PM End of Week Market Commentary - 3/2/2018

By davefairtex on Sun, Mar 4, 2018 - 7:20am

On Friday gold rose +5.60 [+0.42%] to 1323.70 on moderate volume, while silver climbed +0.05 [+0.27%] to 16.54 also on moderate volume. The buck fell again on Friday, down -0.42%, which tells us that gold's rally was just a currency effect.

This week the PM sector map looked a bit confused. The “other metals” (palladium/platinum/copper) group were hit hard, led lower by palladium, which dropped more than 5%. Silver and gold were the best performers in the group, with silver the only component that managed to move higher. Miners were in the middle, continuing to head lower.  All items are below their 50 MA lines, and all the mining share ETFs are below the 200. Only silver is above the 9, making silver's recent out-performance fairly easy to spot.  This is a relatively bearish configuration, especially for risk assets, because of the plunge in palladium and copper.

Name Chart Chg (W) 52w ch MA9 MA50 MA200 50/200 Last Crossing last
Silver $SILVER 0.27% -7.05% falling rising falling rising ema9 on 2018-03-02 2018-03-02
Gold $GOLD -0.51% 7.23% falling rising rising rising ma50 on 2018-02-27 2018-03-02
Junior Miners GDXJ -0.56% -8.11% falling falling falling rising ema9 on 2018-03-02 2018-03-02
Silver Miners SIL -1.61% -13.21% falling falling falling falling ema9 on 2018-02-27 2018-03-02
Senior Miners GDX -2.18% -2.01% falling falling falling falling ema9 on 2018-02-20 2018-03-02
Copper $COPPER -2.29% 16.55% falling falling rising falling ema9 on 2018-02-27 2018-03-02
Platinum $PLAT -3.14% -2.25% falling rising rising rising ma50 on 2018-03-01 2018-03-02
Palladium $PALL -5.25% 28.58% falling falling rising falling ema9 on 2018-03-01 2018-03-02

Gold fell -6.80 [-0.51%], with all of the loss coming on Tuesday. Friday's candle was a swing low, which had an 45% chance of being a reversal. Forecaster concurred, rallying +0.38 to +0.30, which is a buy signal for gold. The weekly gold forecaster remains in a downtrend.

The March rate-increase chances remains at 83%.

COMEX GC open interest fell by -18,711 contracts this week.  Commercials ringing the cash register.

Silver rose +0.05 [+0.27%] to 16.54, chopping sideways this week. Silver's candle print on Friday was a neutral short white candle, while the forecaster rose just +0.01 to -0.03, which pretty much confirms the sideways chop. The weekly and monthly forecasters both remain in downtrends.

The gold/silver ratio fell -0.63 to 80.05, which is bullish.

COMEX SI open interest fell -5,653 contracts.

The miners dropped again this week, with XAU falling -2.21%, and this was after a fairly significant reversal on Thursday. On Friday, the miners gapped up at the open and then sold off all day long, failing to print a swing low which looked to be a sure thing at market open. In spite of the Friday sell-off, XAU forecaster jumped +0.41 to +0.06, which is a buy signal for the mining shares. Weekly and monthly forecasters are both in downtrends, although the weekly improved markedly this week, up +0.22 to -0.06.  XAU remains below its 9 MA.

The GDX:$GOLD ratio fell -1.60%, and the GDXJ:GDX ratio rose +1.66%. That's neutral.


The buck rose +0.05 [+0.06%] to 89.59. The buck tried to break above the previous high at 90.29, but only managed to do so intraday, ending the week more or less back where it started. The buck did issue a sell signal on Tuesday, and by Friday the forecaster remained in an uptrend at +0.30. The weekly forecaster improved to -0.13, but that's still a downtrend, as is the monthly at -0.07.

It was a bit surprising that the Euro ended the week this strong; I would have thought the political uncertainty in Germany and Italy would have resulted in some selling pressure – but perhaps the well-connected big players already knew how it was going to turn out.  Certainly that's what it looked like on Thursday.

US Equities/SPX

SPX fell this week, dropping -56.05 [-2.04%] to 2691.25. SPX printed a bearish engulfing on Tuesday, and then sold off for the next two days, finally recovering on Friday. The thrusting candle pattern was somewhat bullish (28% reversal) but the forecaster ended the week at -0.46, after issuing a sell signal on Thursday.

The sector map looks relatively bearish; homebuilders, materials and industrials did worst, while telecom was the sole sector in the green. Utilities had a bad week too.  It is possible that the homebuilders are giving us a hint about where the housing market might be headed.

VIX rallied +3.10 to 19.59.

