Greyhound Racing New Zealand BOX DRAW REVIEW · 1.2 Processes of the Review 1.3 Basis of this report 1.4 Summary of Findings 1.4.1 Operational Processes 1.4.2 Technical Processes

Greyhound Racing New Zealand

BOX DRAW REVIEW

Prepared for: Greg Kerr, Head of Racing

Greyhound Racing New Zealand

Prepared by: Kim Denny

Gaming Consultancy and Training Solutions Limited.

March 2017

GRNZ Box Draw Review Page 2

CONTENTS

1. EXECUTIVE SUMMARY

1.1 Objectives and Scope

1.2 Processes of the Review

1.3 Basis of this report

1.4 Summary of Findings

1.4.1 Operational Processes

1.4.2 Technical Processes

1.4.3 Influence of Greyhounds

1.4.4 Influence of Trainer

1.4.5 Influence of Other

2. BOX DRAW REVIEW PROCESS

2.1 The Operational Box Draw Process

2.2 The Technical Box Draw Process

3. SUPPORTING INFORMATION

3.1 Technical Process Summary

3.2 Statistical Information

3.2.1.1 Winning percentages by box.

3.2.1.2 Placing percentages by box.

3.2.1.3 Winning percentages by box and track.

3.2.1.4 Placing percentages by box and track.

3.2.1.5 Multiple entries by Trainers.

3.2.1.6 Industry feedback.


1. EXECUTIVE SUMMARY

1.1 Objectives and Scope

Greyhound Racing New Zealand (GRNZ) requested a review of the integrity of the processes in place to select box draws

for greyhound races conducted for Greyhound meetings within New Zealand. The scope of this review was to determine that

there are adequate controls to ensure the integrity of the box draw process.

This was completed by;

Observation of the processes

Discussions with GRNZ management

Discussions with selected GRNZ Racing Club personnel

Discussions with selected Industry members

Reviewing the results of box draws over selected periods

Documentation of the results

1.2 Processes of the Review

This review was conducted primarily by observation of the processes and discussions with various key persons within the

Greyhound Racing Industry. The approach involved;

Understanding the process through discussions and observations with Matt Claridge and Roger Moore of GRNZ

management.

Contacting Industry related persons to invite feedback related to the Box Draw process.

Documenting and evaluating the processes.

Completing a review of the results of box draws over a period of > 5years.

Reviewing the processes to determine the effectiveness and integrity of the Box Draw


1.3 Basis of this Report

This report has been prepared in accordance with the objectives provided prior to the commencement as set out in

1.1 above.

1.4 Summary of Findings

Findings of this review are summarised on the next page.


1.4 FINDINGS & RECOMMENDATIONS

1.4.1 Operational Processes

There are no opportunities for the box draw process to be manipulated in any way.

Changes that are made are tracked and a report is generated. There is no review of why changes are made. I would

recommend a review system for this be introduced.

There are anomalies identified as to the reasons for redraws, this relates to the reasons at club level due to differing criteria.

Clubs operate “points systems” based on criteria such as how many nominations have been made before a race? How many

missed races? I did not obtain full lists from each club but there are definite inconsistencies between clubs.

This criterion (at club level) is then used to determine whether a dog gets into a field and on occasions this information is not

included when the field selection and subsequent box draw is undertaken. To avoid this confusion for all parties and reduce

the number of occasions redraws are required, I would recommend that GRNZ review this process with a view to

standardise the criteria at all tracks. This could also reduce the burden at club level by introducing this information as part of

the GRNZ “On Track” systems.

1.4.2 Technical Processes

Little technical information is available at GRNZ.

The basis of the program that generates the random number selection Microsoft Access is an extensively used program

using algorithm for random selection. Documentation of the Technical processes as provided by Sandfield is included as 3.1

in this report.

From reviewing this documentation and from Industry feedback there is a recurrent issue.


This relates to the Box Draw process that it only selects a box for each dog it does not randomly sort or select the list of

dogs.

