Hi folks,
My name is Steve, aka Steamsharp, I am a Vancouverite and Mathematician and I wanted to bring some expertise to the board solely on the mathematical side of the business.
Since I met Arch on Twitter I have come to respect his views and thought I could offer the board a look into some of the most advanced mathematical methods on obtaining win expectancy for a sports matchup.
Instead of going on and talking bullshit, ill just post our data on the first 1800 games or so of baseball this season. we do this for all north american sports save football since data is scarce
however, xslx data is not allowed due to php.ini restrictions! oh well! glad to be aboard.
S.
Steamsharp Chronicles.
-
ProlinePlayer
- Site Admin
- Posts: 2076
- Joined: Sun Mar 31, 2013 6:33 am
Re: Steamsharp Chronicles.
I for one am really looking forward to your input here!Since I met Arch on Twitter I have come to respect his views and thought I could offer the board a look into some of the most advanced mathematical methods on obtaining win expectancy for a sports matchup.
I've fixed this and the attachment should now be accepted. Please try again.Instead of going on and talking bullshit, ill just post our data on the first 1800 games or so of baseball this season. we do this for all north american sports save football since data is scarce
however, xslx data is not allowed due to php.ini restrictions! oh well! glad to be aboard.
PLP
-
steamsharp
- Posts: 38
- Joined: Mon Aug 26, 2013 8:34 pm
Re: Steamsharp Chronicles.
Compiled WE's attached for download.
How to interpret:
WE in our minds is the precent chance a team has at beating a randomly chosen team performance of a set of 5000+ games. we predict the performance on the field and ask our computer to take the last 10 k home or away games and figure out how many of those team performances would they beat on average. WE is NOT a win probability. It is only a measure of how a teams recent data will perform against random performances drawn from a large set of recorded games in baseball hostory. baseball is random.
when you calculate your WE's you look at the implies WE of pinnacle -> 1/odds and then compare the % value difference between expected WE and implied WE. if there is a lot of signal (high delta) we bet it. you must understand that there is always a lot of Noise in the system so you are looking for enough signal to overcome the randomness in the game. i'll leave it up to everyone to interpret as they will.
Caveat: our WE's are very poor at understanding when a favorite team is going to win despite being ridiculously juiced. use caution when looking at our data. we make lots of money on underdogs but we have a big bankroll that can withstand the dreaded "all juiced day" truly the worst part of betting is a game like the dodgers today where the books are hilarious and give you 1.339 odds. best to avoid this.
Bets of the day: Dbacks, Rangers, Rockies, Twins, Mets, Orioles, Blue Jays, Marlins half dime a peice!
Cheers, S.
How to interpret:
WE in our minds is the precent chance a team has at beating a randomly chosen team performance of a set of 5000+ games. we predict the performance on the field and ask our computer to take the last 10 k home or away games and figure out how many of those team performances would they beat on average. WE is NOT a win probability. It is only a measure of how a teams recent data will perform against random performances drawn from a large set of recorded games in baseball hostory. baseball is random.
when you calculate your WE's you look at the implies WE of pinnacle -> 1/odds and then compare the % value difference between expected WE and implied WE. if there is a lot of signal (high delta) we bet it. you must understand that there is always a lot of Noise in the system so you are looking for enough signal to overcome the randomness in the game. i'll leave it up to everyone to interpret as they will.
Caveat: our WE's are very poor at understanding when a favorite team is going to win despite being ridiculously juiced. use caution when looking at our data. we make lots of money on underdogs but we have a big bankroll that can withstand the dreaded "all juiced day" truly the worst part of betting is a game like the dodgers today where the books are hilarious and give you 1.339 odds. best to avoid this.
Bets of the day: Dbacks, Rangers, Rockies, Twins, Mets, Orioles, Blue Jays, Marlins half dime a peice!
Cheers, S.
-
steamsharp
- Posts: 38
- Joined: Mon Aug 26, 2013 8:34 pm
Re: Steamsharp Chronicles.
Results: 2013-08-28
7/15 Correct - 0.3 units
Tampa bay 9th inning last strike last out bad beat breaks the spread even.
7/15 Correct - 0.3 units
Tampa bay 9th inning last strike last out bad beat breaks the spread even.
-
ProlinePlayer
- Site Admin
- Posts: 2076
- Joined: Sun Mar 31, 2013 6:33 am
Re: Steamsharp Chronicles.
steamsharp wrote:hey PLP. i have some issues with file size. the data block is about 1.4 MB and the php.ini has a block at 256kb.
sorry about that! ill post some raw WE's for today for you guys here in about an hour.
will fix
-
steamsharp
- Posts: 38
- Joined: Mon Aug 26, 2013 8:34 pm
Re: Steamsharp Chronicles.
Compiled WE's attached for download.
lots of ridiculous juices today.
lots of ridiculous juices today.
