Thread by @nickwallis

Morning. We are in court 26 of the High Court’s Rolls Building for Day 18 of the Horizon trial, part of the Bates v Post Office group litigation. Follow here for live tweets…
#postofficetrial

Dr Robert Worden, the Post Office’s independent IT expert has been sworn in. His surname is pronounced as in “word”.

The word “word”, not as in “bored”…

anyway…. usual disclaimer…

These tweets are summaries and descriptions of what is being going on and being said in court. Nothing is a direct quote unless in “direct quotes”.

If you are new to all this, you can have a look at what happened last week here: postofficetrial.com/2019/06/horizo… and click around the website from there.

There’s a lot of faffing at the moment as Dr Worden (DW) is being taken through his statements by Mr de Garr Robinson and confirming all the corrections etc.

I hope to get hold of his reports and joint statements with the claimants IT expert asap.

Patrick Green, QC for the claimants, is on his feet. We’ll call him PG.

DW has a PhD in theoretical particle physics and has spent more than 10 years being an expert witness.

He has a very academic mien.

PG asks him about his statistical qualifications.
DW says he has none, but knows a lot about maths, engineering and therefore statistical techniques.

PG statistics can be about probability and maths is about getting it write or wrong
DW agrees
PG and you are thoroughly comfortable using statistical analysis
DW yes, I did all the maths in this - it’s basically long multiplication and division - it’s not that advanced.

PG takes him to one of his corrections in his report - "minor corrections and clarifications”
PG ”chances of a bug occurring in a claimants’ branch is 2 in 10,000,000 and in fact it is 2 in 1,000,000” so out by a factor of 10.
DW yes. That needed to be changed.

PG have you been involved in group litigation before
DW no this is a new experience for me
PG you’re not a behavioural economist
DW no my knowledge of economics is shallow - I know about business finance

PG takes him to a statement about estimating the value at which a Subpostmaster should call the Post Office when they have a discrepancy
DW yes
PG you know the difference between an estimate and an assumption
DW oh yes
PG pretty fundamental, right
DW have I used loose...

… language here. They’re assumptions.
PG well this is what I’m trying to ask you…
PG well here you say estimates and there you say assumptions
PG did you make your own free-standing estimate at the level of discrepancy at which a Subpostmaster would contact the Post Office?

DW well I made assumptions that became estimates
PG did you make an estimate that confirmed Angela van den Bogerd’s estimates?
DW well you can make ballpark assumptions or estimates and you get to the result and then you look back at what the assumptions or estimates are...

… to see if they need refining.
PG but the question for the moment… you made assumptions and you set them out, but you didn’t make any fact-based estimate of what would happen.
DW no. they were assumptions. and if the court wishes to change those assumptions then the...

… court can drive them through my calculations.
PG to make a fact-based estimate you would have to know how and why people faced things
DW to get precise numbers, yes
PG to make an estimate
DW I made some assumptions, I put them to some maths and they became estimates

PG so you have nearly the facts or the expertise to make any assumptions which fall into the field of behavioural economics
DW I accept that entirely.
PG so you have a PhD in theoretical physics, you are an expert witness, you’ve made detailed statistical analysis
DW not detailed

PG and you made a mistake by the order of magnitude of 10 and you corrected it
DW yes
PG you are a details man
DW physics people don’t tend to like facts, but they like to get into theory and getting that right, so from that perspective I am a details man yes.

PG we are now going in DW’s colleague, who has a degree from Imperial College. they worked together at Logica in 1976.
PG was Credence a Logica solution
DW possibly. I was nothing to do with it. I’ve been involved in data warehousing since the year dot.
PG and data...

… warehousing was big in the late 90s.
DW and has been since then
PG and your colleague is a details man
DW yes
PG but he didn’t check the maths in this
DW no
PG but the rest of it?
DW yes

PG how many times have you read your report?
DW 5 times
PG lots of people have checked it?
DW well I have, Mr Emery (his colleague from Logica) has and I suspect the lawyers have
PG takes him to paragraph 650 in one of his reports about the receipts and payments mismatch bug.

