ACC30003 Investigation into AFL Betting PDF

Preview

Summary

forensic accounting swinburne...

Description

(Sports Power 2019)

Investigation into AFL Betting ACC30003 Forensic Accounting Assignment 3: Team Assignment

Sarah Whalen 102127480 Sophie Storm 102321868 WORD COUNT: 3296

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Contents Introduction...................................................................................................................1 Main Report..................................................................................................................2 Error 1: Missing Data.................................................................................................2 Error 2: Duplicate Data..............................................................................................9 Error 3: Incorrect Spelling of Venue Name.............................................................14 Error 4: Venue Name Change.................................................................................16 Error 5: Incorrect Team Name.................................................................................16 Conclusion..................................................................................................................19 Reference List.............................................................................................................21

1

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Introduction The purpose of this report is to investigate the accused ringleader of a suspected Australian Football League (AFL) betting fraud. By comparing the data between 1987 and 2017, obtained by the Victorian Police and the AFL Big Lists website, will highlight any discrepancies between the two datasets. The structure of the report will follow the steps taken in the investigation, beginning with the preliminary measures taken and further presenting the errors in the order they were found. Excel and data mining techniques such as: COUNTA, Pivot Tables, IF, EXACT, and VLOOKUP, will be utilised consistently throughout the investigation to identify any errors. Following identification, there will be discussion on the methods used to uncover the errors, why the errors have occurred, what fraud symptoms are present, and if fraudulent behaviour has occurred. Considerations in the investigation as to whether the errors were caused when importing, or indication of intentional concealment will also be discussed. Finally, internal control weaknesses will be analysed with recommendations as to how to rectify this. Throughout this report, the dataset obtained from the Victorian Police will be referred to as ‘fraudulent dataset; and the dataset from the AFL Big List website will be referred to as ‘actual dataset’.

Main Report Error 1: Missing Data Prior to investigation, some preliminary measures were taken such as creating copies of both datasets so comparison can be completed without impacting the investigation. Importing the datasets side-by-side for comparison in a new document provides a visual observation of the data and upon doing so; the first error was discovered. The actual dataset was larger than the fraudulent dataset, revealing a clear omission of cells. To ensure efficiency throughout investigating the data, the headings of the both datasets were frozen. This was done by clicking on the view ribbon and then clicking the freeze panes button allowing the heading to stay up the top whilst scrolling through the dataset. To identify the number of cells missing the COUNTA function was utilised by entering the range of cells for each dataset. This function was used as opposed to COUNT, 2

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. as COUNTA prevented any inconsistencies with formatting as it counts all populated cells. Figures 1 and 2 show the formula used in the blue rectangle and the results in red, which revealed the actual dataset’s total of 15200 rows and the fraudulent dataset’s total of 15302. Making it clear there is an omission of 168 cells.

Figure 1 – COUNTA Formula Counting Actual Dataset.

3

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Figure 2 – COUNTA Formula Counting Fraudulent Dataset.

To determine what data is missing, a systematic approach of analysing the date column first was used. Ensuring a careful investigation, the datasets were both split using the text-to-columns function to separate day, month and year. After clicking the text-to-columns function, a window opens, and the delimited option was selected. Upon doing this, it uncovered that the fraudulent datasets date’s format changes halfway through. To combat this, the text-to-column function must be performed twice; once with a dash selected in the textbox and again with a forward slash. Figure 3 and 4 display the before and after results of using this function, with the red rectangle showing the cells affected and the blue highlighting the function button. Since the date format changed from DD-MMM-YYYY to DD/M/YYYY, the formula TEXT(CELL*29,”mmm”) was used. This formula changes the way a number appears, and; in this case, the number needed to be converted to the name of the month to replicate the actual dataset’s formatting.

Figure 3 - Before Using text-to-column Function.

3,

Figure 4 - After Using text-to-column Function.

5

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. A pivot table proved to be the most efficient way to find where the missing data was, as it is easy to operate the table to provide results for certain criterion. Two pivot tables were created for each dataset by selecting any cell in one dataset, clicking the insert ribbon and then pivot table button. In the field list the round was inserted to the value section and the year was added to the rows section, revealing the number of games per year. In figure 5, the blue rectangle highlights this process and the red reveals the columns that the pivot table created. Choosing to investigate by year allows for a more efficient way to examine rows whilst also providing a count of all the rounds for the year highlighting clear discrepancies.

Figure 5 – Entering Fields into Actual Dataset Pivot Table.

