New Member - Indy from Swansea
-
Hi guys, signed up a couple days ago and absolutely love the stats software and team ratings web app, great content!
Who you are? Real name is Indigo (Before anyone asks no I don't have a sister called Violet, but if you've got a new one you won't think I'll have heard then fire away!)
Where you are from? London born, Northampton raised, Swansea bred
What stage are you at in your trading journey? New to the trading side of things however I'm very keen to get involved, I did a lot of research and coding for my Master's thesis a couple years ago where I made a betting system using primitive NBA data which came out to a small profit but could be vastly improved!
What's your background? Currently work as a data scientist so naturally I'm very heavy stats based, the only sports I'm really interested in are combat sports but I'm interested in finding value wherever it may lie. In my experience, combat sports are very hard to make any stats-based decisions on and make a consistent profit.
Why did you join BTC? Very keen to share ideas to make profitable betting / trading systems as a community. Found and listened to a podcast on the way to work and signed up later that night!
What do you hope to get from the community? Further insight into the world of sports betting / trading and work towards a profitable system!
What are some of your trading goals? To eventually have a system backed up statistically that could produce a consistent profit whether this may be betting, trading, or elements of both.
I'm currently trying to find a consistent source of football results so that I could combine the amazing work you've done constructing elo rankings and your other metrics (something I've tried to do before but you seem to have done a much better job haha). A master database with all stats including odds and results would then let us backtest ideas much more easily the way I see things. Just thinking out loud! Great to be here anyway and love the transparency and community vibe already.
Cheers!
Indy -
@indigo-king 100% matey!! Amazing, get cracking on the football
anything you need give me a shout! -
@ryan-carruthers It's definitely something I'd be interested in revisiting in the future, for the moment I'm keen to look into football (soccer) stats and build on the work you guys have done though! Yeah I had so many people tell me that exact thing when I was writing my thesis, as soon as the states get online betting the world won't know what's hit it, there should be value left right and centre! I'll have to have a chat with you at some point mate and share ideas
-
Interesting buddy, I think there might be some data out there we can use (http://footballinsight.co) they also do NBA stats. We haven't used these sports or looked at the stats creation for these at the present time as main sports at the moment for members are Tennis, Football, Cricket & Horse Racing.
I think the States will get online betting in the next few years and then it will become massive. I am sure we can hodge bodge something together that will work
-
Yeah for sure! No reason why it couldn't in my eyes, the only difference is that there is so much data to work with for NBA (and the more continuous nature of the scorelines in theory should provide a more accurate reflection of the skill gap between the teams) that it really gives us a much better indicator of a team's performance. I think a logical extension to my work would be constructing a strong elo system as you have done for football but I ran out of time unfortunately!
The first and main thing in my eyes for any sport using such a stats-based approach that I've come from would be to have a universal database with elo rankings and other performance metrics, odds data and the result. If there's anything like this in the pipeline you've thought of where I could lend a hand or you'd like to share ideas at some point then I'm definitely game
-
@indigo-king Amazing!! Thats the kind of data and algos I like to see.
Do you think this model could be applied to all US sports to make it a year round betting system?
-
@ryan-carruthers Ah the West Coast games can be horrendous! Those 3am games can be a killer.
So I basically constructed a regression model to predict the final difference in scoreline from the perspective of the home team (a lovely thing about basketball obviously being the much more granular scoreline differences due to the amount of points). This used input features constructed from box score data and considered the previous 1, 3, 5 and 10 games as distinct features to see which had the biggest influence (the features considering the previous 5 games were given the highest coefficients and chosen by the feature selection algorithm in the end by the way). I then ran some algorithms to effectively find value given odds data (I used bet365 odds as these were given in the datasets I bought). In the end on the unseen 2016/2017 season I had around a 55% hit rate but when testing on other seasons this wasn't as high and in a couple of cases did not profit, so theres a long way to go to optimise it to be ready for use but it was an interesting project and it gave me a great introduction to the field which eventually led me here!
-
@indigo-king Yes lol the hours are brutal aren't they, I am a huge NFL fan as some of the other guys are and my team is on the west coast which makes it even worse to watch as always the later game.
You say you had a system for betting?
-
@ryan-carruthers thanks mate! Yeah I really liked working with NBA data. You can't beat it for the frequency of matches and the nature of the scores for regression work. The only downside for me anyway is being on the ball for in play trades in the early hours of the morning when I (currently) have to be up for work at 7. Lots of potential there though for sure and I'd be happy to discuss what I learnt from my research
-
@indigo-king Welcome to the group, great to have you here as well matey! Looking forward to seeing some of these ideas, I have started watching the NBA recently I quite like it
