Im sure you have but have you tweaked your model at all tom? And is it doing better recently?

I've mostly settled on things but I'm still changing little things. For example, my most recent change was reducing the degree to which I was regressing a player's shots against. I was previously regressing it 30% (year-over-year correlation was around 70% for this stat), but felt that was too high (remember we're betting intra-season not year-over-year). So that changed my projections 10c or so. I don't tweak things based on results because that's too noisy. I still bet opening faceoffs even though the results sucked because I like my process (take COL +110 at BM if you want to fade me). It's a mixture of how I see my stuff lining up with the betting lines and the bets I end up with vs what I would expect to be betting based upon how I project things. For example, if you pay close attention you'll notice there are a few players I tend to pick on more than others and when that player is in a particularly good spot to go over/under I bet it.

Overall I'm pretty happy with how SOGs have done and bet them myself at various places, although I think Pinnacle is the sharpest book dealing these so I'm least eager to bet Pinnacle lines which is why those get posted here.

Sorry rambling to answer your question yes, the model I eventually mostly settled upon was what got me back to even after being stuck huge.

NJD Jack Hughes u2.5 -202
NJD Miles Wood u2.5 -200
ANA Danton Heinen u2.5 -175

Are there no nba or nhl player props up on Pinny today?

Iowa! Extra CBB YTD: 31-29

McNeese State +4
Grambling -1
Jackson State -3

these XFL lines haven't really moved after Tuesday

is the market efficient now???

I just placed this XFL Future, thoughts???

Los Angeles Wildcats win XFL 2020 +1500

only $500 limit

SOGs Daily: 2-1-0, -0.02u
SOGs YTD: 243-229-2, -0.59u, 579.2u risked, -0.1% ROI

CAR Justin Williams u2.5 -161
TBL Blake Coleman o2.5 +122
BOS Ondrej Kase o2.5 +116
PHI Jakub Voracek u2.5 -202
NYR Pavel Buchnevich u2.5 -134
NYR Ryan Strome u2.5 -125

Note I'm finding value in a bunch of guys who were traded in-season. Could this be a weakness in methodology to improve upon? This is the observe, tinker, improve cycle.

Without knowing more about your model, tough to say.

General question, would it be possible to code a model in python? And what language do you guys usually use for your models? Or do you use just spreadsheets? Sorry, I believe sigs answered this before but I forgot. And the thread is tough to dig though.

I switched everything I do to Python about 11 years back, when Python 3 was released. Before that it was some rigged system with PHP and R. There's built in functions for everything you want to do in Python now which is nice. You used to have to write all your own functions to do basic things like logistic regression.

I also use R occasionally. More for data cleaning or just getting really quick views of something. Everything is stored in databases so python and R can easily connect and pull data from them.

Don't know much about excel but I know some people swear by it.

Late steam coming in on Seattle, down to +11 we doing it guys.

lol 4-0 houston getting boo'd at home after 1 quarter. greatest fans

What BS that is, I already won my bet with +12.5 but there is 2 seconds left and they could tie it up. WTF. The interview with the guy in charge was lol

Easy money on Seattle +13.5, thank you homie

Cash Seattle.

So the databases that you connect to with python, they're just Mysql databases?

i switched to postgres a couple years ago, but before that mysql was fine.

TomG Total Bases v2 died and I figured I'd just clone the backup and make the move then. For things I access non stop (e.g., current lines, projections) they're stored in a local flat file which gets updated occasionally from the database. Just so I'm not always queuing it non stop.

Also multiple databases hosted on different servers for different things.

welp the under was looking good until about 25 seconds ago

70 yard bomb followed by pick 6.

still a chance but not looking good

Lol this game

oh nope

in the last 45 seconds

70 yard pass
pick 6
kick off returned for TD

lol

SOGs Daily: 1-5-0, -6.00u
SOGs YTD: 244-234-2, -6.59u, 587.42u risked, -1.1% ROI

https://www.eviews.com/home.html

anyone ever use this software? my friend who used to be an economy academic said it could probably help me and it seems like it could be legit but 1.6k seems like a lot to figure out if it's for me or not

has anyone here used it? any reviews?

You could probably find a student to hook you up with a $50 university edition license. Let us know what you think.

I've never heard of this software so I can't comment on its usefulness. In general, I'd think you'd be better off digging into R/Python libraries. There are so many pre-buiIt packages that do most of the work for you. The difference is a GUI vs. command line. With R/Python you'll learn actual useful skills that will translate into many other future projects. Plus there is a huge number of other people using those languages so getting answers to your questions on stackexchange is easy.

FWIW coming from someone who has used MATLAB and R most of their life, some of the functions in the major python packages are kind of suspect mathematics wise. One of the major functions in scipy uses the inverse matrix operation instead of solve which is a horrible idea (resource intensive, compounding errors). Coming from a stats / numerical optimization background a lot of their implementation seems inefficient. You can tell what packages were written by mathematicians vs. comp sci people.

Here's a nice post that sums up why we don't want to invert matrices

https://civilstat.com/2015/07/dont-i...x-why-and-how/

They also do weird things like numpy only returns one mode, but it's just the first mode that occurs and doesn't indicate there might be multiple modes (vs. pandas which does that). After using it enough tho you know, ok if I'm interested in mode use pandas, but pandas sucks for this function so use numpy for that. Except pandas wants the data in a dataframe and numpy doesn't, so I guess I'll have to have two versions of my data loaded, etc.

yeah I'm slowly learning more R and python but it's a slog - i've signed up for a demo and will to negotiate a "sports hobbyist" price if i like what I see

for example was trying to do some scraping with R recently and needed the selenium package, selenium had issues with java so i reinstall java but still no dice, via stack overflow discover i need docker so go to install docker only to realize docker only works on windows professional (that's what i get for not using a mac or linux i guess) and so all that time i spent teaching myself how to do some scraping via r and selenium was all wasted

issues like that don't happen with excel, if i can't get excel or power query to pull the data for whatever incompetence i have that day, I can always pull out the old manual copy paste and then move along instead of spending several hours only to learn that I'm trying to use software that's incompatible with my operating system

not sure what the right terminology is but learning to code cold turkey is a nightmare because every resource i use assumes basic competency and months of going through basics of "hello world"

#socialscienceproblems

will report back whatever the outcome

I can't imagine an informed person recommending that over Python or R, which are free anyway.

I've always been an R guy, but am starting to use Python more because industry prefers it for computer science reasons, so I want to be proficient. I doubt that would matter to you, but idk what you're looking to do with your life.