In his report DW notes this bug had an effect on branch accounts, and if Mr Godeseth’s evidence about the bug is not accepted he will revise his opinions accodingly.
PG you have premised the opinion you give on Mr Godeseth’s evidence. So the content of your report focuses on...

… what your opinion will be dependent on the defendant’s evidence.
DW yes
PG you know that the only time we find out if the defendant’s evidence is accepted is when the judgement is handed down
DW good point
PG so you’re assuming the defendant’s evidence is true?

DW yes
PG but here [they go to a different part of the report] in your report you say you have not assumed that any version of events is true

PG so all you say is all you can do is make assumptions based on the evidence you have seen
DW yes
PG do you feel you have made opinions based on the understanding that the defendants' evidence is accepted equal to that of the claimants?

DW well I didn’t feel able to draw much from the claimants experiences [waffles a bit] so I provided my calculations, so if my assumptions need to be adjusted the court can do so. sorry I’m not sure I’m answering your question.
PG let’s go to another example...

PG here you say you proceed on the assumption these figures, the only ones available to me, are accepted by the court. and that if new information comes to light you will have to change it.
DW yes
PG having heard anything in the defendant’s evidence Post Office evidence...

… that makes you want a change of heart about?
DW nothing fundamental, but to come back to the point about precision - my numerical arguments have to be very precise in order to help the court.
PG let’s have a look at some facets of your evidence...

[we go to par 1086]
[it is about POLmis and Credence - two different names for the same app at different times]
[PG is asking about PO evidence that Credence records everything that happens in a branch]
PG was your understanding that Credence records all keystrokes in a branch...

DW I wouldn’t go as far as that, until I read that PO evidence.
That’s a very detailed level of information.
PG Knowing if you can record every keystroke in a branch is important, isn’t it?
DW no - I don’t need to know it goes down to the keystroke level
PG so you made a...

… judgment that keystroke level investigation was unnecessary.
DW yes - MIS (management information systems) takes slices of information and I decided it was enough.

PG [takes him to Horizon Issue 8 about information available to Post Office when investigating what had happened in a branch]. And you decided it was not important to investigate the keystrokes.
DW yes
PG but you accepted evidence it did
DW not quite
PG you said consistent...

with my understanding. When we read “consistent with my understanding” throughout your report what should we take that to mean?
DW in the ordinary meaning of the word consistent
PG okay so wrt to keystrokes, let’s have a look at the transcript of Mrs van den Bogerd’s evidence...

[PG goes to the xe of AvdB where she confirms Credence confirms the exact keystrokes]
PG were you here for mr Patny’s evidence where he didn’t get a Credence report despite being promised one?
DW no.
PG let’s see what AvdB has to say about that shall we?

PG this is Mr de Garr Robinson re-examining AvdB… DG asks AvdB it is the case that Credence logs every keystroke. AvdB says it doesn’t.
PG did you see that?
DW no i missed that.

PG we see later on that Credence doesn’t record every keystroke
DW right - so this is now more consistent than what i know
PG were you then guessing what an MIS is?
DW no my understanding of POL’s MIS is consistent with my experience

PG don’t you think the issue was important to be precise about?
DW youhave to make difficult editorial decisions about what you leave in and leave out of a report
PG why, in response to H8 didn’t you just list exactly what tools the PO have available to them to find out what’s...

… going on in a branch.
DW maybe I should have

[we go to the issue of using a branch user number higher than 32 during remote transaction so SPMR would know what was going on - it turned out that SSC at Fujitsu could go in with a branch number of less than 32]

[PG is taking DW to his report where he concludes it was not possible to go in with a branch number of less than 32]
PG you say [injecting transactions] could not be done without the SPMRs knowledge.
DW yes
PG so you’ve looked at a disputed issue of fact, you’ve accepted the

… defendant’s evidence. you say you’ve established it’s correct and then given your opinion that Mr Roll was wrong. And you shouldn’t have?
DW I accept that was wrong. I shouldn’t have said established ws too strong.
PG well - you’ve accepted the PO’s evidence and said you’ve...

