1. #1
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    NFL system working well... when do I get to declare...it works...

    As some of you know, I have been developing 5 computer driven models to predict the outcome of sports games. College Football, College Basketball, NBA, NFL and Baseball. Well, so far College Basketball worked one season with 2 seasons backtested prior to the winning season. I heard the argument that it doesn't work until it can predict. Well, it did one season. Plus 60%. Do I get to declare victory it is only was season predictive? But it did work... Next I have 5 seasons of NFL back tested and optimized. Using those same parameters, I am now 21-12 this season. Sunday I went 9-2. Again, I have previous seasons tested. When can I declare that it works. Unbelievably to me it predicted the Dallas Philadelphia under and the Houston under. I personally struggled with the Dallas game but bet it. The Houston game I didn't bet because they had a new QB. I lost the Giants game but that was because of new QB also... Any way, I know it's a rhetorical ?...Do I ever get to declare a working system? It seems in this business you are only as good as your last week.... Thanks Dave

  2. #2
    a4u2fear
    TEASE IT
    a4u2fear's Avatar SBR PRO
    Join Date: 01-29-10
    Posts: 8,147
    Betpoints: 35447

    I have a system back tested and ready to say it works. I tested all the way back to 2003 I think, the year where all 32 teams were in existence.

    Even 5 years may not be enough because of the recent rule changes to quarterback hits and offensive point scoring. IF your system can make it through the last ten years, I think you're ok

  3. #3
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Well it has to be tested going forward also...

    I do agree with some of the posters that it doesn't work until it works predictively. But on two programs it has worked. Well, it hasn't made it through the NFL season... but I am 21-12 off to a good start. College basketball has actually worked predicting one season... Thank you Dave

  4. #4
    easyliving
    easyliving's Avatar Become A Pro!
    Join Date: 06-25-12
    Posts: 8,876
    Betpoints: 11

    backtesting especially in the NFL is not too useful as the game has changed so much the past couple years with all the new rule changes that favor offenses. Even the kickoff being moved up 5 yards is a huge deal. so many more touchbacks and a play that could have been a kick off return is now a touchback.

  5. #5
    hutennis
    hutennis's Avatar Become A Pro!
    Join Date: 07-11-10
    Posts: 847
    Betpoints: 3253

    Quote Originally Posted by usma1992 View Post
    . but I am 21-12 off to a good start... When can I declare that it works...
    It is laughable. You ask for a serious consideration and yet you refuse to do one crucial thing required to be taking seriously.
    You refuse to properly post your picks ahead of time. You keep on insisting your words should be taking at face value, which is, as I said, is laughable.

    But even if it is 21 -12 with implied probability of 50%, so what???
    You simply got lucky so far to the tune of 8.1%. That's it!!! Big freaking deal!!!

    It's like making a lucky guess that the first person you meet today in a subway was born in December, and then start asking if it is enough to declare yourself a legitimate psychic.
    That's how stupid your ambitions are.

    Grow up! Or, at least, educate yourself for a change.

  6. #6
    chilidog
    chilidog's Avatar Become A Pro!
    Join Date: 04-05-09
    Posts: 10,304
    Betpoints: 956

    There's huey! ready to pounce on you and bitch-slap you with his mega-dose of negativity. Alright now, thread is now over. Huey has come in and wrote his novel on why you're actually stupid and lucky.

  7. #7
    hutennis
    hutennis's Avatar Become A Pro!
    Join Date: 07-11-10
    Posts: 847
    Betpoints: 3253

    Quote Originally Posted by chilidog View Post
    There's huey! ready to pounce on you and bitch-slap you with his mega-dose of negativity. Alright now, thread is now over. Huey has come in and wrote his novel on why you're actually stupid and lucky.
    Can you at least point out where I am being wrong in my novel?
    Do you even care if I'm right or wrong?
    Or you simply don't want to believe that I'm right b/c it is not want you want to believe in?

