Backtesting and Hedged Portfolios

So, I backtested for all of 2017 using the latest model rules. The model makes money for the year, but there are several losing months, and the model underperforms the market. I could quite easily predict which months would be more profitable and which were more likely to be money losing by looking at their volatility. So, that idea works out of sample.

 

The graph shows the NASDAQ 100 index (close) for 2017 and the model return. The interesting thing is that a hedged portfolio of the market and the model, tracks the market quite closely. The hedging strategy would invest 75% of net worth in the QQQ ETF and use 25% of net worth to trade NQ futures with 3 times leverage according to the model. So it is 1.5 times leveraged with 50% of the total exposure long and 50% traded. Of course, you wouldn't really want all your portfolio into the the QQQ ETF. At least I wouldn't. But this is a step towards seeing what a realistic strategy with investment and trading would look like. Now if we look at 2018:

The hedged portfolio tracks the model, which vastly outperformed the market, closely instead now. It seems that you can get the best of both worlds with this strategy.

Soybeans

To try something different I tried a soybean daytrade. Made $107 in 17 minutes, so not bad :) But I only got the beginning of a much bigger move, so I need to be more patient in future trades i this commodity.

Volatility and Return

The graph shows the average true range (ATR) divided by the closing NQ futures price for all 14 day periods in 2018 so far and the average daily NDX model return over the same period. The correlation is very strong. The model tends to make lots of money when the market is volatile and potentially lose money when the markets are not volatile. This is why the model would have lost money in early 2017 for example and probably why the NASDAQ index produces better results than the S&P 500. Clearly, noise dominates signal when volatility is low. However, the correlation between recent volatility and future returns is quite weak. So, this isn't yet a useful tool for deciding when to trade and when not to trade on a daily or weekly basis. But if you were losing money for a while and volatility was low it would make sense to get out of that market and trade something else until volatility appeared to return.

I'm still holding the strategic long contract. It's up around $4k at the moment. I did a couple of tactical trades netting $215 and $980.

Trading on 2nd August

The market came with 4 points of stopping out the long position but then took off to the upside as soon as the New York market opened. Gain for the day was almost USD 3k. Today, I have added a second contract (@ 7391). Stop remains the same.

July 2018 Report

This month was the fourth month of the futures trading experiment. The first month was the model development phase, while May and June were about ironing out the glitches and training myself to trade the model properly (and not give in to gut instinct etc). In the first half of July I only traded one day and lost but model returns were good in the beginning of the month. Then in the second half of the month I got back into regular trading. Initially the model wasn't doing well but then things improved again as a short trade worked out.

The Australian Dollar fell from USD 0.7571 to USD 0.7432. The MSCI World Index rose 3.05% and the S&P 500 rose 3.72%. The ASX 200 rose 1.39%. All these are total returns including dividends. We gained 1.56% in Australian Dollar terms and 2.12% in US Dollar terms. So, we  outperformed the Australian market and underperformed international markets.

The best performing investment in dollar terms was Unisuper gaining AUD 4k closely followed by Cadence Capital (CDM.AX) gaining AUD 3.9k. The next best in dollar terms was Bluesky Alternatives (BAF.AX), gaining AUD 2.8k. The best performing asset class was "private equity", gaining 2.66%. The second best performer was US stocks, gaining 2.58%. The worst performing asset class was Australian large cap, gaining 0.41%.

The following is table of investment performance statistics computed over the last 36 months of data:

The first two rows gives the annual rate of return and Sharpe ratio for our investment performance in US dollars and Australian dollars. The other statistics are in comparison to the two indices. Based on beta, compared to the MSCI World Index we seem to be slightly geared, while compared to the Australian index we are less sensitive to market movements. We have a slightly positive alpha compared to the Australian and world markets. Finally, we now capture more of the up movements in the international and less in the Australian market and suffer less of the down movements in both the Australian and international markets.

This month I only made a small amount of money trading futures: USD 1.0k. The table compares my performance to the market and the model:

The US markets went up and then down. The model did outperform the market.* In the first week of July I didn't trade as I was in Japan and my phone wouldn't receive the text messages needed to log into the trading account. I actually received all these texts after returning to Australia! Then I traded long on a day when the futures price would suggest to be short and the index values suggest to be long and got stopped out. This made me do some more model research and revise the stops policy, though I found that index values provide better trading signals. After that I got back into regular trading trying to trade double the size but the model was losing at first. Then I started doing strategic and tactical trades, which helped psychologically.