In the technical information available covering the “Box Assignment Process” it states;

“Get a list of dogs to be placed into boxes (maximum of eight dogs). Note that no order is assigned to this list and the result

record set that we iterate through is provided by SQL server without any specific ordering. It is our understanding that in

these situations, SQL Server will use the primary key to apply a “de facto” sort on the records and in this case means that

essentially the records are delivered in the order they were created in the system”.

Based on the above, the randomness of the process is being questioned. The fact that some people’s perception is the field

(list) that is entered into database can be changed. There is also a perception that if a trainer has multiple entries in one race

then there dogs will be entered first so that their dogs are drawn first. The reasoning expressed during the course of this

review included statements like “if my dog is intentionally placed last on the list then all other dogs in the race will be

allocated a number before mine is and my dog will only be able to be allocated the last remaining number?” Questioning

where is the randomness of 1 to 8 inclusive if my dog is 8th on the list? Obviously this does not necessarily mean that the

allocated box will be pre-determined and can still be any number between 1 and 8.

I believe that the above was at least part of the reason that Greyhound racing Victoria (GRV) changed their box draw

system? The system now used by GRV randomly selects a dog then randomly selects an available box for that dog.

This eliminates any perception of whether a dog is being given equal opportunity and provides a more transparent process.

I recommend GRNZ evaluate the option of enhancing the current software to include random selection of a dog from the list

before randomly selecting an available box. I firmly believe this is the single biggest issue identified by persons involved in

the Industry.


1.4.3 Influence of Greyhounds

There is definitely no possibility of the Box Draw being influenced by the Greyhound information either as the registered

name of the greyhound or by ownership.

The process has no knowledge of the name of the dog or trainer and has no bearing on the order of the dogs as they are

assigned boxes.

1.4.4 Influence of Trainers

During this review there were several expression of concern made by Industry related persons regarding the sequence in

which nominations were presented for the box draw process? Particularly when a trainer had multiple runners in the same

event. This is addressed above and also in 3.2.5 showing the results of a trainer having multiple runners in a single event.

1.4.5 Influence of Other

While any software today could possibly be subjected to interference, I am confident that GRNZ have adequate safe guards

in place to prevent or at least detect any interference. Also the nature of the way system is used it has excellent internal

controls and It would be difficult make unauthorised changes.

For transparency I would recommend that wherever possible a redraw is carried out in the presence of a second GRNZ

person and appropriate documentation is maintained for review purposes.


2. BOX DRAW REVIEW PROCESS

2.1 Operational Box Draw Process

The events leading up to the actual box draw are as follows;

A scheduled Race Meeting is entered in GRNZ Access Database. This information is provided from the NZRB Racing

Calendar.

When the time of the first race is entered this automates the opening and closing date/time of nominations.

There are three methods of nominations;

by login to GRNZ website thedogs.co.nz

by phone to GRNZ

by fax to GRNZ

Nominations are processed by meeting, greyhound and preferred race.

There are normally two GRNZ staff allocated to these tasks however there are a number of GRNZ staff who can

perform these tasks.

Prior to nominations closing a recheck of the nominations is made to ensure correctness.

A check is made for each greyhound to make sure that its grade has not changed between nominations opening and

nominations closing. A report is generated which shows all of the nominations for the particular meeting in time order

of when they were nominated and by who.

The report also shows if there are any errors in grading (e.g. Greyhound X is a C4 dog and cannot be nominated for a

C5 event). This, or any other error detected can be corrected by contacting the club before the fields are assembled.


Nominations are then closed and are available to clubs on the website who can commence assembling the fields. The

club will follow the required selection criteria. This includes clubs individual criteria and the “Guarding Rule”.

The club then follows the selection criteria to determine the type of races that will be scheduled and field selection.

Selection is performed using a drop down – in which order they select is up to the club. This could be alphabetical or

otherwise.

GRNZ personnel now access the meeting in the database and the “Box Draw All Races” is now actioned.

The scheduled meeting complete with box draws is normally released on the GRNZ website thedogs.co.nz about

three hours after nominations close. These remain on line as the final field’s right up to the meeting date. This is to

allow time for queries regarding selection criteria. If errors or incorrect entries are identified all such cases are logged

and a full explanation of the reason is recorded, for example a club realises they have not considered the ‘guarding

rule’ and a redraw for particular race may be necessary.