… established it’s correct and you did nothing of the sort…
DW yes - “establish” was too strong.
PG well you’ve said in your report you don’t give any weighting to evidence from either party. Yet here you did. And it was wrong.
DW I accept that is a mistake
PG you based...

… your opinion on a hotly disputed issue of fact, preferring the PO’s evidence to that of the claimant witness.
DW in this case I was wrong
PG no this is in issue your approach was wrong. you rejected the claimants evidence and accepted the defendant’s and a) you shouldn’t have

and b) you were wrong.
DW accepts
PG have you accepted the claimants’ evidence anywhere in your report?

DW I’ve always tried to qualify my approach throughout my report by trying not to accept disputed evidence or give too much weighting to one side over the other on issues of fact
PG answer the question. on this instance your approach was wrong and your conclusion was wrong. yes?

DW accepts.

[we have a break]
[speculation klaxon: PG is going for the **** in **** out approach to DW’s analysis. He’s pointed out one obviously flawed assumption and one obviously flawed conclusion. inference being the maths might be rigourous, but the ...

… incorrect or flawed assumptions going in makes it meaningless]
[well that’s one hour of three days done. let’s see what’s next]

[we’re back]
PG raising a point on remote access by Richard Roll who used to work at SSC for Fujitsu and featured in the Panorama. You said it could not be done. You refused to accept Mr Roll’s evidence in your first report.
PG you rejected it on the basis of the defendant’s...

… evidence that the SPMR would notice because the counter number is higher than 32.
PG then the Post Office accepted Mr Roll’s evidence, and you then said in your second report having spoken to colleagues who confirm it is possible, you accept Mr Roll’s evidence it could be...

… done without the SPMRs’ knoweldge.
PG so you are unquestioningly accepting the defendants’ version of events.

DW demurs.

[so we go through the process again to have a closer look at the documents in question.]

DW’s new gambit - “it depends what you mean by it”
PG Dr Worden…
DW well we agreed in the joint statement that Fujitsu could more or less do anything…
PG let’s go back to your approach - Mr Roll’s 2nd statement and Mr Parker’s 2nd statement (Parker was a Horizon expert)

… and this is your 2nd report “in par 20 Mr Roll addresses a factual point” and that was a factual point you knew to be in dispute
DW two factual points - whether they could do it and whether they could do it without knowledge.
PG and you know this is about the latter point

DW well this about the counter 32 issue, which in my opinion was not the only way.
PG right… let’s take this slowly.
PG the witness statements are saying there is a way that is counter 32 and way that is not and it’s whether or not visible to the SPMR.
DW Mr Roll specifically...

… says that, does he?
PG let’s go back one last time then I think we’re going to have to move on.
PG I am trying to take you to paragraphs 82 to 85 and we never get past paragraph 82.
DW can I have a look at my supplemental report?
PG of course [possibly through gritted teeth]

DW got it
PG right. we were looking at the foot of p20, Mr Roll address factual point about injecting transactions…
PG so a factual witness who worked at Fujitsu was saying it was possible without the knowledge of the SPMR and you knew by the date of this report that Mr Parker..

… had agreed.
DW no Mr Parker had agreed it could be done, but not without the knowledge of the Subpostmaster.
J intervenes - does your opinion differ from Mr Roll's
DW not necessarily
PG you are reading Mr Roll as saying …
DW I’m not reading him - I’m saying I want to know...

…. more.
DW I think both experts agree that very little is impossible, but I don’t think it is possible to know whether or not a SPMR knew about a transaction injection
PG even with the evidence of your own eyes?
DW Mr Roll and Mr Parker certainly agree it is possible...

… but I don’t know enough about it to form a conclusion.
PG so you accept Mr Godeseth’s evidence which supports the Post Office and it is wrong. Yet when Mr Parker gives evidence which supports the claimants and is right you won’t draw a conclusion on it.
DW we don’t know...

… it’s right.
PG that’s a bizarre position to take.
DW I disagree.