We made more progress towards the new long-run asset allocation:

Total leverage includes borrowing inside leveraged (geared) mutual (managed) funds. The allocation is according to total assets including the true exposure in leveraged funds.

The improvement in allocation, came partly due to market movements and partly due to investment activity. We invest AUD 2k monthly in a set of managed funds, and there are also retirement contributions. Then there are distributions from funds and dividends. During the month, I also:

• I added another AUD 10k to the Winton Global Alpha fund, increasing the allocation to commodities.
• I added AUD 50k to the trading account and in the end was moderately successful at trading, increasing the allocation to commodities .
• I closed a small account with Colonial First State, which was invested in the CFS Geared Share Fund, reducing the allocation to large cap Australian stocks. 
• I rebalanced my CFS superannuation account, reducing the allocation to large cap Australian stocks and increasing the allocation to other asset classes.
  
• I bought 75,000 shares of BAF.AX increasing the allocation to private equity and real estate.
• I sold my position in PIXX.AX and bought a smaller amount of Platinum Capital (PMC.AX) and bought more as PMC fell further in price. This reduced the allocation to hedge funds.

* The statistics at the bottom of the table are based on only 4 months of data and so are not at all reliable yet.

Back to the Long Side

Closed short and opened long @ 7261.5. Stop is @ 7211.

The model switched to long today but the signal is close to remaining short and the intraday indicators are signalled short for the first part of the day. So, I kept the short position until now almost 7 hours into the trading day from 6pm New York time (8am Eastern Australian time). I put on a ten point trailing stop for two contracts, so it closed the short and opened a long simultaneously. Gained about 25 points on the short side compared to the market open.

Profit from the closed short was USD 3,000.90. The model gained 1.73% for the trade starting on 25 July and ending 1 August.

Monthly report coming soon.

Friday 27th July Trading

Finally we had some good downside on Friday with the NQ futures falling 126 points to 7299.75 – each point is $20 per contract. I had a strategic short from 7411.75 and a tactical short put on in the morning at 7429.5. My only regret is that I closed the tactical short at 7380 for only a $986 profit rather than setting a wider stop and letting it ride down for another $1600 in profit :) The strategic short is of course still in place as the model remains short. I am now up $885 for the month. Hopefully, I will stay up for the last couple of days of the month. I expect the market will go down further, both the model signals and looking at previous declines this year suggest that there is a lot further to go down before bottoming. All the previous declines went below the 34 day moving average and two went to the lower 34 day Bollinger Band:

If the latter happens, we would be at 6950 or so, $7,000 a contract from where we are now. Of course, given the strong trend it is more likely to be like the declines in April and June, which didn't reach the lower Bollinger Band.

Indications at the moment are that we don't want to do a tactical short during the US overnight on Sunday-Monday. A short near the US market open looks more likely to pay off.

Trading in 2018 is Objectively Better

I computed my average gain per NQ contract traded in 2006-2008 compared to in 2018. In 2006-08, on average I gained 0.46 points per contract traded or $9.31. Commissions on a roundtrip were $4.80 then. In 2018 so far, I made 3.69 points per contract or $73.74. Commissions are now $4.10 on a roundtrip. So, my trading now is almost an order of magnitude better. The average is brought down by lots of small daytrades I've done. As I plan to do fewer of those, the average should improve, I hope. On the other hand, the level of the index is now nearly 4 times higher than it was in 2006-8 and so a given percentage price move translates to more points.

This graph shows the equity curve on NQ trades - the actual number of trades is half this as there is one data point for each opening or closing of a position. Initially in 2006-7 I had a reasonably good increase in profits, peaking around $10k cumulative profit. Then there was a long slow decline into 2008 of a series of small wins, punctuated with larger losses. The big jump is the start of trading in 2018,  when I had a series of big wins. Since then, things have gone sideways, with losses equal to gains.

Tactical and Strategic Trading

After seeing a big profit disappear again a few times, I think I am going to adopt a combination of strategic and tactical trades now that I am trying to trade two contracts. One contract is always held in the direction of the model for as long as the model is long or short. This is the strategic trade. The other contract is in the same direction but can be closed out for the day when there is a big profit already. That is the tactical trade. Yes, day trading but the kind of daytrading where you put a trade on at the beginning of the overnight futures session and close it at the market open or vice versa. I had planned to do this but deferred it to stage 4 or 5 of the experiment. But I think I need the psychological boost now. I will make trading decisions using a chart with 2 to 3 hour candles. On a chart at that frequency most days break down into a rising and a falling period or a weak (when the market goes sideways) and strong period. I will close the tactical trade if it has made a profit and the next half of the day looks like being weak or going in the opposite direction to the model. Anyway, let's see if this works.