  8. #8
    Insoluble
    Insoluble's Avatar Become A Pro!
    Join Date: 05-29-10
    Posts: 71
    Betpoints: 3463

    Quote Originally Posted by hutennis View Post
    Can you at least point out where I am being wrong in my novel?
    Do you even care if I'm right or wrong?
    Or you simply don't want to believe that I'm right b/c it is not want you want to believe in?
    Well you are wrong in your first accusation. You claim he wasn't posting his picks. He seems to be posting his picks well ahead of time in the appropriate forums here.

    As for when OP can declare a method as profitable or working, an adequate sample size will only tell you that your method was successful outside of luck. But that will only be in hindsight. The market is always adapting and a good modeler must keep an eye on his edge over time to see if his model is losing accuracy. Closing line evaluations are a good way to do this.

  9. #9
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Thank you for coming to my defense ...lol

    but no need. Again, I have nothing to hide... Yes, HU... instead of pouncing on me and looking like a dumb*** for the umpteenth time. Check the NFL forum... that is where I posted the NFL picks. Sorry I thought that made sense to post them there. Dude relax a little.. If I am lying about my system it is on me because I will have lost money on the bets. But again, I have been working on this for three years...I want feedback and real critique... I am not selling anything.I just want to make it better and see if I can do it.

  10. #10
    HeeeHAWWWW
    HeeeHAWWWW's Avatar Become A Pro!
    Join Date: 06-13-08
    Posts: 5,487
    Betpoints: 578

    Goto pickmonitor. Post picks. All can then see record.

    Otherwise ......... not sure what the point of this thread is.
    Nomination(s):
    This post was nominated 1 time . To view the nominated thread please click here. People who nominated: evo34

  11. #11
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    In my own defense, HU... I may just be getting lucky. But I am using the same parameters that I used for the previous 5 seasons. So if it is luck, which I am not sure if it is or it isn't, does that mean the system was lucky for 5 years in a row... just saying...

  12. #12
    YourAllAmerican
    YourAllAmerican's Avatar Become A Pro!
    Join Date: 06-25-09
    Posts: 47
    Betpoints: 907

    Quote Originally Posted by usma1992 View Post
    I have 5 seasons of NFL back tested
    As has been pointed out several times, no you don't. You have said you went back to look at a past seasons, then used in-season data to tweak"forecasts." That's not how backtesting works. So you don't have five seasons of results, instead you have a sample size of 33 games.

    Bet your own system and quit worrying about what other people think. But don't try and pass it off here as being validly backtested, either.

  13. #13
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Not really interested in picks monitor right this second...

    ...All American... can you describe me a good method to back testing. I feel like I have done it but maybe I haven't. I have gathered 5 years worth of data and created a system over time that works every past season using the same weightings as the parameters that I am using this season. What is the definition of back testing? I am struggling with this. I thought that was exactly what I am doing... Anyway, I know I am not there yet but I think I am closing in. It is nice however to be 21-12... Dave... If I can college football working that would even be better. I have one maybe two working systems...NBA is ready to launch for testing I am excited about it. College basketball worked really well last year no need to make major adjustments. Enjoy... Dave

  14. #14
    flsaders85
    flsaders85's Avatar Become A Pro!
    Join Date: 11-06-11
    Posts: 68
    Betpoints: 5253

    Use 3 seasons worth of data to develop...and the next 2 to test predicitive power. Probably not even enough data to account of luck and the randomness of football.

  15. #15
    tb1984
    tb1984's Avatar Become A Pro!
    Join Date: 09-11-08
    Posts: 3,112
    Betpoints: 1083

    From Wikipedia: http://en.wikipedia.org/wiki/Backtesting

    Backtesting is jargon used in financial industries to refer to testing a trading strategy or predictive model using existing historic data.


    By reading your post, I think you're falling into the backtesting's pitfall, overfitting (like other posters have pointed out). That is, it is often possible to find a strategy that would have worked well in the past, but will not work well in the future.( From Wiki).

    As always, a model only works when it can predict future.

    So, before last week, your record was 12-10, then maybe(maybe not) you got lucky and went 9-2 last week. I think you should at least test your system for the rest of the season.