[we move on]
PG let’s have a look at your second expert report…
“Mr Roll’s evidence doesn’t cause me to alter my conclusion”
DW yes
PG the one we now agree is both wrong in approach and in conclusion
DW we agree now.

PG you are inexplicably favouring the Post Office.
DW not inexplicably. I didn’t have enough information to make the right call
PG but that is strikingly different to your approach with Mr Godeseth where you just accept his evidence unquestioningly

DW that’s for the court.

PG so let’s go back to the lack of information. Did you pursue it?
DW what do you mean?
PG did you ask for for more information from the claimants? from Fujitsu? or seek directions from the court, which you’re entitled to do?
DW no - let me explain my understanding...

… about the permission to login issue [he does so]
PG did you understand this entire exchange about someone from Fujitsu turning up at the branch and asking someone to log in?
DW no I didn’t understand the entire paragraph of that so I couldn’t draw a conclusion

PG so what steps did you take to clarify a disputed Horizon Issue
DW not beyond this report. it was my feeling it was a difficult area - there levels of depth and complexity in the way H which the experts have not been able to plumb. There’s a way a transaction might have been...

… altered. There’s a whole swamp of difficult questions and I wasn;t going to make progress in that area.
PG even though you concluded in your first report Mr ROll was wrong
DW on the information I had
PG what’s weird is that you concluded he was wrong and then...

… when the defendant’s evidence agrees with the claimant you refuse to go with anyone’s evidence.
DW it’s not weird
PG what threshold of proof were you applying. The claimants and defendants agree and you refuse to come to a conclusion
DW it’s what I can usefully do for the...

… court. I was trying to be balanced and fair.
PG if you had got to this point without first offering a forceful opinion we wouldn’t be here. But you first offered a strong opinion and then refused to correct it. I want to know why
DW want to be balanced and fair

PG is this an example of you being balanced and fair in your reports
DW this is fair example of me assessing the evidence in front of me. In my reports I have tried to be fair in my assessments and when it comes to the claimants I have tried to be biased numerically in...

… their favour. On some occasions.
PG so this is an example of you doing your job well, then. You’re happy for the court to see this as a good example of methodology?
DW yes.

PG did you speak to any Fujitsu engineers.
DW Only on the phone with Gareth Jenkins, but the Post Office lawyers were on the call too.
PG I don’t want to ask you about the contents of the call if it was legally privileged. But you spoke to Mr Jenkins on the phone.
DW yes

PG anyone else?
DW no
PG what about during this trial? there were lots of Fujitsu employees here
DW no the Post Office lawyers set ground rules…
PG I don’t want to ask you about that. did you speak to any of them?
Mr de Garr Robinson: I have no objection to discussing...

… the ground rules if my learned friend wishes to.
[PG doesn’t]
[PG is exploring what has and hasn’t been told to DW by Fujitsu]

PG Mr Coyne made lots of requests - joint requests. But although he was requesting lots of info on a joint document. Why didn’t you support his approach.
DW well - my approach was top down… architecture… looking at things like KELs
PG and that’s your explanation...

… for not supporting requests.
DW I had enough to do
PG and you only looked at KELS
DW not all the time - but they were the most useful to get an overview.

[PG asks about access to documents. brings up a June 2018 letter from PO lawyers which says DW has had “broadly" the same access to documents as JC]

PG JC asked you in an email if you’d had any access to any documents that he hadn’t
PG can you answer that?
DW I was just trying to get on with the job - I’m not a good witness with blow by blow stuff.
PG don’t worry - let’s move on
J let’s not. given you have asked...

… the witness to do some homework over the short adjournment I think it’s best to look at it after lunch. Also suggests that if PG does want to ask anything about the ground rules, he gets copies rather than turn it into a memory test.
Also he didn’t say “let’s not”...

… he politely but quickly interjected before PG could move on.

As I said - everything is a summary or description (ie paraphrase) of what is going on in court. Nothing is a direct quote. And I made that clarification because the original had...

… a whiff of misrepresentation about it.