It probably was necessary to suffer through the pain of seeing a big profit on two contracts disappear a couple of times to be willing to have two contracts on overnight Australian time.... I tried adding one contract tactically before but was too nervous about it to set a wide enough stop.

Today the strong period was during the overnight (the market went down, in the model direction) and the weak period was during the US daytime when the market went up in the opposite direction to the model.

P.S.
I was just stopped out by the Google earnings report... Even more wishing I had closed one contract at the market open... This was a "tactical" rather than model stop. So, I got short again (tactically and strategically) at 7425.5 with the stop at 7441. This is very close, but was the second pivot resistance level when the model originally went short and so with the current model stop rules, that's where the stop stays.

Actually, the model is bit ambiguous today, but following the rules for these situations, we should still be short...

P.P.S.
I was just stopped out at the model stop. That means I'm out for today. Tomorrow morning I will re-evaluate the model direction. This is definitely looking like a losing month, similar to April, which was the initial model development month.

This model trade that was initiated on Friday lost 0.86%.

I researched the previous cases of similar ambiguous model signals so far this year.  There were only two previous cases, which were where the signal said to switch to short but was ignored because the turning point was from a value of the indicator that was close to zero. Both those times, staying long was the right thing to do. Maybe, in the absence of getting stopped out, staying short will turn out to be the right thing to do today. We will see. Either way, it is a very small sample to base any conclusions on.

The Kelly Criterion

There is a lot of incorrect information on the web about applying the Kelly criterion in the stockmarket. It is very different to applying it in a card game where you either win or lose a fixed amount. In that context the Kelly criterion tells you how much to bet on each gamble. But whether you are doing short-term trading or long-term investing that is not the case in the financial markets where there are continuous payoffs. In this paper, Ed Thorp lays out the Kelly criterion for investing in financial markets. It results in a rule of how much leverage to use when investing in a portfolio. That portfolio could be a buy and hold portfolio of stocks, or it could be a high turnover futures trading account. To determine how much of total net worth to allocate to a particular asset class or strategy is a different calculation. I think you should maximize the Sharpe ratio for your total portfolio. Where to set the stop loss in trading is a similar calculation - you want to use stop loss rules that maximize the Sharpe ratio for the strategy. I don't think Kelly tells you how much to risk on each trade in the way it can tell you how much to bet on each gamble.

The Kelly criterion isn't a practical rule in the real world as it requires you to continuously change the size of your position as you win or lose money. The suggested leverage for my trading model – this may be exaggerated because of too short a sample of returns and volatility – is greater than that allowed by the futures exchange. This amount of leverage would immediately blow up in the real world and result in huge amounts of commission and bid-ask spread payments...

Very Good Service from Interactive Brokers

We phoned Interactive Brokers about the login problem. They have a system issue. They set up the account so it can accept a temporary security code which they gave to us. We'll use this until they resolve the issue. The questions they asked to confirm our identity apart from a couple of the typical secret questions were what the net asset value in the account was, what position was in the account (short NASDAQ 100 futures), and what bank we use to transfer money to the account. If we had stolen a password we would have know two of those at least, because you can login into the account on a read-only basis with the password.

I managed to use the temporary security code to set up the mobile app which can produce codes even if it can't receive texts.

Can't Log Into Account

As of Saturday morning I am not receiving the test messages from Interactive Brokers that I need to log into the account. I sent a text to myself using Skype, so it is not the same problem that I had in Japan where I just can't receive texts. If this is still the case on Monday morning we will need to phone the broker to resolve this. They do have a mobile app that can generate the required login numbers even if you don't have phone service, but to set this up you need to get a text from the broker... At the moment this isn't a problem as the model is short for Monday still. In the worst case scenario, I can trade in the opposite direction using my own trading account until the stop is hit at NQ=7441. As my account is much older I have a physical security device - actually a bunch of codes on a card. But I will need to sell stocks/and or transfer money into my account to have enough margin to trade with. And this isn't ideal as profits are taxed higher in my account.