  16. #16
    nikossf
    I dont play at SHIT books
    nikossf's Avatar Become A Pro!
    Join Date: 03-02-10
    Posts: 2,217
    Betpoints: 138

    Just post picks. Let us decide how good it works.

  17. #17
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    LOL...Nickoff ty for the insight...

    I will test it the rest of the season. As I have said I think it is over inflated to the over bets and not enough unders... but I am not going to change it until it fails to work. Enjoy...

  18. #18
    chilidog
    chilidog's Avatar Become A Pro!
    Join Date: 04-05-09
    Posts: 10,304
    Betpoints: 956

    why do you care what we think of your system? if it's working for you, then it's working. who gives a fuk what we think? ppl want you to post ur plays to "prove" to us that ur system work? who gives a fuk. if ur not tryin to sell the system, then have a ball, quietly make your money without worrying about what we think of your system.

    my only question about the backtesting: was it backtested in "real-time" using data available at that time, as if that day were a present-day event that had not happened yet?

  19. #19
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Yes the testing was as if it was present at that day... which means my system was looking at the same data Vegas was ... good night... 0-3 tonight solid... LOL

  20. #20
    chilidog
    chilidog's Avatar Become A Pro!
    Join Date: 04-05-09
    Posts: 10,304
    Betpoints: 956

    Quote Originally Posted by usma1992 View Post
    Yes the testing was as if it was present at that day... which means my system was looking at the same data Vegas was ... good night... 0-3 tonight solid... LOL
    losses happen. i developed my system the same way, backtesting with realtime data at that time. it took awhile, as it had to be done manually, but it's been quite the success lately. and yah, i lost yesterday as well - it happens.

  21. #21
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Thanks... I think I am missing some features in college football because I have yet to win consistently. Dave

  22. #22
    josie88
    josie88's Avatar Become A Pro!
    Join Date: 02-12-09
    Posts: 48
    Betpoints: 2188

    All right I'll bite.
    Consider:
    At a win rate of 63+% betting full Kelly how long before my system would win 1 billion units? Does that time period sound reasonable? Am I kidding myself?
    What is the Z value of my predicted bets? What Z value is my goal? Why?
    What resources can I find regarding sample size? What sample size do I truly need to prove to myself I'm right? What other methods are there that could help me prove my model works?
    Am I using any proprietary variables in my model? How do I adjust them or create new ones to replace the older non-performing ones to compensate for constantly correcting markets?
    Why do I keep posting here, getting slammed for my refusal to post my plays, and still not post my plays?

    It's not that people want you to fail, but if you've been around here for a while you see so many people trying to convince themselves they've discovered the millenium it gets boring.

    Post your plays with the model odds. Go one better. Post your actual tickets.

    Have you actually made any bets? It doesn't appear so. You won't know 'til you bet.

  23. #23
    YourAllAmerican
    YourAllAmerican's Avatar Become A Pro!
    Join Date: 06-25-09
    Posts: 47
    Betpoints: 907

    Quote Originally Posted by flsaders85 View Post
    Use 3 seasons worth of data to develop...and the next 2 to test predicitive power.
    Yup, there's your starting point. Suppose you have data available for the 2008-2012 seasons, use only 2008-2010 to create your formula, then use week-by-week results from 2011 and 2012 for testing. An example:

    Using 2008-2010 data, suppose you notice that it appears you can predict a team's game score by taking their average rushing yards over their previous four games and dividing by 7 (I'm just making that up, so don't take it as being aything real). Then go through each relevant week in 2011 and 2012 using that same formula AND using only data available before the game was played AND not making any tweaks to the formula.

  24. #24
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Fair...ty...

    I will add NFL season 2007 to my data base after the season is over. I had fully intended to do just what all American explained....than as I was adding 2010 piece by piece. I noticed a formula was wrong and had to rerun 2011 and 2012. By the time I noticed the formula was incorrect most of 2010 had already been input and I just decided to leave it as is... Anyway, hopefully NFL stays the course and college quickly catches up...