[the short adjournment]
[I’m going for lunch with @Karlfl from @ComputerWeekly. Back at 2pm]

[we’re back. we go back to ground rules]
PG what were they
DW if we crossed a line the lawyers would stop us talking
PG did that apply to Fujitsu?
DW yes
PG when did the ground rules apply
DW for the trial
PG what about your call to Gareth jenkins
DW that was a year ago.

DW the lawyers were on the line for that but there were no ground rules around that time.
[we move on]

[we are looking at an internal PO document dated 2017 called technology strategy update which describes Horizon as “at the end of its life and needs replacing”]
PG is that fair comment
DW 20 years is a long life and that’s what they’re saying
PG and previous attempts...

… to move away from Horizon with IBM have been unsuccessful.
DW I didn’t really see this document
PG “Fujitsu see this contract as a cash cow”- see that?
DW not at the time - didn’t read that far, but always had the impression were not short of a budget
PG as in - coming from..

PG Post Office.
DW yep

[we move on to another PO document]
PG Horizon software developed in 1996 as part of a DSS project. H was one of the first points of sale. A closed system based on paper process. Calls Horizon “clumsy and operator unfriendly”.
Did you know this was PO’s view of its own system?

DW no
PG “however whilst modernising the front end would be relatively simple…” modernising the back end would be extremely difficult. and so they didn’t complete the IBM deal. It is a “closed and inflexible platform”
DW the back end was more complex.

[we go on to a Post Office 2016 document called Back Office Transformation]
PG “Our back office also struggles with the complication of dealing with each of our many clients. Heavy manual processes… and a lack of flexibility” Poor support contracts, prohibitive cost of change...

… prevented us progressing as business as usual. Did you pick up on this?
DW I picked it up when I looked at the documents cited by mr Coyne
PG did you deal with them in your reports?
DW I didn’t think they were that relevant as they were about the back end IT systems.

PG Back office systems are "complex, unreliable, not suitable to maintain and not fit for today’s business”. Seen that?
DW doesn’t surprise me
PG unreliable
DW well, unwieldy
PG so are you rowing back on unreliable
DW well its what they said

PG okay so reliability is important. because it’s one of the H issues
DW yes but I got a sense that I shouldn’t go near that.

PG from whom
DW from the PO "I was told it was out of scope and I found some difficulty with that because things like reliability of Horizon depends on all sorts of things”
PG yes but the back office systems are relevant to data integrity, transaction corrections

DW and third party stuff
PG like camelot, bank of ireland, etc
DW yes

[we go to a document called Horizon Online Induction Training]

It is a 2009 appraisal:

13yo tech to satisfy a [redundant idea]
A robust service but complicated to change
Slow and expensive to change
H is a system wrapped up in barbed wire making change costly - test everything!

DW this is different from my own understanding of Horizon
PG hadn’t seen this document?
DW no

[we move on]

[it’s an internal Post Office email about Pam Stubbs - Barkham SPMR which came up in the Common Issues trial where the balances are all over the place etc it’s the famous “it IS a Horizon issue” email]
PG accept an email that asserts this is correct that there is an H issue?

DW don’t know
PG is it fair to accept it might be right on the face of the document
DW maybe
[reads on - Mrs Stubbs was told she needs to prove H is to blame]
DW my general view is that H has been built to be robust against power failures
PG so you wouldn’t expect this outcome

PG in this example
DW no
[we move on]

[2001 audit of Mr Alan Bates’ branch]
PG you probably have heard of him - Bates v Post Office
DW I have. he’s here.
PG he is.

PG cash holding - high risk and then under comments - correct assessment of cash holdings could not be made as the H system is adding the previous day’s totals to today’s amounts.
PG have you seen that before
DW no
PG is that consistent with your assessment of how H works?
DW no

[we move on to documents uploaded to the shared space on 31 May 2019 by the Post Office]
PG it was uploaded on 31 May but contains MSC info. Have you seen this
DW no
PG do you know why it was uploaded?
DW no, no idea
[PG takes him through how it works]

PG did you find the MSC logs in their original form difficult to use
DW not too tricky, bit clumsy, but I built a java tool to filter the MSC so I could find what I wanted
PG let’s have a look at this specific one… about an SQL insert - PO have requested this issue be resolved..

