Guess the ISP

Peak-time performance for 5 ISPs, A, B, C, D and E

Can you match them to BT, EE, Plusnet, Sky, TalkTalk?

Example of how to read the graph: How many users at peak-time achieved on average at least 95% of their maximum speed result? ISP A = 89%, B = 72%, C = 47%, D = 42%, E = 4%. The closer to the purple-dashed line, the better the ISP performance. Graph compiled by myself from raw dataset.

data is too old

Post deleted by RobertoS

Edited by RobertoS (Sat 11-Jul-15 10:16:35)

Yes, it's well before G.INP was introduced.

Also I think it's before BT Consumer put it on their website that peak-time slowdowns were normal and to be expected. Which itself was after slowdowns caused by apparent BT Wholesale became common.

The trends are remarkably constant from 2013 throughout the whole of 2014. So this is established performance, it doesn't change from one week to the next, and it is the base on which future results will be founded. This is the only proper analysis of peak-time speeds I know of.

If you have specific guesses about any changes for ISPs in 2015, then nothing stopping you saying them When fresher results come we can see if you were right.

If people know so well how things are right now, what they were half a year ago should be dead easy! Yet no takers so far!

(G.INP is irrelevant and there is no evidence whatsoever of a general BTw issue to affect all 3 BTw ISPs, but more to the point you sure can't see 3 equal looking-graphs anyway!)

Can you just tell us.............

I expect she's been pulling data out of this report.

Oops. More likely this one, which the first one comes from.

Edited by RobertoS (Sun 12-Jul-15 22:48:04)

Wow, how clever.

My second edit gives the actual source I think. My first edit was a half-way house. The one there now has ISP breakdown.

Edited by RobertoS (Sun 12-Jul-15 22:49:44)

Er, no. I did exactly what I said I did, thanks.

PM sent.

Well, there are two possibilities.

The raw dataset belongs to samknows, or it is nothing to do with samknows. The coincidence of November 2014 dates suggests the former.

If you are therefore publishing here a self-compiled analysis of samknows data, that would be interesting in itself. If you aren't, then we have no reason to accept your graph.

Ok, smarty pants.

BT, Sky, TalkTalk, Plusnet, EE.

What's the prize?

P O I N T L E S S P O S T

Don't be hard on yourself. Your post displays advanced use of a space bar.

In reply to a post by RobertoS:

If you are therefore publishing here a self-compiled analysis of samknows data, that would be interesting in itself.

It's no secret, this is declared in the first post and graph itself.

You don't declare the source. That was my point. I assume if it is samknows then you are in a position of sufficient authority to do this without acknowledging it, or have authority so to do, which is unlikely.

Edit - ohhhh, you've edited the graph caption. Why not acknowledge your error dear lady? You are always very quick to challenge other people.

Edit 2 - I think you probably need more than the simple acknowledgement anyway, as the raw data has not been published. Only reports. You need explicit permission, and probably approval. As things now are you are probably committing an even greater offence than without the attribution.

Edited by RobertoS (Tue 14-Jul-15 09:33:52)

In reply to a post by vimto_girl:

Yet no takers so far!

Hi vimto_girl. Any comments on my take:

BT, Sky, TalkTalk, Plusnet, EE (A to E order).

In reply to a post by alext05:

Ok, smarty pants.

BT, Sky, TalkTalk, Plusnet, EE.

What's the prize?

Not bad, 1 right out of 5!

Prizes are to be determined, disappointment guaranteed

Well, this is getting interesting! Thanks.

Look forward to the unveiling of the results.

I think your "smarty pants" comment is becoming more and more appropriate. In addition to the two edits to my recent reply to her, I also wonder who owns the raw dataset and/or the copyright in it. Ofcom would appear to have commissioned the research.

I have a feeling the OP may be even more out of order now she has declared the source. Unless she is sufficiently senior within one of the two organisations, probably samknows, to be fully aware of the contractual position regarding that data and also either to have authority to use it for this unofficial graph or to have obtained such authority.

