The following are the rules of proline I would like to highlight:
- proline requires you to pick three events
- the lowest possible payout for a single event being 1
- the lowest possible payout for a card being 3
If we apply Poker Pot Odds to these payouts and odd, we must be able to pick the outcome of a card at least 1/3 of the time to be profitable. Considering each game to be an independant event we must be able to predict each game with a certianty of only 69.33% to break even in the worst case scenario.
All charts were last updated on most recent commit titled "more data"
The following is a distribution of game win percentages across all odds differences where the lower payout won. what is the odds differece? it is the differnce in payout between the two opposite game outcomes (ie. H and V) consider the example below:
H+ H T V V+
2.1 1.6 5.0 1.3 1.8 (odds difference = 0.3)
It is also important to account for the number of events observed when considering the results above. As you can see move events of lower odd difference did occur, which is important to consider when reading the percentages shown above
The permilinary findings show that events of higher odd difference show the lower odds win more often.
NOTE: All the findings above include all proline sports, except soccer this is due to the rule that if a soccer game is goes into overtime it is only payed out as a tie. While other sports payout the tie as well as payout for the winning team
With MLB season warming up, I will be collecting short and long term season data and ploting its relationship to proline winnings in hopes of finding a correlation. I want to focus more on MLB because the seasons have more games (ie. moarrr data)
Stay tuned for more events containing the following json data
{
<ticketID> : {
...
"mlb_standings": {
"home": {
"rank": <league ranking>,
"points_scored_per_game": <average points scored per game>,
"points_allowed_per_game": <average points allowed per game>,
"win_percentage": <percentage of games won>,
"location_win_percentage": <percentage of home games won>,
"last_five_won_percentage": <percentage won in last 5>,
"last_ten_won_percentage": <percentage won in last 10>,
"streak": <current consecutive wins or losses (+ve for wins -ve for losses)>
},
"visitor": {
"rank": <league ranking>,
"points_scored_per_game": <average points scored per game>,
"points_allowed_per_game": <average points allowed per game>,
"win_percentage": <percentage of games won>,
"location_win_percentage": <percentage of away games won>,
"last_five_won_percentage": <percentage won in last 5>,
"last_ten_won_percentage": <percentage won in last 10>,
"streak": <current consecutive wins or losses (+ve for wins -ve for losses)>
}
}
}
Proline is a sportsbetting game in Ontario Canada. The Proline Project is a data study to find sucessful stratagies for the proline game. The project is broken down into several stages:
- collect Data
- determine Sucess Probablility distribution over payouts
- determine Sucess Probability distribution over
- monitor convergence of data set
- Simulate stratagies over historical data set
The update.py script hits the following two OLG endpoints to fetch payouts and outcomes:
payouts -> proline.ca/olg-proline-services/rest/api/proline/events/all.jsonp?callback=_jqjsp&timenow
outcomes -> proline.ca/olg-proline-services/rest/api/proline/results/all.jsonp?callback=_jqjsp&timenow
This data is collected daily and stored in the following json format:
{
<ticketID> : {
"date": <epochDate>,
"games": {
<gameID>: {
"cutoffDate", <last day to wager>,
"sport": <sportName>,
"visitor": <visitor team>,
"home": <home team>,
"h+": <home shutout payout>,
"h": <home payout>,
"t": <tie payout>,
"v": <visitor payout>,
"v+": <visitor shutout payout>,
"outcome": [<list of payouts won>]
},
...
}
It is still to early to apply any ML techniques to this data till strong enough correlations are seen across a large enough dataset, and an informed decision can be made on the inputs. But the long term objective is to combine the proline odds with basic league standings from the short and long term for a model