Once the pivot table was completed for both datasets, the results were copied and pasted next to each other for comparison. Again, the comparison lead to visual observation was that the table did not match, unearthing the omission of an entire year. To investigate further, a separate table was made, and the IF function was utilised to test whether the cell for the actual dataset’s year matched the cell for the fraudulent dataset’s year and additionally for both the round’s cells. If it did match, then it would populate the cell with a dash and if not, it would populate with ‘NO 6

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. MATCH’. In figure 6, the blue rectangle indicates the formula used, the purple displays where the discrepancy was, and the red indicates the difference between the two datasets results, ultimately underlining that the data from 1991 is missing from the fraudulent dataset.

Figure 6 – Comparing Pivot Table Results.

Since this error was an entire year and not a random date, this missing data could have been deleted through human error and without fraudulent intent. The change of date format could, however, be evidence of potential attempt to conceal the 1991 omission. Due to the size of the dataset and no segregation of each year, it is easy to miss an accidental omission; however, also provides an opportunity to conceal fraudulent behaviour. The error highlights an internal control weakness as human error of this size can affect the betting odds by skewing them. An internal control of independent checks with excel functions, such as COUNTA, can ensure data is correct. Whilst segregation of duties and having another person to sign off on the data import, can improve the integrity of the betting odds and eliminating the 7

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. opportunity for fraud (Albrecht et al. 2019, p. 112). Additionally, another internal control of only allowing authorized personnel to adjust historical data or override data will limit who has opportunity to alter data (Albrecht et al. 2019, p. 111). Earlier a difference of 168 cells was discovered. As seen in figure 6, the number of rounds for the missing 1991 data is 172, therefore revealing four rows of duplicated or added data elsewhere. This consequently prompted further investigation leading to the next errors discovery.

Error 2: Duplicate Data Due to the omission of 1991 entirely, the IF function displayed multiple cells with ‘NO MATCH’ as seen in Figure 6. To ensure a thorough investigation into the four extra data cells, the fraudulent data results must be moved in line with the actual dataset to account for the omission and highlight the discrepancy location. This process is shown in figure 7 with the red rectangle indicating the data to be moved. Figure 8 displays the result of the move revealing 2015 to be the year with four extra cells, 210 rounds as opposed to 206 in the actual dataset.

8

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Figure 7 – Moving the Datasets in-line.

9

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Figure 8 – Comparing Pivot Table Results for Year.

To further investigate where the extra data is, the months field was added under the rows section in the pivot table field list. This was completed for both the actual and fraudulent datasets, and then compared against one another. Since it is known that the data extra data is in 2015, the filter option was used to show only the 2015 results. Figure 9’s blue rectangle displays the month field being selected to appear as a row underneath the year and the red displays that utilising a filter provides means to concentrate on 2015. Figure 10 presents the pivot table results for each dataset copied side-by-side with a cross check column for the IF function. This 10

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. exposed that in October of 2015 the fraudulent data had four more entries than the actual data within the same month.

Figure 9 – Adding Months to Pivot Table.

Figure 10 –Pivot Table Results Reveals Month Error.

11

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. To investigate what the extra cells were, a filter was again applied on the fraudulent dataset only. In figure 11, the blue rectangle indicates the function that was used, and the purple indicates the year column has the 2015 filter applied and the month column has the October filter applied. The red reveals the extra cells were duplicate entries of the 2015 Grand Final. Since in October the actual dataset has only one round, the error could have been a system error that resulted in the round of October to duplicate in the fraudulent dataset. The only suspicion is that it is repeated four times. The error was able to manifest itself and occur due to the size of the data being present and the duplication being hidden. This highlights an internal control weakness, the use of independent checks with the COUNTA function could detect fraudulent behaviour, eliminating fraud opportunity. Similarly, to the first error, segregating duties and having authorized personnel to sign off on data once input and eliminating ability to override data unless person of authority signing off would provide greater security (Albrecht et al. 2019, p. 111).

Figure 11 – Filter reveals Duplicate Errors.

Now that the discrepancy of the 168 rows has been accounted for there should not be any more missing or duplicated data. Prompting the question of the integrity of the remaining data, since there has been missing or duplicated what else could have been modified?

12

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Error 3: Incorrect Spelling of Venue Name Since the fraudulent dataset’s integrity was in question, another question surfaced as to whether the data had been intentionally changed in other sections. Beginning at the last column ‘venue’ to be and following a systematic approach of column by column, the EXACT function was utilised to present a comparison of venue names across both datasets. The function works by searching for a specific text to be located in another cell’s position. If it can be located it will produce an output of ‘TRUE’, however if the text cannot be found it will populate the word ‘FALSE’. Due to the missing data, to decrease the number of false results, the fraudulent data starting from 1992 was again moved to be in line with the actual dataset to account for the missing data. Compensation for error two’s duplicated data must also be made by removing the duplications in October 2015. This EXACT formula is made of two components: 1. Text 1 – Where should the text be located? (Venue in fraudulent data/column R) 2. Text 2 – What text is it looking for? (Venue in actual data/column I) By adding a filter, see figure 13, a reduction can be made in the number of rows shown by hiding all the cells that match correctly thus leaving the compromised data cells (figure 14).