Switching to Short

The model has switched to short as at the open of today's Globex session (8am Eastern Australian Time, 6pm New York Time). I went short 2 contracts in an attempt to move to Stage 3 of the experiment.... The stop is at NQ=7441 and am short from 7383, so risk is relatively low (compared to what it might be), though the nearer the stop the greater the chance of hitting it...

Selling Everything

Well, in my mother's former account. Apparently the main (international) bank doesn't care that we the estate hasn't yet completed probate. Another local bank is, by contrast, very concerned about that. If we sell and go to cash, apparently we avoid paying this investment bank's very high fees. The account has returned practically nothing after fees in the last three years. August and September are historically bad months for equities (though only about 20% of the account is in equities). As we want to sell in the end anyway, it makes sense then to sell now. The plan is to hold everything in US Dollars in the interim.

Stopped Out Again

So, I put on my second trade of the month - long NQ - and was stopped out again, losing $1100 this time. The stop actually saved about $300 this time. But the model is still long for 17 July and so I put on a new long trade at 8:00am Australian time at 7320, which is up $190 at the moment. Stop is 7253 on this trade, currently at 7329.75. Down about $1700 for the month so far. NASDAQ 100 model is up 3.6% and NASDAQ 100 index 4.5% but I'm down 4.5% (due to two bad trades only and leverage). I'm determined to stick to the model now... let's see how I do.

Model Decisions for the Year So Far

The chart shows each short and long decision the model has made in the NASDAQ 100 index since the beginning of the year. The letter S or L is placed on the first day the model was long or short in each trade. So in theory you should get long or short at the previous close. Most trades were winners, though in late February, for example, the model got short on a big up day and then switched back to long the next day, which turned out to be the top. That long trade was also a loser. There were also stop outs along the way, which aren't marked here as new trades. Some times the model picks the exact top or bottom, at other times it misses it by a couple of days.

So, all I need to do is trade exactly like the model :) I put a new long trade on this morning.

New Investment: BlueSky Alternatives Fund

I had read back in April about BlueSky's battle with activist hedge fund Glaucus. As a result, the share price of the management company (BLA.AX) collapsed and they undertook a thorough review and independent valuation of all their investments. Listed investment company (closed-end fund) BAF.AX, is a fund of funds, investing in BLA managed investments in real estate, private equity, agriculture, and water rights. The price of this fund also fell, though not as dramatically. The valuation of all but one of its investments is now complete and the net asset value is AUD 1.13 per share. On Friday the stock was trading around AUD 0.80. The company is buying back a lot of stock which is supporting the price. I made an initial investment today and could add more if my thesis that it should rise, plays out.

Position Size

One of the main ideas in traditional momentum trading wisdom is that as volatility increases your position size should decrease. This is one of the key ideas in the Turtle Trading System, for example. If you have no idea what will happen, then higher volatility likely will result in higher losses as well as higher gains.

But if you do have some ability to predict the future, that trading signal might be stronger when volatility is higher and weaker when volatility is lower. Then you will have more losing trades when volatility is low and a higher proportion of winning trades when volatility is high. This seems to be the case with my system. Higher volatility means higher risk but also a higher probability of being right. In this case, position size maybe should be constant regardless of volatility.

P.S. 22 July

I calculated the Sharpe ratio for constant position size and for strategies that reduce position size as volatility increases and and increase position size as volatility increases. The constant position size strategy has the highest Sharpe ratio confirming my intuition. The strategy with a negative correlation between position size and volatility has the lowest Sharpe ratio. The strategy with a positive correlation is in between. So, for the moment I will stick with constant position sizing.

Turtle Trading

I have been reading the Complete Turtle Trader, trying to get some inspiration. Back in the early 1980s, futures trader Richard Dennis hired a bunch of relative novices (some actually had trading experience) and taught them a trend-following method of trading. He then got them to trade some of his assets using the methods. The idea was to see if trading could be taught. During the next few years, many of them generated extraordinary returns, as documented in the book. Then the experiment ended after Dennis suffered major losses and shut his fund.

Some of the "turtles" went on to run their own investment firms. The star pupil seems to be Jerry Parker who founded Chesapeake Capital. However, subsequent performance has not really been that good.* The fund has underperformed the S&P 500 and has had about twice as much volatility. Taxes would be much higher on Chesapeake's strategy than on buying and holding the index. Why does voltatility matter? Because I could have used leverage to invest in the S&P 500, increasing volatility to the level of the Chesapeake Capital fund, but increasing returns far beyond its returns.