… and then there’s a reference.
PG on p2 "POL approval required Y or N?" It says no - "POL aware" and gives same reference
PG so on a fair reading PO request a change and don’t need to be informed of it as they’ve already requested it. yep?
DW yep

PG you can see impact, testing security implications, do users need to be informed of the change? “No”
DW yup
PG so they’ve got two fields of info cross references from the three MSC spreadsheets provided, but they can’t find the third. yep?
DW yep.

[we go to compare it with another MSC doc]
[this one appears to show the problem Mr Bates was having in his branch which the auditors wrote about from a Fujitsu perspective]
PG there’s a fix proposed and Post Office approval again is not required… yes?
DW yes

PG there’s a delete SQL being applied to remedy a problem in the database
DW yep
DW What we don’t have in this upload is the data from the other two spreadsheets, so we need to look in the spreadsheet itself...

[sorry that was PG not DW]

[we go into the spreadsheet]
DW so this is all theimpact stuff
PG it is all the impact stuff, not in the document provided by Bond Dickinson. So:
impact: N/A
users informed: No
testing: N/A
risks: N/A
do users need to be informed? No
Do you see that?
DW yes

[PG takes him to a PEAK]
DW does this refer to the MSC?
PG yes - shall we have a break whilst you look at it
DW have you got a paper copy?
[we don’t]
J says we have a break - he is going to ensure counsel has paper copies for DW in future

[we’re back. we’re looking at a PEAK related to an MSC report. I say we, but I can’t see any of this]
PG “branches will be forced to declare stock when they don’t want to” causing “spurious discrepancies”
DW can we go back to the MSC?
PG let’s just take it in stages...

DW but it is the MSC we’ve just looked at?
PG yes
DW that’s my puzzlement
PG you see the PEAK refers to the MSC
DW yes
PG “spurious discrepancies”
DW have you see this document
PG can’t say I’ve seen it before

[oops sorry got my DW and PG flipped in the last exchange]

PG reads the PEAK to say that there is a problem here and the fix hasn’t in certain cases worked. and it will affect at least 36 branch accounts
This is not H working as it should is it?
DW no

PG goes on to read the solution is to get a new fix in, and do it quick so Fujitsu don’t have to tell the Post Office, as they’d already told the Post Office this problem had been fixed. Agree that’s what’s going on here?
DW yep probably

PG reads: risk of not doing anything - previous year’s balance will get dropped into accounts and cause stock discrepancies in live branches. see that?
DW yep
PG so Fujitsu is attempting to get a fix in that neither the PO or SPMS will be aware of
DW agree

[we move on to Horizon Issue 1 and DW’s approach to it. I will call it H1]

PG the q posed it "to what extent is it possible or likely for bugs errors or defects to have the potential to cause apparent or alleged discrepancies or shortfalls”
DW yes
PG words “to have potential”...

… are important?
DW well the word potential reflects a lack of knowledge
PG well it’s not the most transparent task to engage in - difficult to plumb - you’ve used the word swamp
DW it’s a big complex system
PG so if something DOES cause a problem is also has the potential...

… to cause a problem [elsewhere?]
DW yes
PG you focused on actual impact rather than potential. and you added in “lasting” and you focused on claimants’ accounts
DW agrees
PG but you saw it said SPMRs branch accounts

… in H1
DW I thought extent in issue 1 required calibration, so it needed numbers.
PG but the helpline and manual processes of Horizon was not what you looked at
DW as experts we did look a bit at the helpline.

PG did you know the helpline people read out scripts?
DW wouldn’t be surprised
PG but did you know
DW no
PG did you know it was referred to as the hell-line for example
DW no.
[we go to a forum post by Mr Tanq]

[Mr Tanq was a claimant witness]
PG Mr Tanq describes £176 discrepancy which he pays up himself because he can’t cope with going through the stress of the "hell-line”. Seen this?
DW no - I did look at 2 weeks worth of helpline logs
PG and that was it?
DW yes

PG your answer to the robustness of H is about TCs and countermeasures which reflect that process and you effectively say those are important because that’s the way errors are corrected.

