Member Info for taufour

Oops and apols. For at least the last several days the homebrew app at luku.org/tau (ref previous posts from me) has been rendered barely usable by its service provider sporadically killing it on grounds of 'excess use of resources'. At the moment it works about a tenth of the time; other times it might either stop with an error message or just freeze (but the 'cancel' button should still work).

Gah.

Apparently it would be OK if it were on its own server but that would cost at least 5 dollars a month and I object on principle.

I am going to split it into parts via a really cool redesign so that each part should fit within various cloud providers free tier service. I will post again if and when the app resumes reliable operation.

Hi. If a 15 minute delay is acceptable there is the stock exchange's own data feed at

https://www.londonstockexchange.com/exchange/prices-and-markets/stocks/summary/company-summary/GB00B0DG3H29GBGBXSTMM.html

with recent trades (though lacking bid/ask info) at:

https://www.londonstockexchange.com/exchange/prices-and-markets/stocks/exchange-insight/trade-data.html?page=0&pageOffBook=0&fourWayKey=GB00B0DG3H29GBGBXSTMM&formName=frmRow&upToRow=-1

Sometimes it is interesting to focus on high-volume trades. I've added a new viewing mode to the home brew app at http://luku.org/tau

Start it in the normal way to get a diagram of recent trades, then press the button at lower right marked 'Mode'.

Select 'Volume Kings' and enter a number into the volume box.

Then only trades at or above that volume will be displayed. As usual you can adjust the date range (button marked 'Range')

Several people have commented today along the lines of "Today's drop is all about fear and uncertainty".

But according to http://www.lse.co.uk/SharePrice.asp?shareprice=SXX&share=sirius_minerals the statistics for today are "Vol. Sold" 12.5 million and "Vol. Bought" 22.2 million.

Looking in more detail, A-trades today have been balanced, whereas the O-trades show a whopping discrepancy. According to my home-brew app (which is only catching about 80% of trades but is useful as an indicator), O-trade volumes today are 20% in the low quintiles and 74% in the high quintiles.

So this is not looking like selling pressure.

Intesting comment from brexit (16:57) "Bought another 5k, pounds that is, on line late this afternoon but took an hour before I got a quote".

My provisional inference is that MMs are *trying to create* fear and uncertainty by dropping the price.

Hi. 'I'm getting a "404 Not Found" with ...'

Myo's link is a 'deep link' but the spiffy new map/animation feature at https://map.siriusminerals.com doesn't accept deep linking.

"Deep linking is the use of a hyperlink that links to a specific, generally searchable or indexed, piece of web content on a website (e.g. "http://example.com/path/page"), rather than the website's home page (e.g., "http://example.com")." (wiki)

Not accepting deep links means that visitors need to enter at the front ( https://map.siriusminerals.com ) and click on the pulsing buttons (presumably after putting in the work to make such a nice feature, the designer wanted visitors to experience it in full)

Can anyone with the right kind of knowledge give us an idea of how many shares the market makers would typically hold? I've put a snapshot of SXX for this week (well, up until mid Friday afternoon) at http://luku.org/snap/SXX/pk38zzac/

It indicates that while A-trades have been balanced, O-trades have been 8 million in the low two quintiles and 15 million in the high two quintiles, suggesting a depletion in MM stock of about 7 million shares over the week.

Is that enough of a disparity to be worth noting, or is it fairly normal?

Hi PAAA, thank you for detailed explanation. Where I was going wrong was in thinking that the spread is defined as the interval between the highest MM bid quote and the lowest MM offer quote, and imagining that SETS trades should be buzzing along in a tighter band than this.

If the spread is actually defined as the interval between the highest buy order on the book and the lowest sell order ... then yes I can see how by definition SETS trades would tend to take place on the bid or on the offer.

Now to check my understanding ... how does it come about that *some* A-trades take place *between* the bid and the offer? Is it perhaps due to matching buy and sell orders coming in at the same moment and being executed immediately instead of being put in the order book?

Not sure what you are actually asking, but as far as I understand it the concept of buys *versus* sells is only meaningful for O-trades.

It would be good if you could explain what you are actually trying to find out, bearing in mind that there can't really be significantly more or fewer buys than sells over the course of a year ... it just has to balance out, approximately.