Once confirmed a redraw is required GRNZ personnel can go into a race and select ‘redraw field’, the details are then

documented including reasons for the redraw. There is no record of any review process of this although I do believe

there should be to ensure the integrity of the process.

From the information that was available during the review it is apparent the main reasons for redraws being required

is due to clubs overlooking the individual clubs criteria or the ‘guarding rules’. Not every clubs rules are the same and

can lead to confusion. Whereas if standard rules were in place at all clubs this would eliminate confusion and allow

this information to be included as part of the GRNZ database providing a further measure to prevent greyhounds

being placed in an incorrect event.


2.2 Technical Box Draw Process

The software and applied functionality for GRNZ’s On Track System was provided and is supported by Sandfield a

software company based in Auckland.

A summary of the technical box draw was provided to Crowe Howarth in 2013 when a previous review was

undertaken. Discussions with Peter Ammon of Sandfields has provided confirmation that there have been no changes

made to the methodology of the Box Draw and the documentation provided by Sandfields in 2013 is still relevant.

The application is a Microsoft Access ADP and uses Microsoft Visual Basic to generate Random Number Generator

(RNG). The functionality of the On Track system which encompasses the Box Draw process applies a standard

Microsoft Random Number Generator to select box draws. Technically the selections are ‘pseudo-random’ as truly

random selection cannot be calculated using an algorithm.

As noted in the previously published review of this system, the random number generator has no record of the name

of the greyhound or the name or details of the trainer.

When a greyhound is selected, it is randomly allocated a box between 1 and 8. If a box has already been allocated

then it generates another number until a free box has been found. This process is completed without advice or

recording of how many sweeps are performed before a box is allocated.

A strong point of criticism from a number of Industry related people I spoke to during the course of this review was

that although the draw is selected randomly the entry of individual greyhounds is not. The list of dogs is determined

purely by the sequence in which the nominations are received. Therefore if a trainer were to nominate four

greyhounds for a same event they could appear on the list sequentially, for example the first four on the list. This

appears to be the biggest cause of concern to people spoken to, that their perception is that this process allows these

four greyhounds to receive box draws next to each other?


In comparison, Greyhound Racing Victoria Fast Track system works differently in that it firstly randomly selects a

greyhound from the list and then randomly selects an available box. Whereas GRNZ process only randomly selects a

box for each greyhound. This is a major point of contention with a number of people associated with Greyhound

Racing in New Zealand and this will be a recommendation in this report that considerations be given to changing this

process.

Even taking into account the same general principles are used for both systems the VRC system is arguably more

transparent.

Attached as 3.1 is a summary of the Technical Box Draw Process.


3. SUPPORTING INFORMATION

3.1 Technical Draw Process

The process below is a summary of the technical box draw process provided by the Stanfield Solution Manager.

Confirmation that there have been no software changes or enhancements made to this process since the previous review in

2013 was received from Wayne Hyman Solutions Manager Stanfield on March 29th 2017.

GRNZ BOX DRAWS

Pseudo Random Number Generator

The Race Day System application is a Microsoft Access 2007 ADP and therefore uses Microsoft Visual Basics for

Applications (VBA) pseudo Random Number Generator (RNG) function called “Rnd”. This function returns a random number

between 0 and 1 and this is converted to a decimal number to an integer between 1 and 8 inclusive.

Note that all algorithmic RNG’s are “pseudo” RNG’s as they use mathematical formula and algorithms to generate random

numbers. They are not truly random and they do not have a period of repeatability (although this period is typically an

enormous number before the pattern repeats). Truly random numbers can’t be calculated using an algorithm. An accepted

way to obtain truly random numbers is via the use of atmospheric static and generators exist on the internet that use such

approaches (for instance www.random.org).

RNG Seeding

It is common to “seed” a random number generator before first use. We seed the RNG using the “Randomize” VBA function

called without parameters. When this function is called without parameters, the current system time is used as the seed

value. It is our understanding from the Microsoft documentation on this function that the milliseconds component of the

current system time is used by “Randomize”.

http://www.random.org/


This means that even if the Randomize function was always called at the same general time of the day, we always get high

variance on the number used as the seed as it is the milliseconds component (0.000 and 0.999).