Name Chart Chg (W) 52w ch MA9 MA50 MA200 50/200 Last Crossing last
Telecom XTL 0.63% -1.17% falling rising falling rising ema9 on 2018-03-02 2018-03-02
Technology XLK -0.89% 28.33% rising rising rising falling ema9 on 2018-03-01 2018-03-02
Cons Staples XLP -1.12% -2.91% falling falling falling falling ema9 on 2018-02-27 2018-03-02
REIT RWR -1.83% -11.82% falling falling falling falling ema9 on 2018-02-27 2018-03-02
Healthcare XLV -1.98% 10.78% falling falling rising falling ema9 on 2018-02-28 2018-03-02
Gold Miners GDX -2.18% -2.01% falling falling falling falling ema9 on 2018-02-20 2018-03-02
Financials XLF -2.34% 14.45% falling rising rising falling ma50 on 2018-03-01 2018-03-02
Energy XLE -2.69% -7.17% falling falling falling falling ema9 on 2018-02-28 2018-03-02
Cons Discretionary XLY -2.70% 18.27% falling rising rising rising ma50 on 2018-03-01 2018-03-02
Utilities XLU -2.80% -5.30% falling falling falling falling ema9 on 2018-02-27 2018-03-02
Industrials XLI -3.28% 13.40% falling rising rising falling ema9 on 2018-02-28 2018-03-02
Materials XLB -3.80% 11.88% falling falling rising falling ema9 on 2018-02-28 2018-03-02
Homebuilders XHB -5.23% 9.91% falling falling rising falling ma200 on 2018-02-28 2018-03-02

Gold in Other Currencies

Gold fell in most currencies, with gold in XDR falling -7.68.

Rates & Commodities

The long bond moved mostly sideways, up just +0.19%. Bonds did well on Thursday, but lost most of the gains on Friday. TY did even worse, falling -0.05%. TY forecaster issued a buy signal on Thursday, but Friday's sell-off erased almost all of Thursday's gains. TY remains in a daily uptrend, but the forecaster closed the week at just +0.05. Weekly TY forecaster improved (+0.29 to -0.17) but remains in a downtrend, while the monthly forecaster continues to look quite bearish.

JNK mostly tracked equities, falling -0.72%, and issuing a sell signal on Thursday.

CRB fell -0.96%, with 4 of 5 sectors moving lower, led lower by energy (-3.71%).

Crude fell -2.14 [-3.37%] to 61.43. The 3 days of selling in crude roughly paralleled the drop in equity prices. Certainly part of the problem involved a bearish-looking EIA report on Wednesday [crude +3m, gasoline +2.5m, distillates -1m]. Crude did seem to find support down at the 61 level, and the doji candle print on Friday was neutral. The forecaster wasn't impressed either, falling -0.10, ending the week at -0.20. Crude remains in a downtrend.

Physical Supply Indicators

* Premium of SGE over COMEX was +8.12 vs COMEX.

* The GLD ETF tonnage on hand rose +4.72 tons, with 834 tons in inventory.

* ETF Premium/Discount to NAV:

 PHYS 10.77 -0.40% to NAV [down]
 PSLV 6.11 -1.88% to NAV [up]
 CEF 13.24 -3.12% to NAV [unch]

* Bullion Vault gold (!/orderboard) shows no premium for gold and a 1% premium for silver.

* Big bars premiums were: gold [1kg] 1.01% and silver [1000oz] 2.56%.

Futures Positioning/COT

In gold, the commercial net position rose 15k, with 8.4k from short covering, and 6.6k of new long buying. This wasn't a very substantial change. Managed money net fell by 25k contracts, evenly split between 13k longs sold, and 12.5k new shorts. Another 6 weeks of this and we might be nearing a COT low for gold.

In silver, the commercial net rose by 5k contracts, with 3.9k longs sold, and 8.9k shorts covered. The commercial net position is at its highest since mid-2015. Managed money net fell by 5.9k contracts, mostly it was about 6.4k new shorts, but also 476 new longs. Managed money net is now the most bearish on silver  ever in the history of the timeseries, which dates back to 2006.

A low for silver should be near.

Grey Swan Status

  • Italian Elections are this Sunday, with results known prior to the start of trading on Monday morning. The black swan outcome (courtesy of Mish): M5S and Lega Nord form a coalition to pull Italy from the Eurozone. Rumor has it that Italian polls are inaccurate; Lega Nord also suffers from “optics” problems, portrayed by the media as being just shy of Benito Mussolini in terms of political positioning, but given the 600,000 economic migrants and an unemployment rate north of 10%, people may well secretly vote for Lega Nord in private while decrying them in public.