It all seems rather odd that an anonymous person can publish on the Internet a graph compiled by themselves from this proprietary raw dataset.

Except I haven't edited anything at all, and have all the approval I need, thanks

Liar. The graph did not originally mention Samknows. If it had, I would have been questioning your permission in the first place, not trying to trace your source as per my earlier suggestions.

I'm never happy with graphs that don't show the full axis. VERY misleading.

"Unless you like very tall empty graphics (even log 0 tall), don't show unnecessary zero points. Real scientists show data, not zero points."

"Don't spend a lot of empty vertical space trying to reach down to the zero point at the cost of hiding what is going on in the data line itself. (The book, How to Lie With Statistics, is wrong on this point.) For examples, all over the place, of absent zero points in time-series, take a look at any major scientific research publication. The scientists want to show their data, not zero."

Professor Edward Tufte, leading authority on visual statistics and author of the classic text The Visual Display of Quantitative Information.

I understand some people are y-axis Nazis, if you are one of them, you can always print the graph out and move the x-axis down to show a lot of blank space If I did it in the image, there would simply not be the space or resolution to show the data intended. It would be the same if I was plotting Lewis Hamilton's lap times over a grand prix on a F1 forum, it would be silly to insist it starts at zero. Caution is good, but rigid rules are something else, especially when they lead to useless graphs or ones the size of a small village.

I'm not lying, let alone a liar, thank you very much. Anyone who looked at the graph can back me up. AFAIK it is even impossible to change an image uploaded to postimg, you have to do a new upload which gets a new link and seen as my post has not been edited, so that proves that.

I'm not surprised you admit your only planned contributions are to distract from the results at any cost. Trolls gonna troll

In reply to a post by vimto_girl:

Caution is good, but rigid rules are something else, especially when they lead to useless graphs or ones the size of a small village.

Agreed. It's all a matter of context and whether the chart represents the data appropriately. Sometimes charts are used or manipulated to emphasise the point, rather than merely demonstrate it.

In reply to a post by vimto_girl:

In reply to a post by alext05:

BT, Sky, TalkTalk, Plusnet, EE.

What's the prize?

Not bad, 1 right out of 5!

Prizes are to be determined, disappointment guaranteed

Guess I am winning so far - no more takers for some reason.

In reply to a post by RobertoS:

I think your "smarty pants" comment is becoming more and more appropriate. In addition to the two edits to my recent reply to her, I also wonder who owns the raw dataset and/or the copyright in it. Ofcom would appear to have commissioned the research.

I have a feeling the OP may be even more out of order now she has declared the source. Unless she is sufficiently senior within one of the two organisations, probably samknows, to be fully aware of the contractual position regarding that data and also either to have authority to use it for this unofficial graph or to have obtained such authority.

It all seems rather odd that an anonymous person can publish on the Internet a graph compiled by themselves from this proprietary raw dataset.

unless of course it were published as open data.

In reply to a post by vimto_girl:

Graph compiled by myself from raw dataset.

Unlikely that dataset has been published, but the OP is of course welcome to post a link if it has.

no more takers for some reason

For what is is worth vimto PM'd me (and probably others) the answer (yawn)

It is very old data and the differences are really quite small, on all 5 ISP's >80% of user achieve >80% of Max download speed at peak time; impressive for an inexpensive service.

Ian

Or from the graph 80% achieved 82%, even more impressive.

What tickled me in the Ofcom report in my second link earlier is that Plusnet had the highest incidence of disconnections. Maybe we should tell samknows about the multiple rapid channel hopping attempts that may be distorting that figure .

>80%

COUGH...COUGH

Could not be bothered to work out the %age but thanks....

multiple rapid channel hopping attempts

In November 2014??

http://community.plus.net/forum/index.php/topic,1353...

The capacity issues thread started on new years eve.......2014 ....

You could be right about the bulk use of it, but it was well known much before then that some gateways were better than others, and that sometimes they got clogged up.Then Plusnet used to bump a few users off to rebalance the system.