This doesn't encourage me to adopt a long-term trend following strategy. The assumption of this kind of model is that the future is entirely unpredictable... Eckhardt is cited in the book as saying that random entry into a trade is just as good as long as you follow exit rules. That's true about most momentum trading strategies I think.

It's notable that none of these turtle related firms are very big in terms of assets under management.

* The "LV" fund performed better but still underperformed the S&P 500 on a risk adjusted basis. Salem Abraham's – described as a "second-generation turtle" in the book – fund has gone nowhere in the last ten years.

The Index Gives Better Trading Signals than Futures Prices Do

It turns out that the NASDAQ 100 Index gives better trading signals than the NQ futures prices themselves do. I think the reason for this is that most trading takes place when the stock market is open and that is usually when big moves happen. The "out of hours" trading is mostly noise then reflecting what is happening in other stock markets and after hours earnings reports etc. The futures prices still provide signals that "beat the market" but not as well.

I did find again, that stops mostly detract from performance and I am introducing a new stops policy. When we change direction we set the stop loss at the the second pivot support for a long or the second pivot resistance for a short. We then keep that stop until either the direction of trade changes or we are stopped out. This results in far fewer stop outs.

It's likely that in commodity markets such as oil or gold the futures prices do provide good trading signals. Well, there isn't anything else to use anyway.

Futures Prices vs. Index Values

I didn't trade while I was in Japan because my mobile phone wasn't receiving the text messages I needed to log in to my trading account. When I got back to Australia I dithered about getting back in for a couple of days, missing a nice rally. Then this morning I decided to make the plunge (on the long side) and 2 hours later I was stopped out. Apparently there is negative news on US tariffs on trade with China.

After the cash market closes at 4pm New York time, the stock index futures trade for another hour before closing for one hour. The futures closing price can, therefore, be quite different to the index closing price. This was the case today where the futures plunged around 30 points in the last ten minutes of the futures trading session. Using the index data for the 4pm close, my model said to stay long. However, if we had knocked 30 NASDAQ points off to reflect the futures closing price, it would have switched to short. So, I think I need to get historical futures data and re-estimate my model with these. I should be able to get these from Quandl. An additional advantage of using futures prices is that I can do the analysis one hour later -  currently from 7am Australian Eastern time rather than 6am Australian Eastern time. The futures market then shuts for an hour and reopens at 6pm New York time or 8am Australian time.

However, on Saturday morning the futures market closes at Friday 5pm New York time and then doesn't reopen till Monday morning at 8am in Australia. So, I will need to do the analysis with index closing data before 7am on Saturdays unless I want to get stuck in possibly the wrong direction over the weekend.

June 2018 Report

This month was the third month of the futures trading experiment. The first month was the model development phase, while last month was about ironing out the glitches and training myself to trade the model properly (and not give in to gut instinct etc). It turned out that this month was more of the same and I am still on the second stage of the experiment, which is learning to consistently trade the model and iron out the glitches. The third stage is to reach a level of profits equal to my salary, while the fourth stage would be to maximize returns beyond that. I had planned to move to trading two contracts this month, but mostly traded one contract still.

June is the month when our Australian managed funds pay out their main distributions at the end of the Australian financial year. These usually have large tax credits associated with them. In this report, I have estimated the likely tax credits, which won't be known till later in July.

The Australian Dollar fell from USD 0.7571 to USD 0.7391. The MSCI World Index fell 0.50% and the S&P 500 rose 0.62%. The ASX 200 rose 3.63%. All these are total returns including dividends. We gained 3.16% in Australian Dollar terms and 0.71% in US Dollar terms. So, we  underperformed the Australian market and outperformed international markets.

The best performing investment in dollar terms  was CFS Geared Share Fund gaining AUD 23k. The next best in dollar terms was IPE, gaining AUD 19k. The best performing asset class was "private equity", gaining 7.79%. The second best performer was Australian large cap stocks, gaining 3.21%. The worst performing asset class was hedge funds, losing 0.46%, the only asset class that lost money.