  • The SPD announced the results of their election, approving the grand coalition with a surprisingly high 66% of the party members voting in favor. It looks like the currency markets were able to correctly sniff out the “Euro-positive” results of the SPD poll. This outcome takes the German grey swan off the table.

  • US Congressional Elections, 2018. The generic ballot shows Democrats 46.8% [+8.8%] Republicans 37.9%, a widening of 0.6% in the spread vs last week, entirely due to a decline for the Republicans.  A Democratic victory in 2018 would almost certainly lead to impeachment proceedings against Trump.

  • Mueller Investigation: no new developments, but a whole collection of rumors gleefully provided by the CNN/CNBC news axis, with the most interesting one being an investigation of Trump's business dealings in Russia prior to the 2016 campaign which allegedly might have given the Russians compromising material on Trump.


Risk assets and crude fell this week, but bonds were able to gain no advantage from the plunge, and the commodity group and PM moved lower even though the buck was unable to break higher. The Euro did not look particularly worried about a bad political outcome in Europe; it seems that the “no-gro-ko” rebellion in Germany's SPD was more hype than substance. Perhaps when their party moves into 5th place after the next election, they will see things differently. AfD will be leading the opposition.

Gold and silver big bar shortage indicators are showing no sign of shortage; premiums on big-bar gold and silver are normal, GLD tonnage rose slightly, while ETF discounts were mixed.

The gold COT report showed the start of a modest “COT rinse” cycle; commercials rang the cash register, while managed money dumped longs and went short, but at this rate it will take weeks for a low to form for gold. Silver shows managed money at the lowest net position ever in history, while the commercial net position is at multi-year highs. This indicates a low for silver should be upon us.

The fact that silver is performing the best of all the PM group supports the whole “buy silver at GSR > 80/COT low” thesis. In spite of the relatively ugly environment for PM, someone is out there accumulating silver – even if it is just paper silver at COMEX.

The positive aspect I see right now for PM is that the buck has failed to rally through 90.29. This is suggesting that Europe will avoid disaster one more time. The market could be wrong about this – we still have the Italian elections to go, but that's the way things are shaping up at the moment. I'm not sure I'd run out and buy a big bunch of mining shares, but certainly silver seems to be popular among the bankster group at the moment.

Next week we have Nonfarm Payrolls on Friday – that's about it for economic reports excitement.

Forecasting code – weekly – is pretty negative; virtually everything is being sold:

Uptrend: miners.

Downtrend: crude, copper, gold, silver, SPX, 10-year treasuries, BAA-grade debt, USD.

Note: If you're reading this and are not yet a member of Peak Prosperity's Gold & Silver Group, please consider joining it now. It's where our active community of precious metals enthusiasts have focused discussions on the developments most likely to impact gold & silver. Simply go here and click the "Join Today" button.


davefairtex's picture
Status: Diamond Member (Offline)
Joined: Sep 3 2008
Posts: 5874
yuuuge result for M5S

Current exit polls show the following results for the Italian elections:

  • M5S: 34%
  • PD: 18%
  • Lega Nord: 16%
  • Forza Italia: 14%
  • 4 smaller parties, combined: 11%

As a reminder, pre-election polls 2 weeks ago were (roughly): M5S at 28%, the PD at 23%, LN at 14%, FI at 16%.  Looks like PD lost voters to M5S, and there was a better than expected result for Lega Nord.

All this is in line with the "populism" trend.

And remember, this is during an expansion - a global growth phase.

MSM's headlines are "hung parliament", but the real story - I believe - is: "populism on the march."


davefairtex's picture
Status: Diamond Member (Offline)
Joined: Sep 3 2008
Posts: 5874
recession leading indicator: credit card charge-offs

Another tidbit courtesy of Wolf Richter.  This is a chart of credit card charge-offs - a coal-mine canary that has been a leading indicator for the past 3 recessions.

Note that there was one headfake in 1995, so let's say this one is only 75% reliable.  I like my chart better than his because it covers a longer time period.  It is curious how the smaller banks are showing worrisome credit signs before the big banks this time around.

davefairtex's picture
Status: Diamond Member (Offline)
Joined: Sep 3 2008
Posts: 5874
CGP Grey on Machine Learning

His latest video - actually 3 months old.  Its a pretty good description of what I do.

jussaumm's picture
Status: Member (Offline)
Joined: Feb 16 2016
Posts: 21
Machine Learning

Hi Dave,

That video is great; thank you for sharing. I have been thinking about your machine learning lately and . One of the questions that I have is, do you provide your bots all of the sources of data that you can get your hands on, even if you don't think that there is correlation between the data?

