...

1) Avoid stats that are in stats. I will give one example a QB's passing yards per game encompass the WR receiving yards per game on that team, and vice versa. So counting and using both is not a viable option for coming up with accurate results.

2) Only count stats that have a direct bearing to the outcome of a game and are not statistical anomoly. Like for example in Baseball, a hit, any hit has a bearing on a game. Calculating how often that hit is a triple is not. Most baseball players will tell you that you cant really focus on hitting a triple. They are harder to hit than home runs for 99% of players and have a lot to do with the defensive shift, a bounce, and luck.

3) Find the obscure stats that make a difference to the outcome. For example in basketball some referees have a tendency of calling a lot of fouls. That in turn increases free throws, the length of a game, etc...

4) Keep track of them for as long back as possible and never lose them. At first it starts with a lot of data enrty but after a little while you find you are actually maintaining better stats on sports, teams, players, than you can find anywhere else. We have an unbelievably large databases of statistics of college players, pro players, teams, games, and much more. I am sure other people have more but ours is starting now to get where we can probably sell it to Pro teams wanting scouting stats on college players before a draft.

5) Its time consuming yes, but you dont have to start big, start small. Pick one sport, if its college basketball, pick one conference. Or even pick one team and try it with them become an expert on that team.

Either way gathering info is the least sexy but most important step. That is all I can think of at this moment on that step lets move on to the next.

 

[otschamp]

2) Come with a score

This is the most intriguing and exciting step for a budding system developer. Don't look at the lines, don't look at the totals. Just look at who is playing, take your stats and come up with a predicted score for the game. You can start easy, take PPG and def PPG and do a little algebra and see what you get. You will find that its pretty raw but 50% of the time when you print out your sheet it will be close to the vegas lines. Sometimes it will be way off, and sometimes somewhere in between. Its a start, then its time to fiddle and faddle until you find what you think influenced the actual final score and the actual line and total to be different than what you had.

Eventually you will need to start doing some things, which come back to the meticulous keeping of records and stats, like creating filters (at least that is what we call them). For example, how do you decide a College Football spread when Florida is playing Citadel. Are the statistics for each team equal? Did they have equal competititon. Is a 300 yard passing game the same in the SEC as the Southern Conference? The answer to that is NOOOO. So you have to use these filters to modify the score accordingly. We use over a dozen filters, some will do nothing unless the situation calls for it. Like increment weather, some are strength of schedule and strength of statistics depends on the situation. Over the years it has become pretty complicated, which leads to my next point. I am not going to lie, you will need to learn some level of math, algebra, programming, etc... Why not, if you don't think you are good at it, go for it. Heck you might invest your money every single day into your bets, why dont you invest some of yourself into it as well.

Now that you are developing modifiers for your formula, its becoming a little more complicated but you will find you are right, if more and more of your score predictions are almost in line with the opening vegas odds. When we moved to Vegas over 3 years ago we made some contacts with some bookmakers and worked in conjunction with them to develop overnight lines. It was a great experience, and it wasn't hard to do, since it seemed like everyone and their mothers booked bets. I was playing in a poker tournament once in Caesars palace and it was the day of the Kentucky Derby and the line was too long to get bets in before the race started. Two guys at my table ran around the room started booking bets 5 minutes prior to the race starting, there were about 300 people in the room and they booked about 100. I guess they are lucky the poker room heads didnt care to call gaming on them or security but my point is being around them you start to realize they go through a similar process with a couple differences, but that is neither here not there.

Bottom line, come up with scores, test them, keep track of your results and you will be able to start predicting with the best of them.

On to step 3... I bet the people who asked for tips are like, hey buddy I didnt ask for all this bs, stfu. Well I got started so I am going on!

 

[otschamp]

3) The lines oh the lines

First off I want to apologize for any typos, I still believe my college education wasnt worth the paper my degree was written on, let alone the mass amounts of debt I graduated with. I digress, so back to step 3.

Well this part will look easy compared to the other 2 steps, but it is crucial you do something here that is just as important as the other two. If you guessed keep track of them, all of them, for all games you are capping, then you guessed correct. Meaning the opening lines, the high and the low, and you will see why later that is important.

It is a little obvious what you do here as the next step:

1) Compare the line with your score. Now you might say, well duh you told me my spread should be similar. Yes exactly, you need to find the ones that arent similar and examine why. Is it public betting patterns, an injury you missed, something like the coach got herpes from his star point guard and there is a lot of animosity in that locker room, etc... If there isn't any viable reason except you are now calculating your scores a little better than you should for certain teams, or vegas is valuating a certain team too high or too low because they have to consider betting patterns into their initial lines. When you find those games you got yourself an opportunity.