The following is table of investment performance statistics computed over the last 36 months of data:

The first two rows gives the annual rate of return and Sharpe ratio for our investment performance in US dollars and Australian dollars. The other statistics are in comparison to the two indices. Beta expresses the change in investment returns for a 1% change in the market. Compared to the MSCI World Index we seem to be slightly geared, while compared to the Australian index we are less sensitive to market movements. Alpha shows the risk adjusted excess annual return. This is how much we are beating the market (or not) adjusted for risk expressed as beta. We have a slightly positive alpha compared to the Australian and world markets. Finally, up capture and down capture breaks beta into the response to positive and negative months in the stockmarket. A greater up capture than down capture ratio is desirable. We now capture more of the up movements in the international and less in the Australian market and suffer less of the down movements in both the Australian and international markets. A hedge fund like return would show this positive skew and a positive alpha. We show some hedge fund like properties across the markets.

This month I only made a small amount of money trading futures: USD 1.2k. The table compares my performance to the markets and the models:

The US markets went up and then down, ending quite flat. The models did outperform the market.* Through a series of missteps I performed worse than the models given that I was using leverage. This is mostly because I picked the wrong contract to trade with for some of the time. I think one way to trade in strongly trending markets is to act more tactically, trading in the direction of the model when other short term indicators (using a chart with 3 hour candles) show it is advantageous and then closing the position when the odds move the other way. More than once I was up USD 2k and then gave it all back... On 21 June I did exactly the wrong thing, throwing in the towel for the day and closing my short just as the market was about to reverse and go down... Seeing that happen did increase my faith in the model a little bit more. Gut instinct is not as good as the model. But then the same thing, kind of, happened on the last day of the month. The models were short, the market went up, but I capitulated at almost the worst point, because at the market close the indices were way down from the highs.

The best I can say is that I didn't lose money for the month as a whole. So it looks like more of the same for next month. I'll try to trade one contract exactly according to the model and one tactically.

We made a little more progress towards the new long-run asset allocation:

Total leverage includes borrowing inside leveraged (geared) mutual (managed) funds. The allocation is according to total assets including the true exposure in leveraged mutual funds.

The improvement in allocation, came partly due to market movements and partly due to investment activity. We invest AUD 2k monthly in a set of managed funds, and there are also retirement contributions. Then there are distributions from funds and dividends. During the month, I also:

• I shifted money out of trading when I shifted the account I am trading with. This decreased the allocation to commodities.
• Added another AUD 10k to the Winton Global Alpha fund, increasing the allocation to commodities.
• I sold 500,000 shares in IPE and bought a small amount of OCP.AX, reducing the allocation to private equity.
• I sold some Platinum Capital (PMC.AX) and bought a lot of PIXX.AX, which is the equivalent ETF, because PMC was particularly overvalued. This increased the allocation to hedge funds.

* The statistics at the bottom of the table are based on only 3 months of data and so are not at all reliable yet.

Trading Update: Tokyo Edition

So far it looks like the "old" more systematic model won today. US stocks are down. I'm not trading as I was flying to Japan and now my phone's connectivity is dodgy and I need it as a security device. Anyway, Wednesday is US Independence Day and Tuesday is already a short trading day so, I'll wait till Thursday or when I am back in Australia next week,

My hotel is the blue tower in the background.

P.S.
The stockmarket turned and went up. So, the "new" model was vindicated in the end. Still, I don't like trading rules that don't make logical sense.

Changing the Model Back and Researching the Model

I'm not happy with this model rule that I added on 27th April. It's not very systematic. So, I dropped it, while keeping rules added more recently. The stock market switches between two states - trending and cycling - and the model uses different indicators in each state. Deciding what state we are in is a little problematic. So, I am going to have another look at the rules for that and also possibly smoothing out very small moves, which are noise. The rule I just dropped was ad hoc - it switched to the indicator for the cycling state when we are in a trending state based on one period lagged positive performance.

Monday is a good test of the two algorithms - the simpler model is short and the model with the "ad hoc" rule is long. Let's see which wins.

I also thought of using the volatility indices, VIX and VXN, as indicators. The basic idea is that volatility is highest at bottoms in the market. But I couldn't see a way to do this. This graph shows why:

VXN (and VIX) rose throughout January while the stock market rose too. This was a warning sign that a correction was coming. But a model that shorts the market when VXN is expected to increase would have lost money all January. Also, my existing forecasting model is no good here...

Gold 2048: The Future of Gold

This report seems bullish for the price of gold over the next 30 years. Continued growth in India and limited gold discoveries recently seem bullish. On the other hand, some regions haven't been explored much and technology could enhance extraction, though the latter likely balanced by increased environmental restrictions.