  25. #25
    indio
    .
    indio's Avatar Become A Pro!
    Join Date: 06-03-11
    Posts: 751
    Betpoints: 3954

    While there is no definitive magic number as an answer to your question, one can analyze the probability of certain occurrences and make a rational judgement. If you are judging the validity of a systems ability to produce positive equity bets on 50/50 propositions, then you need to determine the chances that your success was just variance of a 50% probability.

    If you were to go 60-40 through 100 picks of 50/50's, there is a .028 chance that was random, or approximately the same chance as rolling snake eyes on a dice roll. Hardly conclusive.

    If you were to go 180-120 through 300 picks of 50/50's, there is a .000317 chance that was random variance of a 50/50 (not factoring the probabilities that you have only a 53/47 edge, which after vig makes it non profitable). Even though the chances are 1 in 3,154 that this was random, even this is not conclusive.

    If you were to go 350-250 after 600 games on 50% bets, the chances of this being random variance of a 50/50 is .000026 which is eerily the exact chance of making a straight flush on pick'em poker (38,460-1). You can start to feel pretty confident your system is valid when the chances of it not being random variance is 99.9974% But believe it or not, that is still not considered statistical certainty in the world of statistics.

    If you can win 850 out of 1500 50% bets (56.67%), the chances that this was random variance occurring is about 10 million to1, I think you can rest pretty easily in knowing you definitely have a system producing an edge when you start to achieve numbers like that.

    As far as practical usage of a system, one must realize any system that produces conclusive edges will eventually be countered. A perfect example of that is the Wong teasers. While bookies fell asleep at the wheel thinking they always had an edge in teasers, statistical analysis showed a player edge in teasers that went through the 7 and the 3. This took many years to be near conclusive, and bookmakers adjusted by opening lines at 9.5 with adjusted money lines, reducing payouts, and other methods.

    So if by some miracle you actually have a "system" that produces a concrete edge strictly obtained by numerical equation and nothing else, you probably want to keep that information to yourself.
    Last edited by indio; 10-26-13 at 05:25 PM.

  26. #26
    usma1992
    usma1992's Avatar Become A Pro!
    Join Date: 08-02-11
    Posts: 870
    Betpoints: 2220

    Indio... ty... great feedback...

    And yes... once I am there I will keep it to myself.

  27. #27
    nolocks
    nolocks's Avatar Become A Pro!
    Join Date: 12-17-13
    Posts: 1

    Quote Originally Posted by indio View Post
    While there is no definitive magic number as an answer to your question, one can analyze the probability of certain occurrences and make a rational judgement. If you are judging the validity of a systems ability to produce positive equity bets on 50/50 propositions, then you need to determine the chances that your success was just variance of a 50% probability.

    If you were to go 60-40 through 100 picks of 50/50's, there is a .028 chance that was random, or approximately the same chance as rolling snake eyes on a dice roll. Hardly conclusive.

    If you were to go 180-120 through 300 picks of 50/50's, there is a .000317 chance that was random variance of a 50/50 (not factoring the probabilities that you have only a 53/47 edge, which after vig makes it non profitable). Even though the chances are 1 in 3,154 that this was random, even this is not conclusive.

    If you were to go 350-250 after 600 games on 50% bets, the chances of this being random variance of a 50/50 is .000026 which is eerily the exact chance of making a straight flush on pick'em poker (38,460-1). You can start to feel pretty confident your system is valid when the chances of it not being random variance is 99.9974% But believe it or not, that is still not considered statistical certainty in the world of statistics.

    If you can win 850 out of 1500 50% bets (56.67%), the chances that this was random variance occurring is about 10 million to1, I think you can rest pretty easily in knowing you definitely have a system producing an edge when you start to achieve numbers like that.

    As far as practical usage of a system, one must realize any system that produces conclusive edges will eventually be countered. A perfect example of that is the Wong teasers. While bookies fell asleep at the wheel thinking they always had an edge in teasers, statistical analysis showed a player edge in teasers that went through the 7 and the 3. This took many years to be near conclusive, and bookmakers adjusted by opening lines at 9.5 with adjusted money lines, reducing payouts, and other methods.