2) Now you can assign values to them. The ones further apart have a higher value then the ones that look dead on.

3) Look at the lines like they are a stock market, which they really resemble. Studying that in school and originally coming from that, boy walking into the Hilton Sportsbook on a football sunday sure reminded me of it when i first got to vegas. Just like stocks, they are predictions, they are not always right, a stock price is basically calculated by a simple formula between EPS and P/E ratio. Usually it is dead on, but sometimes it is not, things like investors confidence, expectations of lowered earnings, etc... will fluctuate the stock price. It doesnt mean the stock price should be lower it just means it is, same thing with lines. You will find certain teams lines are always a little off, no matter what. Why? They are popular, like the Yankees, Red Wings, Cowboys, Lakers. You have to compensate for it, but knowing when to make money from it and knowing which way a line is going to move prior to the lines coming out could make you a lot of money.

There is more but my hand is now officially tired, and I will guage whether to finish this post based on any rants, raves, or flames I get. The point is you dont need a capper you just need time and determination and some know how and you can be totally self sufficient. The most important question to ask yourself prior to betting is are you in this for the ego or are you in this for the cash. If its the cash you want then dont compare results, dont focus on what others say, be helpful to who you can, but develop a system you feel comfortable with, and work it to the promise land. Sounds a little corny, geez I do sound like a retarded Tony Little, sorry cant help it.

 

[otschamp]

Didn't get too many flames or tomatos thrown at me so here we go to 4

4) Pick em right.

Ok so now you have a score, you have a line from vegas what do you do? Well tonight we had the Air Force - Wyoming game at 62 - 76. Vegas started the line at -9. We said hmmm 76 - 62 = 14. The line is -9. So the raw value of this pick is 14-9= 5. See how simple that is? The problem is after you pick em and assign a value then you must put that value and that pick through the ringer by putting on to it some deviation modifiers, previous result modifiers for that exact situation. So for example this season the % of the time our program was correct when it picked Wyoming and they were the home favorite was 50%. So and the number of times it was right when it said to bet against Air Force covering in this situation was 100%. So using that info you put that value through a modifier to change its value. Do I mean multiply it? Or divide it? or average it? The answer to that is totally dependant on the type of calculations you did to get to that point. We have a total of 6 modifiers at this point that raise or lower the value. You need to figure out what your modifiers are going to be, but start with a raw value and through trial and error you will learn the pattern and the adjustments you need to make.

Ofcourse dont forget its important to keep track of your results, and we will be analyzing them in the next step. I know it seems like a lot, but dont you want to have a quick fundamentally sound starting point when you look at 140 games on saturday ncaab's. Not just start going, hmmm what do I think looks the best. The work you put in will pay you back 100 fold in games that you will suddenly avoid and opportunities you would have missed just playing the more visible games.

On to 5...

 

[otschamp]

5) The devil is in the details

These results you are getting or not getting are so key to keep that they will be the fundamental difference between winning and losing over the long run. That is the key, over the long run, not short term gains. If you are a maniacal risk loving person, then forget about everything I said and go pick ever +200 or higher underdog in Hockey with large bets and do a couple 8 team parlays as well and once in a while you will get lucky.

For a consistent gain you must know what your strengths and weaknesses are and how to bet them. To explain further what you will find is that any system that is statistically viable will indentify some patterns based on its mathematical biases. For example, our system last College Basketball season was almost undefeated in the MAAC, and did real well in the SEC, yet struggled mightily in the Big West and Ivy Leagues. Why? Well it seems like our system identifies which teams play a consistent brand of basketball, either consistently good or bad according to the way our score predictors calculate scores. The ones that are consistent with our method will seem to pop up on our radar as teams to bet. The reason you need this analysis, is you need to help your system help you avoid teams that just seem to inconsistent to track and raise your bet on teams that you just seem to have nailed this season. Once you have indentified it and assigned higher values to those teams in certain situations (because maybe your system identifies certain teams strengths in specific circumstances) then you can raise the value of that pick or lower it accordingly to max profits.

So keep records, analyze them, scrutinize the results and see where the strengths and weaknesses are and use that info to make some dough.