Figure 12 – EXACT Formula Comparing Venues.

13

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19.

Figure 13 – Filter to Highlight Errors.

Figure 14 – Error 3: Venue Name Change.

A clear reveal of the incorrect spelling of the venue name can be seen in Line 265, whereby Corio Oval has been changed to ‘Corioioioio Oval’. Corio oval was Geelong’s home ground in 1900 and altering this data can have a severe effect on their home ground statistics (Ryall, p. 176). Whilst proof of fraudulent behaviour cannot be stated for certain over human error, line 265 ‘Corioioioio Oval’ required a further six keystrokes to be pressed to create the error. Human error cannot be wholly ruled out however, due to the letters ‘i’ and ‘o’ being next to one another on a qwerty keyboard. A crosscheck of number of games per venue could have located

14

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. this error that was able to go unnoticed. Additionally, removing overriding controls once data is finalised could prohibit fraudulent data being entered. The methods undertaken to discover error three, also revealed error four which was a venue name change prompting further investigation.

Error 4: Venue Name Change Since discovering the incorrect spelling of a venue within error three, an additional error in line 14432 found Manuka Oval altered to ‘Manuka Honey’. A change in venue names highlights evident internal control weakness with data being altered to potentially heighten gambling odds for specific venues for teams. Similarly, Manuka Oval is the home ground of GW Sydney and altering the Venue Name can then have an impact on future premiership odds. Again, human error cannot be entirely discounted due to ‘Manuka Honey’ being a well-known type of honey created in New Zealand and Australia. Internal controls are weak within this area and errors can manifest due to venue’s being entered from scratch, if all venues were pre-loaded to be chosen from a list this would inhibit venue mistakes from occurring. Furthermore, to prevent future data from being altered overriding controls must be put into place. The venue name errors prompted an investigation as to whether perhaps other columns had also been altered to augment betting odds. Continuing the systematic approach of working column by column (to ensure data was not missed) located no errors within the scores. The final column to be analysed was team names.

Error 5: Incorrect Team Name The incorrect spelling of venue names prompted an investigation as to whether any of the teams at play had been altered. A similar approach to error three is followed to check if any of the team names had been misrepresented. The easiest way to investigate the team names was to utilise the VLOOKUP formula and compare between both datasets. Again, removing/rectifying errors previously found, the VLOOKUP formula looks for a specific value within a range of data and if found returns the value contained in a designated column. If the specified value is not found the formula will return ‘N/A’ indicating an anomaly. The formula is made up of four provisions: 15

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. 1. Lookup Value – Select what data to look for in the fraudulent dataset (Team 2 in the actual data/Column G) 2. Table Array – Where to search for the data (Team 2 in the fraudulent data/Column P) 3. Column Index Number – Specifies the column in the table array to return (column 1) 4. Range Lookup – This indicates as to whether to search for an approximate match or an exact match (exact/false) The formula must then be applied to every row of data to show either a team name or ‘N/A’, and a filter is applied so that only ‘N/A’ results are shown.

Figure 15 – Filter to Highlight Errors.

Within figure 16 and 17, the blue rectangle highlights the formula used and the red displays three rows with an ‘N/A’ anomaly. The figures reveal the first two rows in the fraudulent dataset is identical to the actual dataset only the two rows are swapped 16

102127480 Sarah Whalen, 102321868 Sophie Storm, Assignment 3: ACC30003, Investigation into AFL Betting Fraud, 16/09/19. (rows 4174 and 4175); this signifies that although there has not been a manual change in team names there has been an issue when importing. This will not affect betting odds and therefore there is no intention to commit fraud however it does highlight the need to segregate duties and have a person of authorization confirm accuracy of the data being inputted. This may be confirmation of data being altered deliberately with this trail of evidence remaining.

Figure 16 – Actual Dataset Rows for Comparison.

Figure 17 – Fraudulent Dataset showing Errors.

Moving on to the last row the formula also uncovered, which is likely to be an importing mistake. Shown in figure 17, the purple rectangle reveals row 8475 team two has been misrepresented. The actual dataset shows Melbourne vs. Carlton; however, the fraudulent data has been changed to represent Melbourne vs. Hawthorn. Whilst there is certainly a chance the data could have been changed by mistake; the score provides insight into a reason as to why someone might intentionally alter which team played. Melbourne kicked nine goals worth six points each and seven behinds equating to one point each making their total 61 points, however, Carlton kicked 19 goals with 26 behinds resulting in a total of 140 points. Not only is there incentive to change which team played by giving them an extra four points on the ladder and improving their statistics, there is also a huge incentive to change the team name based on the pe...