We have also conducted testing around one reseed and reseeding between each draw. As seeding a pseudo-RNG can

have a big impact on the random numbers that are generated, where and when the pseudo-RNG is seeded is vital to the

overall “randomness” of the numbers generated. Therefore we only seed the pseudo-RNG when the Race Day application is

opened and not before each box draw.

Box Assignment Process

In general terms the process is like this:

1. Get a list of dogs to be placed into boxes (maximum of eight dogs). Note that no order is assigned to this list and the

result record set that we iterate through is provided by SQL server without any specific ordering. It is our understanding

that in these situations, SQL Server will use the primary key to apply a “de facto” sort on the records and in this case

means that essentially the records are delivered in the order they were created in the system.

2. Take the first dog in the list and randomly pick a number between 1 and 8 inclusive. If the box is available (not assigned

to another dog and not to be left vacant if less than eight starters) then the dog is assigned to that box.

3. Continue to iterate through each dog in the list and randomly pick a box between 1 and 8 inclusive until a valid box is

found, then assign the dog to that box.

A key point to note is that the process has no knowledge of the name of the dog or the trainer and this has no bearing on the

order of the dogs as they are assigned to boxes.

We have also conducted testing whereby the list of dogs is also randomised before iterating through them and randomly

placing them into boxes. On large samples (10,000 box draws) we could detect no discernible difference in the

“randomness” by first randomising the list of dogs and the theory would suggest that doing so should not alter the

“randomness” of the draw.


Comparisons with Greyhound Racing Victoria (GRV) Box Draw Code

After reviewing the GRV code (specifically their code for non handicap races as they have specific logic to handle

handicaps) we found the same general principles were used. The following differences were noted:

1. GRV used the Microsoft.Net pseudo-RNG function in place of the Microsoft VBA pseudo-RNG function used by GRNZ.

2. GRV randomly selects a dog then randomly selects an available box whereas the GRNZ process just randomly selects a

box for each dog.

As noted above, we have conducted testing between a random dog list and an implicitly unordered (as provided by SQL

server) dog list and found no detectable differences in “randomness” of the box draw. However it must be pointed out that

this was a subjective assessment of the resulting box draw distributions and does not constitute a detailed mathematical

assessment. That said, logic would suggest that if selecting a box is “sufficiently random” then first randomly selecting a dog

would not make the process”more random” or any “less fair”.


3.2 Statistical Information

In order to quantify whether the actual draw provides an advantage for a dog statistical information from the GRNZ website

www.thedogs.co.nz was collated.

If there is a perception within the Industry that some trainers dogs are being advantaged by the draw they receive then surely the

statistical information available would reflect which draws are the most successful and also the information would reflect that some

trainers receive more favourable draws than others.

From this review and others completed previously both here and Australia shows conclusively that it is impossible to favour a dog

by trainer or any other information as this information is not included as part of the box draw process.

During this review confirmation was received from the Sandfield Solution Manager stating “A key point to note is that the process

has no knowledge of the name of the dog or the trainer and this has no bearing on the order of the dogs as they are assigned

boxes”.

The only other assumption that can be made, as mentioned previously and repeatedly in this report, is the only avenue not

controlled by the automated process is the sequence in which dogs are entered in the system before the box draw process is

requested.

3.2.1 Winning percentages by distance.

The information for this spreadsheet was obtained from the GRNZ website combining the percentages listed for individual

distances at every track.

This information highlights what percentage of wins has been achieved from all boxes showing a range of distances.

This was intended to highlight whether there is any consistency of a particular box(es) being favoured over particular distances at

all tracks.


3.2.2 Placing percentages by distance.

The information for this spreadsheet was obtained from the GRNZ website combining the percentages listed for individual

distances at every track.

This information highlights what percentage of placings has been achieved from all boxes showing a range of distances.

This was intended to highlight whether there is any consistency of a particular box(es) being favoured over particular distances at

all tracks.