Finally have a betting system that you stick with. We have a multiplier betting system and we try and follow it religously and we at least keep the results of that system public for all to see what would have happened if you would have followed it. Whatever yours is, stick to it and use it. An example of one is like ours where you multiply your value bet my a dollar amount you feel comfortable with and you bet that amount on the games that are a certain value or higher. So for example $10 is what we use as a multiplier, and if the value is 8 we bet $88 (to win $80) etc... We dont bet games lower than 3, and if we are betting a 3 at least we know we are betting less so if its more volatile it will cost us less. This might sound elementary to many of you pro bettors, but I found after only being on this board a short time that many people are open to suggestions and gladly welcome it. So as fundamental as it might seem, if there is anything that you read here that might save you even $1 or make you $1 then I feel it was time well spent on my part. Hopefully you feel the same.

Good luck with all your ventures,

and remember with a system of your own,

You won't just Beat the spread, you'll OWN IT.

 

[bjh21]

Wow, for 6:30 in the morning that was a bunch of insight but very well said. I have a few questions for you, though. Basically I just started with the BIG XII and I used statsheet.com to gather all the meaningful statistics to predict the total score of a certain team. Without going into alot of detail I was able to find a regression that had an r-sq value of over 97%, which means over 97% of the variability can be explained with this equation. The problem is I feel like I am missing two very fundamental columns in opp fg percentage and opp 3 pt percentage. The problem is I just developed this last week and it's tough to find what the opponent's were allowing coming into a game last december. Do you have a site or link that would give me some of that info??

Also, a basic question in the statistics area do you take each individual game and try to develop an equation from "what" happened or do you guys take YTD numbers and develop an equation that will predict what "will" happen?? All of the info is appreciate and I do feel like I am close. One last question is where you get all of your data?? Is there a site better than statsheet.com?? Thanks in advance!

 

[otschamp]

To answer your questions bjh21...

Boxscores, and situationally compartmentalizing them. We raw input boxscores and then we have a series of scripts, depending on what sport it is, store those statistics into different places for further analysis. For example with college football... For three years now we have every single players offense and defense, how they did against conf, non conf, div IA, div IAA, home, away, if applicable neutral site, and more then we break down all the applicable meaningful stats, like in what weather conditions, in day game or night game, after loss, after win, etc... Its mind numbing either way we no longer do it manually, scripts process the info it takes our server a lot of work to do it and we dont even use all the info in our calculations but I still feel its better to have too many stats. So have I seen any site that has these extensive stats, I would say I havent, hence why we had to create our internal scripts/programs to do that work. Can I post them on our site, I don't feel comfortable doing so, for one major reason. We had an unusual situation last year when I was freely giving them out here locally to people, I was contacted to cease and desist, and that openly distributing statistics of specific players and blah blah was in violation of licensing this and that. I don't know exactly if that was legit if I went to court, but it wasn't worth fighting for. I don't currently have to time to even create a delivery method for people to see it other than just printing massive amounts of rows of data off a database. I would need to refocus on that and I cant at the moment.

I guess to answer your last question, we use a variety of current, historical, but mostly situational stats. The one thing we found is that human behaviour is typically predictable as far as their sports performance. Some people are very predictably un predictable, so we have to vary what stat we use based on the team, coach, player and situation. We create a factor similar to a beta in stock calculation that defines the player/team's statistics. That is also a factor in our deviation modifier. Either way the short answer is that its all situational, and we found the more data we have per viable situation the more accurate it becomes.

Saying all that, it still a game and last night we had the BYU game tied at 65 so although the total was 141.5, our computer said that there is a high enough deviation chance of those exact scores that it ranked that game as a low value over instead of an under. It was right of the OT not right of the score. The goal should be to get within 1 basket of the score at least 25% of the time and within 3 baskets over 60% of the time. If you can average les than 6 point deviation in basketball you can use your results for nice prediction. The problem with last nights game is it still had BYU as a medium level cover, and look what happened in OT. I wanted to throw our servers out a window and urinate on them when I saw that. I actually placed a decent size bet on them. Anyhow I am not sure if that even answered your question, I did just want to vent on that annoying loss. I think i need to retweek the what if we go to OT calc, but its tough to base a prediction on going to ot.

Good luck with what your doing.

Remember...

Don't just beat the spread, OWN IT.

 

[otschamp]

Two final notes...