Meanwhile here are some possibly interesting observations. The app at luku.org/tau has data (in some cases) going back to last July and for SXX there is a snapshot up to the end of 2018 at luku.org/snap/SXX/7xe98gs3/

The idea is that if we divide trades into quintiles according to where each one lies, proportionally, between the bid and the ask, then the lowest quintile (0 to 20%) are probably sells; the next quintile (20% to 40%) are more likely sells than buys; the middle quintile (40% to 60%) could be either; the fourth quintile (60% to 80%) are more likely buys; and the top quintile are probably buys.

At least, that logic makes sense for O-trades. As I've commented in recent posts here, I am not very confident about what any of this means for A-trades.

Also, please note that the app is currently only finding about 80% of the actual trades, so it is at best only an approximate indication of trends.

But laying those doubts to one side, in summary the app is saying that over the last five months SXX has been fairly balanced between O-trades and A-trades, with total volume getting on for a billion below the midprice and a billion above it.

SXX for 2018, monthly

https://uk.finance.yahoo.com/quote/SXX.L/history?period1=1514764800&period2=1546214400&interval=1mo&filter=history&frequency=1mo

Date Open High Low Close* Adj. close** Volume
01 Dec 2018 22.78 23.18 19.31 20.80 20.80 213,408,528
01 Nov 2018 22.90 24.80 22.00 22.78 22.78 255,803,713
30 Sep 2018 28.10 28.30 21.22 23.04 23.04 523,079,822
31 Aug 2018 36.54 36.76 22.50 27.94 27.94 761,141,893
31 Jul 2018 37.18 39.78 35.20 37.24 37.24 301,489,417
30 Jun 2018 32.78 37.56 31.90 36.38 36.38 282,261,666
31 May 2018 34.40 34.92 31.02 33.08 33.08 284,685,921
30 Apr 2018 31.28 35.08 30.90 34.40 34.40 383,783,520
31 Mar 2018 30.36 32.88 26.42 31.00 31.00 455,566,371
01 Mar 2018 28.00 30.86 26.20 30.36 30.36 595,619,837
01 Feb 2018 23.56 29.00 21.10 27.50 27.50 296,108,906
01 Jan 2018 23.50 24.70 22.00 23.56 23.56 262,687,656

Adding up the twelve numbers in the last column gives between 4 and 5 billion

Thinking further about why A-trades tend not to occur at the mid price ... I never thought I'd be the kind of person who would come up with a conspiracy theory but that's what I'm going to do. Seriously, I can't at the moment think of any other explanation.

The theory is that A-trades really do occur at the mid price, but they are not reported as such for fear of fomenting ill feeling among private investors who would see all these trades going through at prices better than what they are able to get.

So the A-trades are reported as being either on the bid or the ask, on pretty much a random basis.

Hi PAAA, but then why don't A-trades tend to be priced near the midpoint? Overwhelmingly they tend to be priced near the bid or near the ask. Whereas O-trades have more of a propensity to be priced in between. Not doubting your explanation, just finding it very puzzling. Can you cast light?

Here is my current understanding of how A-trades work. *NB* this understanding must be wrong, because it does not fit the facts. Can anyone cast light?

A-trades are (I think) the result of an automated bid/offer matching system whereby a seller can sell directly to a buyer without a middleman. Using such a system, a seller can get a better price than if they sold on the bid to a market maker, and a buyer can get a better price than if they bought on the ask from a market maker.

However, if that were the case, then A-trades should tend to occur at or near the mid price between bid and ask. But they don't, so far as I can tell. Looking at the numbers it seems that A-trades overwhelmingly tend to occur on the bid or the ask, whereas O-trades are more evenly spread out in between. I can post some diagrams but first I'd like to try to clear up my misunderstandings to some extent.

Tia.

Hi Alan, here is my attempt at quantifying the long/short position for Wednesday, which was when you said "Checked up on Official stock exchange figures more buys than sells and SP red it's a funny old game !!"

luku.org/snap/sxx18k28b/

The script is currently finding only about three quarters of the trades; for Wednesday it shows 9.1m trades, whereas https://uk.finance.yahoo.com/quote/SXX.L/history shows 12.4m

So treat with caution. But there may be enough information to try making some tentative inferences.