IIRC the BT Wholesale system was just dropping us on using a round-robin system, not a load-balancing one.

In reply to a post by ian007jen:

Can you just tell us.............

In reply to a post by ian007jen:

For what is is worth vimto PM'd me (and probably others) the answer (yawn)

Ha, so much to say after you got the result and clearly didn't like it Why didn't you just say your piece before huh?

The differences are not small, take your 80% peak-time speed target (one you picked out all by yourself), that is a reduction from 38 Mb/s to 30 Mb/s in the evening. With the best ISP there was 0% chance of not hitting this target, with the worst there was a 13% chance, more than 1 in 8. That's a big difference in my book.

And the data is not very old, it is from Nov 2014, and results prove performance remained remarkably consistent through 2013 and 2014, so this shows established performance and trends. And without past results, future results have no reference or significance, and any changes specific to 2015 can't be captured.

Edited by deleted (Tue 14-Jul-15 21:11:01)

This is probably the most boring and uninteresting post ever on a web site that has multiple boring, repetitive and uninteresting posts.

us....

Not just me....

I suppose we could look at the document http://stakeholders.ofcom.org.uk/binaries/research/b...

Figure 2.8 on page 23...I did remember seeing this before but with the colours of the lines (and my colour blindness) I found it very difficult to make out the ISP's, I remember thinking it would be easier with dotted or dashed lines.

This graph does show the the small variation of the top 5 (up-to 38 and 76).

Just noticed the x and y axis have been swapped, can you re-plot your graph this way round, and with the vertical axis in 5 steps (0,20,40.....100).

Ian

This is probably the most boring and uninteresting post ever on a web site that has multiple boring, repetitive and uninteresting posts.

Could not agree more

(yawn again)

In reply to a post by rogerfp:

This is probably the most boring and uninteresting post ever on a web site that has multiple boring, repetitive and uninteresting posts.

That's ok. I can deal with that.

It wasn't your post Alex, I was referring to, but the OP. Sorry if you got the wrong impression. I was just agreeing with you that the whole thread is pointless and a waste of space.

Edited by deleted (Tue 14-Jul-15 22:18:48)

So you're saying the graph published by Ofcom is the same as the graph in the OP with the axis swapped?

Yes, I might have misunderstood you. Cheers.

with the axis swapped?

I think vimto-girl will have to answer that one....but that what it looks like to me; although she did process the raw data herself.

Access to the raw Ofcom/SK data would be very useful to verify whether the money spent is worthwhile or just an exercise in ensuring Ofcom has some data to publish (which as its Ofcom is difficult to question), e.g. what speed are the lines they are testing and where they are so it can be determined how much information is real data rather than modelled

The location is very important when doing drop due to peak speeds, e.g. are the areas where people moan about virgin media speeds in the list etc

they in the ofcom tests yes.

VM refused to add me as a tester, I emailed ofcom to ask if they blocking VM bad areas, then got a reply saying its upto VM with what they do on their own tests but they were happy to have me as a ofcom tester.

Samknows do testing separately for VM and ofcom (and possibly other isp's).

In reply to a post by ian007jen:

with the axis swapped?

I think vimto-girl will have to answer that one....but that what it looks like to me; although she did process the raw data herself.

Noticeable that the results here are restricted to FTTC 40 Mbps products only, so fail to show the abject performance of two of the VM product lines.

On the plus side, these results include TT performance, which isn't in the Ofcom chart.

And better, partly as a result of not bothering to put zero in the chart, this chart highlights the differences above 90% - which is where behaviour alters the most.

I'm not exactly a zero-axis nazi ... but I do prefer it if the axis gets a zig-zag wiggle to help make it clear.

The SamKnows reporting dashboard fails to plot zero, by default - and a first glance at a graph of download speeds makes it appear as though mayhem is in charge. Until you remember ... and it turns out that the speeds actually differ by less than 0.5Mbps.