Last edited by steamsharp on Wed Aug 28, 2013 3:12 pm, edited 1 time in total.
-
steamsharp
- Posts: 38
- Joined: Mon Aug 26, 2013 8:34 pm
Re: Steamsharp Chronicles.
Sure PLP, thanks for eliminating the negative elements. I think to discuss what the math guys are doing these days, there are two concepts that need to be entertained
A) Pinnacle can be beaten and they are still subject to market pressure
B) the "edge" is created by connecting historical data algorithmically to expectation of future performance.
lets break it down.
A) It seems like a lot of people dismiss betting pinnacle. In math based approaches, syndicates calculate a hard Win Expectancy by connecting up historical data using machine learning, statistics aka: poisson functions and the like. this is their calculated break even and its a hard number regardless of the odds on pinny.
Pinnacle is creating odds by responding to market pressure. so for instance, if 90% of the public is on the rays today, well that drops the price to X and if we see that our statistical expectation implies a price that is better then X, we bet that game. its really just that simple. if you have lots of data and you are great at creating WE models, you are guaranteed to make money betting the entire index over the long term. This is mostly due to pinnacle over juicing their lines in response to the weight of money and its a REALLY big weakness in how they do business. A great analogy is facebook stock. if you shorted it at the IPO, you got stupid rich.
so we understand that sports are pretty noisy so we intentionally bet every single game (like a mutual fund) and do beat the closing lines at pinnacle. we do this by betting games where we feel the "signal" beats out the noise. any given team can win on any given day. it seems for us, all of the value is locked up in mispriced underdogs. (for instance Kershaw last night at 1.313 was ridiculous, notice our data clearly pointed to value on the offensive end).
now i read up on the parlay systems here and i agree OLG is pretty juicy! god they suck at PS but i assumed like BCLC they took paddy powers lines which means your edge no longer exists. ill look into that. syndicates these days just stick to large steam on asian markets using systems like i described above.
B) This is extremely important! i have humorous conversations with Arch about edges on twitter and I can never explain to him properly what we define as an "edge". we define this as a mathematical discrepancy in the closing lines caused by lack of attention to the historical data. for instance take the last 90 innings of sports data and create a distribution of hits. find the expected number of hits and correlate that to the expectation of winning with respect to a training set of 10000 previous historical games. do this for many statistics that correlate well to winning and you can often create edges (almost always on overjuiced games) that can be mixed with professional knowledge to generate high returns over time. The hardest part of this is having the good judgment to stick to underdogs (like the cubs last night) and ignore bets where the square fave is going to win. very hard!
please ask me any questions I am happy to discuss.
S.
A) Pinnacle can be beaten and they are still subject to market pressure
B) the "edge" is created by connecting historical data algorithmically to expectation of future performance.
lets break it down.
A) It seems like a lot of people dismiss betting pinnacle. In math based approaches, syndicates calculate a hard Win Expectancy by connecting up historical data using machine learning, statistics aka: poisson functions and the like. this is their calculated break even and its a hard number regardless of the odds on pinny.
Pinnacle is creating odds by responding to market pressure. so for instance, if 90% of the public is on the rays today, well that drops the price to X and if we see that our statistical expectation implies a price that is better then X, we bet that game. its really just that simple. if you have lots of data and you are great at creating WE models, you are guaranteed to make money betting the entire index over the long term. This is mostly due to pinnacle over juicing their lines in response to the weight of money and its a REALLY big weakness in how they do business. A great analogy is facebook stock. if you shorted it at the IPO, you got stupid rich.
so we understand that sports are pretty noisy so we intentionally bet every single game (like a mutual fund) and do beat the closing lines at pinnacle. we do this by betting games where we feel the "signal" beats out the noise. any given team can win on any given day. it seems for us, all of the value is locked up in mispriced underdogs. (for instance Kershaw last night at 1.313 was ridiculous, notice our data clearly pointed to value on the offensive end).
now i read up on the parlay systems here and i agree OLG is pretty juicy! god they suck at PS but i assumed like BCLC they took paddy powers lines which means your edge no longer exists. ill look into that. syndicates these days just stick to large steam on asian markets using systems like i described above.
B) This is extremely important! i have humorous conversations with Arch about edges on twitter and I can never explain to him properly what we define as an "edge". we define this as a mathematical discrepancy in the closing lines caused by lack of attention to the historical data. for instance take the last 90 innings of sports data and create a distribution of hits. find the expected number of hits and correlate that to the expectation of winning with respect to a training set of 10000 previous historical games. do this for many statistics that correlate well to winning and you can often create edges (almost always on overjuiced games) that can be mixed with professional knowledge to generate high returns over time. The hardest part of this is having the good judgment to stick to underdogs (like the cubs last night) and ignore bets where the square fave is going to win. very hard!
please ask me any questions I am happy to discuss.
S.
Re: Steamsharp Chronicles.
