1 Introduction

This report will try to state and visualize the performance of our current model in from the month the Start of the season until the January the 12th. And how tuning/filtering the matches that we invest on for the Optimum Odds is key in obtaining profit with our sports betting model.

The model that we will use will be model_favud_6.

2 Reading the Data and General Performance

  • The data decembertest.rdsis a dataframe consisting of all the matches from the start of the season untill Decemberr 1st 2019 (First Quarter).
  • The data decembertest_2.rdsis a dataframe consisting of all the matches from December 1st 2019 until January 19th 2020 (Second Quarter).

2.1 Confusion Matrix

From our confusion matrix we can conclude that we have a general accuracy of 53.89%.

With a 52.95% win percentage from fav predicts, and 54.82% win percentage from our ud predicts.

#> Confusion Matrix and Statistics
#> 
#>           Reference
#> Prediction fav  ud
#>        fav 296 263
#>        ud  253 307
#>                                           
#>                Accuracy : 0.5389          
#>                  95% CI : (0.5092, 0.5684)
#>     No Information Rate : 0.5094          
#>     P-Value [Acc > NIR] : 0.02592         
#>                                           
#>                   Kappa : 0.0777          
#>                                           
#>  Mcnemar's Test P-Value : 0.69196         
#>                                           
#>             Sensitivity : 0.5392          
#>             Specificity : 0.5386          
#>          Pos Pred Value : 0.5295          
#>          Neg Pred Value : 0.5482          
#>              Prevalence : 0.4906          
#>          Detection Rate : 0.2645          
#>    Detection Prevalence : 0.4996          
#>       Balanced Accuracy : 0.5389          
#>                                           
#>        'Positive' Class : fav             
#>

Below is the table consisting all the matches available to bet on in the first quarter of the Season.

2.2 Bankroll Statement

With our 53.89% win rate, we are still recoding a loss of -61.31 units per the 12th of January.

The graph below shows the progress of our overall Bankroll/Financial Resource vs the time.

3 Problems and Solution

3.1 Problem

The problem,percentage of wins of our model, 53.89%, is still recording a loss over time in our bank loss. It seems as if we place all bets we are not obtaining the right value of the odds payout. To explain a little about value of odds, a simpler example is a coin toss. A coin toss provides a 50 50 chance for the outcome heads or tails. If we bet on heads, we should expect a breakeven payout at 2.00. This means everytime the outcome is Heads, we get 2.00 x bet unit. In the long run, statistically we will break even, no profit or loss.

But in the betting market, the house/bookie sets a commision for providing the bets. This usually ranges from 5-15%. This means, that the original payout of 2.00 becomes 0.95. Bookies also apply adjustments according to the supply and demand of their market. When the majority of his bettors places a bet on heads, he will have to lower the payout for heads and increase the tails payout in order to balance his banks and avoid losses.

So further on, the 0.95 can become 0.8 while the tails payout becomes 2.05. This is what we mean by value, in the long run, you will statistially record a profit because your chance is truly at 50% while your payout is at 2.05. Upon 100 coin tosses you will win 50 tosses, while obtaining a 2.5 units of profit.

To put it simply, what we are trying to accomplish is to see value in the market where everybody else doesn’t. In this case, the odds payout.

3.2 Solution

We optimize the odds for each possible type of bet, in order to maximize profit from the market.

3.2.1 UD home , fav A odds < opt_udhome2

#> [1] 3.575
#> [1] 121

#> 
#>         L         W 
#> 0.4725275 0.5274725

3.2.2 UD Home, favodds > opt_udhome

#> [1] 7.215
#> [1] 145
#> 
#>         L         W 
#> 0.2372881 0.7627119

3.2.3 Fav Home, favodds > opt_favhome

#> [1] 1.7
#> [1] 162

#> 
#> W 
#> 1

3.2.4 Fav Home, favodds < opt_favhome2

#> [1] 0.085
#> [1] 11

#> 
#> W 
#> 1

3.2.5 UD away, favodds < opt_udaway2

#> [1] 0.07736486
#> [1] 81
#> 
#>   L   W 
#> 0.5 0.5

3.2.6 UD away, favodds > opt_udaway

#> [1] 0.45
#> [1] 175
#> 
#> W 
#> 1

3.2.7 Fav Away, favodds < opt_favaway2

#> [1] 0.14
#> [1] 176
#> 
#>         L         W 
#> 0.4150943 0.5849057

3.2.8 Fav away > opt_favaway

#> [1] 0.8333333
#> [1] 30
#> numeric(0)

4 Overall Return After Optimization

#> 
#>         L         W 
#> 0.4107143 0.5892857

#> [1] 0.09727679

5 We test these parameters to the second quarter of the season

5.1 Reading Data

#> Confusion Matrix and Statistics
#> 
#>           Reference
#> Prediction fav  ud
#>        fav 202 157
#>        ud  127 167
#>                                           
#>                Accuracy : 0.5651          
#>                  95% CI : (0.5261, 0.6035)
#>     No Information Rate : 0.5038          
#>     P-Value [Acc > NIR] : 0.0009812       
#>                                           
#>                   Kappa : 0.1295          
#>                                           
#>  Mcnemar's Test P-Value : 0.0852811       
#>                                           
#>             Sensitivity : 0.6140          
#>             Specificity : 0.5154          
#>          Pos Pred Value : 0.5627          
#>          Neg Pred Value : 0.5680          
#>              Prevalence : 0.5038          
#>          Detection Rate : 0.3093          
#>    Detection Prevalence : 0.5498          
#>       Balanced Accuracy : 0.5647          
#>                                           
#>        'Positive' Class : fav             
#>

5.2 Bankroll Statement Before Optimized Parameters

With our 56.51% win rate, we are still recording a profit of 3.635 units per the 19th of January. This is without filtering the matches to our optimum value parameters. We hope with applying the odds parameters, we can hopefully reach a higher profit.

5.3 Applying our Optimum Value Parameters

5.3.1 UD home , fav odds < 2.21

#> 
#>         L         W 
#> 0.4464286 0.5535714

5.3.2 Fav Home, favodds > 2.62

#> 
#> L 
#> 1

5.3.3 UD away, favodds < 1.81

#> 
#>         L         W 
#> 0.4324324 0.5675676

5.3.4 Fav Away, favodds > 1.3

#> 
#>         L         W 
#> 0.5063291 0.4936709

6 Overall Return After Applying Optimum Parameters

#> 
#>         L         W 
#> 0.4739884 0.5260116

#> [1] 0.1324566

6.1 By Week

Our return on investment if we decide to take a higher risk and divide our bets to the number of games in each week accordingly. Achieving an active rollover of our whole bankroll every week.

7 Conclusion

We conclude that a higher winrate will not guarantee you a profit from the betting market. Filtering and finding value in the odds/payout is the essential part in this business.

As you can see, applying our optimum odds value parameters yields greater profits in the second quarter of the season.From the Initial 3.635 units to 22.595 units.