And then, how often do you tweak your bots? Is it a nightly build thing, where every morning you get a new "campion bot" for each question you are asking?

Just some curious thoughts; thanks for the work you are doing here.

jussaumm's picture
Status: Member (Offline)
Joined: Feb 16 2016
Posts: 21
Machine Learning

Duplicate deleted

davefairtex's picture
Status: Diamond Member (Offline)
Joined: Sep 3 2008
Posts: 5874
bots & correlations

Providing the network with irrelevant information always causes trouble.  It makes it harder for the net to learn.  I try to toss out as many things as possible and still arrive at a good answer.

I can check the progress of the training session and sometimes things just slowly get better the longer you train it, and sometimes they just don't.  The process is more art than science right now - educated guesses, trial-and-error.

The reason I know so much about what things correlate with what is because I pre-qualify my inputs by selecting the ones that correlate most strongly (either negatively, or positively) with the thing I'm trying to forecast.  Gold has a different set of inputs than silver, or copper, or SPX, or the 10 year treasury.  Its a really instructive process in how the different markets interact with each other over time.  At some level it represents the actual workings of the real economy.

I do retrain my models every now and then, which is why you'll see the forecast values change sometimes.   More (and more recent) data is always better.  That, or I'll have an inspiration and try to see if it helps the results.

I retrain the bitcoin/ETH models a lot, since there is so little data for them, they need every bit of improvement they can get.  Especially ETH.  Plus, I think the underlying market has changed pretty dramatically over the past year so I worry that I probably shouldn't be relying too heavily on the 2013 trading data. 

If I wanted to construct a "recession predictor" I might well include that credit card charge-off number.  Only problem is, the timeseries only goes back to 1985 - which means I wouldn't have enough a long enough series to give me any comfort that it would come up with a properly general answer.  I.e. it would properly predict the 1992, 2001, and 2008 recessions, but it wouldn't be general enough to predict the 2018 recession when it appears.  To do a real recession predictor, I'd want to use monthly data, and I'd (ideally) want it to go back to 1920, if possible.

Hmm.  Maybe I should try it.  :)

Nate's picture
Status: Platinum Member (Offline)
Joined: May 6 2009
Posts: 618
davefairtex's picture
Status: Diamond Member (Offline)
Joined: Sep 3 2008
Posts: 5874
the 2015-2016 recession

Ok so I tried constructing a recession predictor.  Fascinatingly, it predicted a recession that never happened...but SPX sure sold off.  And INDPRO dropped, and BAA rates rose - it looked and smelled like a recession was on the way.  But we never had a recession - at least not one that was recorded anyway.

Fascinating stuff.

jussaumm's picture
Status: Member (Offline)
Joined: Feb 16 2016
Posts: 21

Thanks, Dave. I guess it will take time before machine learning gets to a state where you can tell a bot what to optimize for and where to get information, then the bot decides what information to keep or ignore during it's continuous learning process?

I guess is good news from a Skynet perspective :)

bronsuchecki's picture
Status: Bronze Member (Offline)
Joined: Apr 22 2012
Posts: 97
Silver $1330.50

First sentence "while silver climbed +0.05 [+0.27%] to 1330.50 also on moderate volume" seems like the "buy silver when GSR>80" trade really paid off laugh

So Dave you are running a "builder bot slaughter house"? Not very PC, soon you'll be told you need to treat your special need bots equally.

davefairtex's picture
Status: Diamond Member (Offline)
Joined: Sep 3 2008
Posts: 5874
bots, etc


Yeah, while he makes them out to be cute little things, to me they are a collection of floating point values with connections, inputs, and outputs.  There is really no anthropomorphism, probably because they are really quite stupid at some level.  Great at finding patterns - as long as you are very careful to pre-digest the data for them, and toss away all the irrelevant stuff, and then scale the data properly.  Skip one of those steps, and things stop working - or work much less well.  No self-directed skynet yet.  I'm much more worried that some human will come along and hijack some "purpose-built bot" and direct it for their own purposes.

How would I even start to make one that learned on its own?  I have no idea.  They have a hard enough time learning when you pre-digest everything for them.  I think that requires another breakthrough that we haven't seen yet.  Probably a good thing.  I don't particularly want to throw in yet another factor into this crazy world.


Thanks, sometimes my editor doesn't catch my mistakes.  :)

We aren't yet at the point yet where people feel that "every neural net is sacred"...fortunately.  Where does artificial life begin anyway?  At conception?  Heh.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Login or Register to post comments