    So if by some miracle you actually have a "system" that produces a concrete edge strictly obtained by numerical equation and nothing else, you probably want to keep that information to yourself.
    Hi Indio, or anyone else, how did you do the math on estimating the amount of randomness on 50-50 propositions?

  28. #28
    indio
    .
    indio's Avatar Become A Pro!
    Join Date: 06-03-11
    Posts: 751
    Betpoints: 3954

    Quote Originally Posted by nolocks View Post
    Hi Indio, or anyone else, how did you do the math on estimating the amount of randomness on 50-50 propositions?
    `I used a binomial calculator. My answer and methods were fairly simplistic, hence determining the chances of a true 50/50 probability occurring at higher rate through various sample sizes. There were many variables that could have been added for a better answer, but it really wasn't needed for the original question.

  29. #29
    possum11
    possum11's Avatar Become A Pro!
    Join Date: 10-20-11
    Posts: 172
    Betpoints: 948

    usma if you want some advice on what your college football system might be missing I might be able to offer some. I have hit 60%+ the past 5 seasons.

  30. #30
    steel26
    steel26's Avatar Become A Pro!
    Join Date: 11-14-12
    Posts: 53
    Betpoints: 2163

    Hey where did you go?

  31. #31
    v1y
    v1y's Avatar Become A Pro!
    Join Date: 05-02-11
    Posts: 1,138
    Betpoints: 1633

    You don't.

    You just roll in the money, and smugly acknowledge to yourself it works.

  32. #32
    TheSuit
    TheSuit's Avatar Become A Pro!
    Join Date: 03-17-14
    Posts: 7
    Betpoints: 48

    For sure V

  33. #33
    noyb
    noyb's Avatar Become A Pro!
    Join Date: 09-13-05
    Posts: 971
    Betpoints: 6821

    wow, retarded thread (even though it's old). you supposedly developed a winning model that can potentially earn you a lot of money, and all you care about is when you can declare victory (whatever the penetrate that means) to some random people on the internet you have never and will never meet. i can't believe i spend 5 minutes of my life reading this.

  34. #34
    suzie
    suzie's Avatar Become A Pro!
    Join Date: 05-09-14
    Posts: 22
    Betpoints: 12

    ok i have a friend who developed systems for each sport. This year for NFL he added two more systems using past 11 yrs results

    this is how he does it he asks a question that involves all teams like How does team perform after scoring 100+ points within 5 games EXPLAINING the 5 games that's the max amount of games used he said most times happens within 4 games some teams can go a full season or more without totaling 100 pts or more

    So far he is done up to the Colts and he showed me his book with results in that one catergory that will boggle your mind



    DO not try to do this on your own as in adding up the game totals there's a method to be used.

    he had to redo every team after completing 5 of them took him another week.

    anyway good luck on your model.

  35. #35
    suzie
    suzie's Avatar Become A Pro!
    Join Date: 05-09-14
    Posts: 22
    Betpoints: 12

    OK as far as back testing goes my buddy boy started with 1999 season up to 2013 some things you need to use past 15 years because the systems do not have enough results in just past 5 years

    like favored after a s/u dog win many teams are always dogged or always chalked rare to be dogged

    1 team is tops to fade over past 15 years this team is 1- 12 ATS 7 times lost the game outright and are 100% losers 0-4 VS a div. team
    that's game and spread they lost the $ and the game.

    Another thing I know no computer can read into teams emotional state and any software used in handicapping any sport is not to be used fully you need your eyes ears and resources on TV and radio in whatever sport you bet on.

    Like can a computer tell you that yu darvish is owned by div. Rival A's ??? I dont think so. Saw it come up on ESP ticker was 1-6 vs them.
    now is 1-7 Computers cannot read or hear they only do what you tell them to do and as always Garbage in = Garbage out

    This is main reason for every store errors the goofballs didn't program scan codes correctly and you look at your receipt when you get home and see you got charged $10 more.

Top