i think dan is right to use publicly avail power ratings and predictions to get the approx. proper spread and then go from there. not sure you should model the entire process unless you enjoy it.
-
Comment -
the thing is the door is wide open for people that maybe dont have the scrape tools or skills to set up a model. they don't need to, they can use a basic model and some common sense to create a pretty fair line. I am not saying this method is foolproof and simple but it takes a true square and makes them a ton sharper. Most people bet and they have no idea what the line should be, they say I don't need a line is like a fireman saying he does not need a Scott Airpak just because he doesn't know how to use it. He might survive forever and prosper without a Scott Airpak but he would definitely be a better fireman with more optiosn if he had one and learned how to use it!Comment -
Poor druggie rudyComment -
Comment -
of course not alfa there is not a single model out there that does not require human interaction.Comment -
sure the input parameters would be the same and it would spit out a base model but very few people I know that model go by only the model spit out. they make adjustments themselves after that.
do you really believe that there is much difference in most advanced models? you can literally find 100s of models and papers on models on the internet with a simple google search. Most college math students build a model and publish that model online. the beauty or art or separation in models is the adjustments after the factComment -
Heehaw do you use a fully automated model?
if Lebron is on a minute restriction your model automatically adjusts for that?Comment -
not sure if dan will agree..... but i think on college football (for instance), get the predictiontracker.com and some other power ratings predicted spread...
AND THEN,
something you could look at is mediocre team with good run D (but terrible pass D) vs. good team that predominately runs the ball. basically that mediocre team is getting shredded by most team with its passing defense, but it'll be ok against a run team..... i think dan will say that's already in the line.... this is a bit one-off, like playing Army or Navy, but i could relax the idea some and get more ideas on good bets.
i think EVERYTHING is in the line, but is it properly priced?...............
it's like warren buffett and investing, he doesn't appear to have any special skills but he sensed something others didn't, like the power of insurance company and consumer staples stock's ongoing profitabilty. sort of an obvious idea, or not such an obvious idea? i think everyone had the idea but not the strength of the underlying trend.Comment -
just an aside, but the Army-Navy UNDER has been going on forever now. never even comes close to the closing total, let alone some insane 47 point total that lasted for 5 minutes.Comment -
I think most things like O and D and comparing is for sure in the line already. A base model is good to get a base line but then you need to adjust for injuries, rest and weather.
also if I do find a factor that I believe is not in the line, the first thing I have to do is determine its value. A players mom died last night, I assume that is not in anyones model and in a sense I think it actually has line value but the question is how much and if I cannot quantify it close, I will just skip it or the event in particular.
example last year
the WNBA Aces did not get into a hotel they had flight delays and crap and were still scheduled to play, I know this moved the line but how much it should move the line, I dont know so I had to skip that game
A its in the line
B if its suppose to be in the line and isnt how much is it worth, if I dont know I got to skip the game but take note of my believed change it had on the line for future reference.Comment -
Dan, do you believe certain people with enough work and innate ability have a superior "feel" for pricing/sentiment/bias/etc and can win long-term on a large number of games? p.s. i don't mean myself.
one reason though that academics love betting markets for research is that it's very tangibly short-term results-oriented whereas the stock market has the "beauty contest" (maynard keynes) angle.Comment -
cal now down by 10 and that is the line too
oops, wrong threadComment -
of course look at pokerjoe, he makes his living off adjustments and he use to use EA sports as his rating system for players, I am not sure if he still does that but yeah of course. Player benefit is very subjective to the game, quality of the team as a whole and his backup. these are not math problems these are more like an art. But most art projects can be turned into skillComment -
i have fully autonomous models running on stand alone systems that perform scraping, predictions, evaluations, bet executions. automation is important.
One thing about models... you can have 7-8 years of success, and lines can adjust and your model can lose and become worthless. Just because you put a lot of effort into it, and you had success, doesn't mean it can't expire. Models are like milk. Unless you stay up with it, and work to improve it, it will eventually fail after some number of seasons.Comment -
hey Miz and your models adjust for Lebron being on a minute restriction all by themselves? if so that is fricken awesome, never seen that done automatically!Comment -
Fight between the sharps
Who will win
Who will have a better ROI
To be honest Dan, the answer to that situation is probably to simply avoid Lakers lines rather than try to incorporate that unique scenario into a model. As sharps, you likely have models for several sports. Avoiding one team is not going to destroy your profitability and the uncertainty of a player worth 3-5 pts in the line makes for a higher variance in those games, reducing your likely edge since you never know how long he will play.Comment -
I agree but the issue is a fully automated model would have to be pretty slick to account for this by itself. I think this kinda opens the door of the need for human review of the model and of course adjustments.
if Lebron plays say an avg 40 minutes normally and if out is worth 6 if he played only 20 I would assume its worth about 3 or 4 and that would not be adjusting for the other obvious issues that cause this in the first place!Comment -
no but definitely not as much as you think it does but that isn't the point
You can either lose a little edge here and there running it automated or do all the little shit but put it in a lot more effort...
This is about money at the end of the day and having a high $hourly is the end goal so...
Like I said before this ain't the forum for youComment -
"lose a little edge" I work all day nad night to get any little edge I can and you can afford to just lose a little, that is crazy to meComment -
I think lots of games have situational changes and that can drastically effect your overall edge. I cannot see how it is not wise to manually review. if you design some mega monster bot that can go and just crush the line everyday, yeah I could see how a few things like Lebron only playing half the game has little effect, wait what?Comment -
I think some of us might model different so the importance to you may be drastically different than the importance to me of some of the situational stats that I am implying require manual review. so this argument might be really not be an argument and just different methodology
I think most good cappers, do both, they dont just listen to a model and they dont not listen to it either. a little sweet science with a dash of art!
\if the packers were -8 on my model and they were currently priced -2, I would be hesitant after adjustments to bet that. I also have like a 1.5 pt rule but if I see some intangibles I might take the packers at -6.5 when according to my model they should be -6 for me to take themComment -
I used to think that too. Then I tested pretty extensively, over several thousand bets, and was surprised: most of these extra little elements do nothing. Some were already accounted for by other features, or combinations of them. Some I was able to add extra info to account for them directly.
At this stage you then learn that most common wisdom and understanding in sports are based on myth and lazy clichés.Comment -
15 years of experience in statistics has taught me that if you have a bias to find evidence in your numbers nothing can stop you. in the betting market it is possible to find value in numeric data, but it is not a simple mathematical equation nor a formula you can rely on year in year out.Comment -
If you are using those laughable ratings you mentioned earlier, like KenPom or Sag, or then I suppose it's a problem.
But I've created ratings and inputs that, in the end, show that so many of these situational changes contribute to and and are just noise in the big picture.
Successful experience may be necessary to properly learn those lessons.Comment -
There is not much to look at overall. You can use python to scrape the data you need from publicly available sites... and get opening and closing lines from the books that matter. Then you have to develop a hypothesis (relating output to inputs) Then put the idea together (a long time ago simple regression worked but not these days so much) ... then back test and check out z-scores etc. See if there is any reason to think you are datamining or overfitting. When you test it, use out of sample test set. Look at your graphs and sweep the result versus the delta between line and your prediction... if you see an improvement in performance that is a good sign. If you do better against openers than closers that is also a good sign. if you see 54% to 57% then you probably have something real.... ... this is all very general info
Elihu's book is a pretty good starting point if you're interested in learning the basics. some people talk shit about it to feel smart but the book is pretty good in my opinionComment