Sharp I completely understand how your going about finding "edges". It boils down to the most simple explanation which is u have your break even # and If pinny is significantly off from your number you bet it. Its not Rocket Science I totally get it and have for many moons.
What my concerns are is that you may have the Public/Square factor represented at pinny as a tad to high, And I and many here would argue that the "Public" factor 99% of the time for Major high limit events Does NOT move odds at pinnacle close to gametime.
Are you familiar with Elihu Feustel? Pinny hired him many years back and he vouched that the odds on high limit events are only moved by Smart$ on BOTH SIDES. The public only has a significant effect in the more limited markets at pinnacle. Take MMA A bunch of fans betting $25 each may move their MMA odds LOL That wont happen for Champions League
What my concerns are is that you may have the Public/Square factor represented at pinny as a tad to high, And I and many here would argue that the "Public" factor 99% of the time for Major high limit events Does NOT move odds at pinnacle close to gametime.
Are you familiar with Elihu Feustel? Pinny hired him many years back and he vouched that the odds on high limit events are only moved by Smart$ on BOTH SIDES. The public only has a significant effect in the more limited markets at pinnacle. Take MMA A bunch of fans betting $25 each may move their MMA odds LOL That wont happen for Champions League
-
steamsharp
- Posts: 38
- Joined: Mon Aug 26, 2013 8:34 pm
Re: Steamsharp Chronicles.
i got you. well here is the kicker i think.
there are many many methods to connect historical data to break evens. in financial markets there are hundreds of HFT algorithms that compete against each other. only a few succeed. Using financial markets as an analogy, what we are concerned mostly with is having better arms then the other guys.
we don't really care about smart money or moving lines.. to me it seems beyond the point. just need a price that makes sense. We just gladly bet every single game (which most people say is retarded) but lo and behold after 2000 bets we are up 50k on an 100k bankroll. average odds bet = 2.24. win accuracy = 47.43% $500 a bet.
amazing! a raw, emotionless, non conversational algorithm is just doing its thing and ripping out those big upsets. ill gladly post the data soon. Note in contrast to several years ago, this approach would fail terribly since teams could actually hold leads back then. in the new age of drug enhanced sports and home run swings and bullpens that implode worse then a new Johnny Depp movie, it now pays a lot of money to bet against favorites. i love that folks on this board have invented their own heuristic to beat OLG badly, but I fear as well you wont be getting those prices for long. BCLC has wised up and soon they will to.
that is my two cents anyway, ill let the data speak for itself over time.
there are many many methods to connect historical data to break evens. in financial markets there are hundreds of HFT algorithms that compete against each other. only a few succeed. Using financial markets as an analogy, what we are concerned mostly with is having better arms then the other guys.
we don't really care about smart money or moving lines.. to me it seems beyond the point. just need a price that makes sense. We just gladly bet every single game (which most people say is retarded) but lo and behold after 2000 bets we are up 50k on an 100k bankroll. average odds bet = 2.24. win accuracy = 47.43% $500 a bet.
amazing! a raw, emotionless, non conversational algorithm is just doing its thing and ripping out those big upsets. ill gladly post the data soon. Note in contrast to several years ago, this approach would fail terribly since teams could actually hold leads back then. in the new age of drug enhanced sports and home run swings and bullpens that implode worse then a new Johnny Depp movie, it now pays a lot of money to bet against favorites. i love that folks on this board have invented their own heuristic to beat OLG badly, but I fear as well you wont be getting those prices for long. BCLC has wised up and soon they will to.
that is my two cents anyway, ill let the data speak for itself over time.
Re: Steamsharp Chronicles.
Well Said Steam!
I totally agree that the age of drug induced sports especially in MLB were pitching is the most impacted by those juiced up batters will significantly distort older data. I wonder if the Sossa/Mguire era saw same sorta correlations to MLB data the last few years?
I also dont doubt that your team is profitable because you told me u like betting with Matchbook and many times they offer even better odds then pinnacle.
Have u guys thought about opening multiple accounts at SIA? If your model favors alot of "Dogs" you be stupid to not try and take advantage of SIA's Lazy odds on MLB dogs LOL They many times have 10%-20% edges on MLB dogs. If pinny has +130 they sometimes hang over +150 on same game
I totally agree that the age of drug induced sports especially in MLB were pitching is the most impacted by those juiced up batters will significantly distort older data. I wonder if the Sossa/Mguire era saw same sorta correlations to MLB data the last few years?
I also dont doubt that your team is profitable because you told me u like betting with Matchbook and many times they offer even better odds then pinnacle.
Have u guys thought about opening multiple accounts at SIA? If your model favors alot of "Dogs" you be stupid to not try and take advantage of SIA's Lazy odds on MLB dogs LOL They many times have 10%-20% edges on MLB dogs. If pinny has +130 they sometimes hang over +150 on same game