PG the experts have differing views on branch impact. Mr Coyne sees it as anything that has an impact on branch accounts, you see it only as something which happens if the TC fails.
DW yes… [qualifies]
J just so I’m clear - if a TC takes several months that would be a transient

… inaccuracy?
DW yes, and for that period there would be a discrepancy in the branch.
[we move on]

PG you have written a report on how you deal with H1 in section 8 of your first report. You’ve made assumptions to arrive at a number. Yes?
DW yes
PG let’s look at a couple of them… one of the assumptions is that all branches whether claimants or non-claimants are affected...

… evenly. and another is basically the same point - on the premise that it affect an even distribution of bugs you’ve considered a scaling factor, haven’t you
DW yes
PG it seems implausible to me there is a special factor about claimants branches that makes them more...

… susceptibly to bugs. and you’ve given a lot of consideration to this. as it’s important to your report.
DW yep
PG and you’ve written later that you have SHOWN this to be the case. Do you mean by “shown” you’ve thought about it and cant’ see any reason why not
DW that’s fair

PG so you wracked your brains and the only thing you can think of is that claimants make more errors than non-claimants. Or that bugs are better handled by non-claimants. Then you discount them.
DW i called them of secondary effect
PG did you hear from Mrs van den Bogerd...

… when she said that the Post Office suffers from User Error Bias
DW no i didn’t
PG do you think you’re suffering from that here?
DW no
PG no because you rule it out

DW I consider it and rule it out
PG but you do think that claimants branches on the whole are smaller
DW yes
PG and so you put that in your report assumptions
DW yes
PG so you come up with the conclusion that the chances of the claimants being correct are absurdly unlikely

DW yes, but its important to quantify words like likely which is why I gave the court some numbers. And if my assumptions about claimants branches are wrong then the court can use the calculations to come to a different conclusion.
PG okay - let’s do a statistical exercise.

[they work out if there are 120 people in a country of 60m called Penny Black then that’s 1 in 500,000. You go to a dinner of 50 people. There is a 1 in 10,000 chance of someone being there called Penny Black. A 1 in 10,000,000 chance of 2 people there called Penny Black]

[there is a 1 in 1tr chance of three people of that name]
[but if all 50 people were invited were called penny black (as they were at a Royal Mail anniversary dinner) - then you have a near 1 in 1 chance of meeting a Penny Black]

DW that’s because you do statistical analysis in the absence of specific information. The specific knowledge changes the whole ball game
PG “The specific knowledge changes the whole ball game” Yes.
Do you know how GLOs work?
DW no
PG they are advertised - it is no surprise...

… that they have one thing in common.
DW I don’t follow this.
PG this is your specific information which you say you don’t have.
DW no I haven’t I am working on the basis of what a bug in Horizon might do not what claimants believe has happened to them

PG okay lets move on to your methodology - your scaling factor and your maths because maths is right or wrong - it’s not a matter of opinion
DW it’s a great advantage for the court and expert witnesses.
PG you’re scaling factor concludes that there is a 0.45 chance...

… of an claimant SPMR being hit by a bug which may have affected branch accounts [I think that’s right - it’s the concluding total which has been mentioned many times in cour before. I now finally have the report so I can see exactly what it is meant to represent shortly]

[we are now looking at the source material which informs one of the factors which led to the 0.45 calculation]

PG explains how two of them are different from what he says they are.
DW accepts
PG asks if he was in court for when AvdB accepted that the figures that were extracted for her in a report were wrong
DW wasn’t
PG asks if he was told
DW heard about it
PG did it worry you?

DW I did think about it, but the level of precision I needed for this particular calculation was very low.
PG when the corrected data came through did you use it?
DW I must admit I’m not sure as there were other areas where the precision of my data was more of a concern...

… right.
PG well take it from me you used the column with the gaps in it
DW difference is of 4%
PG yes but you used the wrong numbers
DW well it was not important. when you are making an engineering estimate - one is the precision of the result and the other is the precision...