3.2.3 Winning percentage by box draw.

The information for this spreadsheet was obtained from the GRNZ website combining the percentages listed for all distances at

each track.

This information highlights what percentage of wins has been achieved from all boxes regardless of the distance.

This was intended to highlight whether there is any consistency of a particular box(es) being favoured regardless of distance and

regardless of the track raced at.

3.2.4 Placing percentage by box draw.

The information for this spreadsheet was obtained from the GRNZ website combining the percentages listed for all distances at

each track.

This information highlights what percentage of placings has been achieved from all boxes regardless of the distance.

This was intended to highlight whether there is any consistency of a particular box(es) being favoured regardless of distance and

regardless of the track raced at.


3.2.5 Examples of trainer with multiple entries

One of the most frequent comments made by Industry related persons spoken to during this review was directed at the draws

received by trainers with large numbers of dogs in their kennels. There were two main comments, firstly that there is a perception

that these trainers receive more favourable draws and secondly when there are multiple numbers of dogs in an event with the same

trainer they are likely to be drawn next to each other?

I am not aware of whether an advantage is gained by having kennel mates drawn alongside each other and secondly if there were

any advantage how many dogs would need to be placed in boxes side by side to achieve this advantage.

The attached spreadsheet was completed by identifying races where a trainer had three of more dogs entered in the event and

collating the information highlighting what boxes were drawn and whether there was any consistency of whether there was any

basis to the perceptions of favouritism?

In total 54 races in a three month period were identified where the same trainer had three or more dogs in the same race.

The statistics show that there is spread of assignment to boxes varying from 9.5% being the lowest for Box 1 and Box 7 showing

16% as the highest. This would considered to be an expected result considering the randomness of the process and does not show

that this particular trainer received more favourable draws during this period over any other dog in each event.

Of the 54 races analysed there were 9 occasions where 3 or more dogs from this trainer were assigned boxes next to each other.

From the 54 events the dogs from this trainers kennels raced in events as follows;

On two occasions there were 6 starters from these kennels, in both races 4 dogs were drawn alongside.

On six occasions there were 5 starters from these kennels which in two races 4 dogs were drawn alongside.

On twenty occasions there 4 starters from these kennels which in two races 4 dogs were drawn alongside and in two races 3 dogs

were drawn alongside.

On twenty six occasions there were 3 starters from these kennels and on only one occasions were all 3 dogs drawn alongside.


3.2.6 Industry Feedback

Ideally I would have liked to have been able to have spoken with more people as part of the this review, however I do believe that

the people I did speak to provided a good mix from within the Industry both from their roles and their personal and professional

opinions as well as geographically. I spoke with Trainers and Club Officials in both the North and South Islands.

I am extremely grateful to each of them for their time and for expressing their opinions candidly.

There is no doubt that there is anybody who considers the present systems, including the Box draw process to be in any way to be

manipulated.

It is very evident however that a number of parties spoken with do believe that the current Box Draw process has one flaw in that

the perception is the listings of the dogs prior to the automated box draw could provide trainers with a preferred draw depending on

the placement of their dog on the list?

The general opinion is that this part of the process is completed by manual input and therefore could be altered at any stage,

maybe not with intent to manipulate but the draw could be affected by the placement on this list.

As it can be seen throughout this report there is concern by a number of people within the Industry in regard to this process and a

number of people made reference to the fact GRV have a system in place which as well randomly selecting a box it first randomly

selects a dog.

Having a system that randomly selects a dog before selecting a box draw may not make it any more random but it would eliminate

the perception that maybe the existing process could provide a bias towards dogs dependent on the positioning on the list.

A suggestion was made that GRNZ could provide more information and assurance in regard to the integrity of the box draw by

putting more information on the GRNZ website showing how the process works or even having a segment on “Dogzone” where the

presenter could be provided with information for the audience? Maybe even a representative from GRNZ could appear on Dogzone

and outline not only the Box Draw process but also the objective and findings of this review?

Greyhound Racing New Zealand BOX DRAW REVIEW · 1.2 Processes of the Review 1.3 Basis of this report 1.4 Summary of Findings 1.4.1 Operational Processes 1.4.2 Technical Processes

Documents