Remember there will be biased in the way you calculate your formula, and that is not only expected, but necessary. Let say your formula tens to favor the "better team" or the favorite, you will find that your picks will be leaning towards lines that are favorable for those type of teams, and vice versa the other way. The point is you need to know what your formula leans towards and adjust your modifiers and filters accordingly to create accurate values for your picks. Its all in the modifiers after the initial score calculation that makes all the difference.

Second, expect fluctuations. As in life things happen that are unpredictable, like for example Shaq hitting 9/10 free throws the other night against the Suns. Or the Pacers and Magic scoring 249 points, you might think they are high scoring teams I can see that happening. Well if you knew how often one team let alone two teams shoot at almost a 60% fg percentage for an entire game. Its rare, but the point is things happen, don't trash your formula just because of one night or even one week. We have been going at it for thousands of games in all 4 major sports and both major college sports over multiple full seasons, and you should expect it to be time consuming. Just as in our real lives who knows what might happen or has already happened out of our control. I never though 3 years ago when I bought my house that it would appraise last month for half, and who knew the Celtics last year when they were 300-1 prior to the season to win the championship would make a couple trades and poof would win. You job is not to predict the exception, just the rule. The rule tends to be right more than 55% of the time, you will see that if you kept track of your results. That is good enough to make a nice profit.

Good luck.

 

[Kluter24]

That was a very nice read, thank you for the time you put into that! Definately shined some light on a few questions I had regarding getting started doing this the right way. Once again, I appreciate the time you took to help out people with questions, for free.

 

[otschamp]

No problemo

I was worried it was too long. Appreciate the feedback.

 

[jamesmouton]

Yah, thanks for taking the time to shed some light on your system. Wealth of knowledge in this thread.
General question for you. What is your opinion on tempo-free stats (Dean Oliver, kenpom.com, etc)? Do you use them at all, or have you found them useful?

I like them a lot as a way in comparing teams that play at different pace (speed) as far as possessions per game, etc. But do you have any insight into how useful tempo-free stats can be as far as predictors?

 

[otschamp]

Thanks for the response as far as tempo-free stats

It is where we all started, and it is where this journey for us started 3 years ago. The problem with it, especially in college games is that the factors that affect a possible outcome of scores in relations to game lines and betting trends are much more complex than the singularity of a fairly generic formula that doesnt take a those multitude of factors into consideration. Many times those factors dont matter but many times they do, then your formula needs to have layers in that there is a formula to scrutinize the formula after it first comes out. Ill give you an interesting example, we have a similar internal rank as kenpom.com has, so like I said its a great start, but with modifiers I think it would be hard pressed to disagree with our rankings. Back to the example, tonight is one of those nights where the final reformulation of our score after the original similar tempo-free calculation we start with was the key. Some very tricky games tonight, and I feel our last calculation tonight was the most important it completely changed many games values from the original calculation. Some nights it is almost meaningless some nights it isn't when it takes into consideration the very factors that the starting formula doesnt.

I guess it is about your school of thought and what you want to predict, we need to use 6 very specific factors with college basketball that as far as I know are the only ones using these factors.

Thanks for the response, and good luck tonight on all your bets.

Except the ones that go against mine .

 

[woodypaige]

You keep saying we in all your explanations. Does this mean that your a service or others who do the inputs:think2:

 

[Paddywack]

You keep saying we in all your explanations. Does this mean that your a service or others who do the inputs:think2:

Woody, he has announced he is going tout on 31 January.

 

[otschamp]

Yes

The answer to your question is our guys do the inputs.

 

[otschamp]

Hey Paddy

Is that a good thing or bad thing from your perspective. Going tout could mean different things, I am wondering where you are coming from.

 

[Paddywack]

Is that a good thing or bad thing from your perspective. Going tout could mean different things, I am wondering where you are coming from.

Being a tout to me means someone who sells picks, or offers subscriptions for picks. That is your plan at your website isn't it?

 

[otschamp]

Yes, I am sorry, new to that term. From my understanding I thought a tout was someone who begs for money and annoys people into giving them money for services they didnt ask for in the first place. Like those gypsies in Venice who stand around in costume and when you take a picture of them they get abrasive about wanting a tip. I was just trying to clarify, sometimes I think too technically. I apologize if I came across the wrong way with my response.

Appreciate your response and good luck with all your plays tonight.

 

[Paddywack]

One thing you have going for you is displaying all of your knowledge and expertise, should garner you some customers. Maybe you should build up a following first before going tout. You are relatively new here, and not many people know about you yet, or what kind of results you get.