… you can achieve. you have to look at it and ask yourself how precise you have to make. They do not worry about spurious precision.
PG did you worry about spurious precision before you decided to ignore the correct figures
DW I decided it was not necessary to worry about those

… details.
PG why decide by 561 - number of claimants?
DW yes
PG even though there is only 496 on the spreadsheet
DW yes
PG so you knowingly presented to the court the wrong figures
DW it doesn’t matter for statistical purposes
PG let’s go to the heading of session...

[sorry heading of the document]
PG you’ve decided the number of sessions is the number of transactions, yet we know there are multiple transactions in a session
DW yes about 1.7
PG so you knew this and you scaled it down anyway.

DW well the definition of transaction is rather fluid. what the 1.7 refers to is how many things a customer does in one session. but I agree there are big uncertainties in my calculations
PG none of this is stated in your report
DW it’s not a big uncertainty. the biggest...

…. uncertainty is the number and impact of bugs.
PG yes but even on your evidence here you could have doubled your scaling factor.
DW this claimant scaling factor doesn’t really matter - in my report I focus on the biggest areas of uncertainty in my analysis.

PG okay - in AvdB’s report she says there’s 47bn transactions. You say there’s 48. Is that just sloppy.
DW it’s effect is so small it doesn’t matter.
PG okay - so all these mistakes - you say they’re not, we say they are. all these mistakes are against the claimants.

PG is that just bad luck? if there’s a 50/50 chance of making a mistake it’s just bad luck if they’re all against the claimants
DW yes, but I do make large assumptions, by a factor of 30% in favour of the claimants.
PG yes you flag that up a lot, but there’s not a mention here...

… of the decisions you made against the claimants. You understand the difference here?
DW No because these are tiny by comparison.
[we move on]

PG let’s look at the smaller branches scaling factor that you used as your methodology in your report. par 630 - in your first report.
reads: claimant branches have fewer transactions on average so they have less likelihood of them being hit by a Horizon bug.

PG so you assume that errors are a direct consequence of bugs in transactions
DW yep
PG now we have data of branches which were affected by actual bugs that we know about
DW we do
PG let’s take the Dalmellington bug which affected 88 TINY branches

DW well outreach from other branches
PG so shouldn’t we look at how they were affected shouldn’t we?
DW It’s the next level of detail down
PG it affects the control variable
DW yes I could have done it
PG why didn’t you? what do you say about why it’s acceptable to leave...

… that out.
DW because you make an engineering assumption about what is required in a result and I decided it was not relevant.

[Ah okay - the .45 is the chance of a claimant branch being hit by a bug compared to an average branch. PG says if you look at outreach branches, then because of Dalmellington, outreach branches are 10 x more likely to be hit by a bug]

PG suggesting out that the premise of his scaling factor - it is designed to control for a branch having a lower chance of a branch being hit by a bug, because of the smaller number of transactions
DW if a bug affects rollover it’s got just as much chance of being affected...

… as a bigger branch.

[PG onto final point “maths”]
PG back to the column without the gaps being taken. That was a mistake - you looked at the column headings and saw 2007 GAPS and took that column.
DW for the level of precision I need you can take either column.
PG I’m suggesting your evidence...

…. earlier was wrong. It wasn’t a deliberate choice. It was a mistake. Was it a deliberate choice to take a column which had gaps in it
DW no
PG was it a mistake?
DW it didn’t matter which one I took so I didn’t worry about it.

[we are mopping up. all I can suggest is that the moment I post it up you read the transcript because today has been fascinating and these tweets are just a flavour of what’s going on]

The transcript will be posted on postofficetrial.com later. I will tweet it when it’s up.

Report to follow. Thanks for reading. This is crowdfunded reporting. If you have enjoyed reading any of this and you haven’t donated before please do go to the above website and...

… chuck a few quid in the tip jar.

I’ll get this threaded. We are back in court THURSDAY, not tomorrow. Mr de Garr Robinson has an urgent matter tomorrow which has meant shifting the last two days of cross examination along to Thursday and Friday.

@threadreaderapp unroll pls