The basic idea is that if we divide trades into quintiles according to where each one lies, percentage-wise, between the bid and the ask, then the lowest quintile (0 to 20%, near the bid) are probably sells; the next quintile (20% to 40%) are more likely sells than buys; the middle quintile (40% to 60%) could easily be either; the fourth quintile (60% to 80%) are more likely buys; and the top quintile are probably buys.

The script gives 3,282,579 probable or likely sells, and 3,888,118 probable or likely buys. So at first glance it agrees with your "more buys than sells" comment. But if we look deeper, it doesn't really agree.

Note in particular the different behaviour of the O and A trades. The A trades are more disciplined, almost all in the top or the bottom quintile, whereas the O trades are much more all over the shop. And although the A trades have a much smaller average volume than the O trades (5k versus 31k), there are far more of them (589 A trades versus 154 O)

It looks to me like a relentless drip of small A trades, 1.8m in the lowest quintile and only 1.3m in the top quintile, that could have walked the sp gently down throughout the day.

Thanks, very interesting. I imagine that probably by the terms of your subscription you cannot relay actual numbers. But if someone writes a script that tries to get as much of this information as it can from publicly available data, could you let us know from time to time if the results reported by the script are reasonably close or are way off?

AlanG "Checked up on Official stock exchange figures more buys than sells"

Can you explain please? Noting PAAA's reply to you, I suppose it's possible that when you say "more buys than sells" you are intending it as mere shorthand for "more trades above the midpoint than below the midpoint". But as far as I'm aware, even that information is not easy to find on the "Official" stock exchange free data feed, which does not show bid and ask (which, presumably, are what define the "midpoint").

I think that londonstockexchange.com also offers paid-for data feeds, some of which perhaps could include this information. Is that what you are getting?

Re: 1,060,000 share trade

John2015: "@ 16:49 apparently a buy but the sp is falling ?" (I think you mean @ 15:49)

kingrav. "Then it was a sale surely ?"

Maybe you are being darkly ironic. But I think it really was a buy.

Granted, one cannot be sure, but just from looking at the graph it seems like it was clearly a buy. There is a diagram of SXX from 15:15 to 16:15 today at http://www.luku.org/snap/SXX18k28/

Hi Milo: "there will be some dilution ... In my opinion, it will be approx 20%, and could be either in extra shares or a royalty / revenue share form. It will not decimate our shareholdings, just dilute it a bit ..."

As it's weekend I'll use up some bandwidth to mention that 'decimate' strictly speaking means "reduce *by* one part in ten", not "reduce *to* one part in ten".

> The sense "to reduce to one in ten" is etymologically unsound and omitted by the OED but increasingly common (wiktionary)

Hi ProphetW, "Plenty of buys but price is not moving much, must be a big seller in the background" ...

Not so sure about "plenty of buys".

Looking at luku.org/snap/THR18k07 I think the string of trades around 1.67 today could be sells.

Overall the situation is looking quite balanced. Probs best ignore the low outlier 1.57 at 15:36pm which was for 2 pounds.

radsatser: "Don't forget, after the material has arrived at Wilton and been processed, it will then have to be transported on another Mineral Transport System to the port loading site. This will be a wibbly wobbly way MTS running 5km presumably overground which is not a straight line."

Not 5km but 3.5km.

Hi. I have tinkered some more. The home made app at luku.org/taunine/ is now on version 4.

In case I broke anything, the previous version is saved at luku.org/tau/v3/

Volume bars are now solid for A trades, clear for O trades.

The new version can plot further back into the past, for shares that it already knows about, being based on cron jobs that periodically gather data.

It starts out exactly like the previous version, until you click on the 'Range' button. It then offers an 'earliest' date as far back as it can.

For example, you could now plot some 90,000 SXX trades going back to July. Except that, as the widest possible plot is only about 32k pixels, it wouldn't show much. But you can choose a reasonable-sized stretch of time and plot that.

By the way, it might seem like a terrifically unimportant detail, but when you start producing large plots the shape of the Bayeux tapestry, it helps a lot to have precise control of zooming in and out. Double-clicking on the image zooms in towards the point where you double-clicked; you can keep the cursor still and double-click several times in succession to get to an orignal-sized view of that area. Single-clicking on the image is similar, in that it sets a focus point on which the '+' and '-' buttons operate.

E.
(Enjoy.)