Fun and Profit; Predicting the price of GLD

You will find this accurate prediction of the weekly HI and LO of the gold ETF quite profitable. (2-3 times a year it will be off on both sides). It is useful for trading GLD or buying/selling of actual gold. Use your skills and imagination. YMMV

Good luck to all–cogman, the Blind Squirrel

For GLD the Blind Squirrel prediction of hi/lo for this week ending 9/29:

HI: 125.61
LO: 122.07

Each week the actual results will be posted:


Week Ending 9/29

ACTUAL HI: 124.66

ACTUAL LO: 121.55

1 Like

For GLD the Blind Squirrel prediction of hi/lo for this week ending 10/6:

HI: 122.04
LO: 119.45

Can you include the accuracy of past predictions?

Because, you know, predictions are very difficult. Especially about the future.

1 Like

Didn’t he, with the “ACTUAL” prices? Or you want a percent?

I think he means longer than just one week.

1 Like

Gotcha. Thanks.

I started he GLD statistical model in Nov 2013. The statistical software (SPSS) back tested the model at a correlation level in the hi .90’s. The algorithms are my own.

32 weeks ago I started tabulating the predicted vs the actual to check the back-testing. The correlation for the HI is .965 and the LO is .975. For those of you without statistical training, this means it is about 97% accurate. This is very high for financial quantitative predictions. 2-3 times a year it will be much further off on both sides. This is due to the tendency of financial data to have sudden spikes and sell offs that cause unpredictable “outlier” points in the data.

I’m without statistical training. Presumably the model requires input, and the output can only be as good as the input. What data are you feeding into it?

It’s a combination of statistical methodologies. The predictor variables ( what you called the data input) are custom and intellectual property.

It seems like what you are saying is that GLD trades within a narrow range, with some (3%, you say) extreme outliers. Those outliers likely make it disadvantageous to trading within that narrow range.

Also, the weekly price thing is weird. You are saying that if it starts at the “high” on Monday you should short it, just because it has 4 days to drop? What about if its at the high on Friday?

I use it as an option writing (selling strategy) or as a trading strategy on IB. Do not use it to buy options as you will likely lose money to the power of time wasting.

The most accurate prediction is daily, but it leads to the weary exhaustion of day trading. The model works for two-four weeks spans, even three months, but is less accurate. The best balance is one week for accuracy and profit.

Quantitative models are an aid to intuition. I usually trade when the price approaches the threshold. “Approach” is something you estimate on your own. If the threshold hits on Friday, I wait til the next week. Same goes for the weeks it does not approach target, as there’s always next week.

GLD is in a relative narrow range right now. It’s not usually volatile. Money is made gradually with this model. …and its not high risk. YMMV

If you’ve discovered a goldmine :punch: , why are you sharing it with the world?

For the same reasons we all share here.

1 Like

For GLD the Blind Squirrel prediction of hi/lo for this week ending 10/6:

HI: 122.04
LO: 119.45

Here’s last weeks results:


HI: 121.34
LO: 119.78

For GLD the Blind Squirrel prediction of hi/lo for this week ending 10/13:

HI: 123.16
LO: 120.66

For GLD the Blind Squirrel prediction of hi/lo for this week ending 10/13:

HI: 123.16
LO: 120.66

The actual hi/lo for the week:


HI: 123.86
LO: 121.55

So whats the prediction for week ending 10/20?

I build preliminary model on the weekend, then refine it on Monday. The market and gold prices on Monday mornng contribute to the prediction.

I have no idea what’s happening in this thread. GLD is supposed to be based on the underlying value of gold stored by the ETF fiduciary. Are you somehow predicting the price of Gold as a metal?

Or are you arbitraging NAV inefficiencies?

Personally, since Gold is part of my overall portfolio, I invest in the ticker CEF which is a closed-end fund that holds gold and silver. Historically, when gold is doing poorly, it trades at a 5% to 10% discount to NAV. And when gold is doing well, it trades at a 2% premium to NAV. The reason for this is that it’s a closed-end fund that cannot be liquidated by share holders for the underlying metal. Otherwise it would never drop to 10% below NAV since a hedge fund would come in and buy them all and liquidate. (Sprotts has done this with another fund and has been attempting to do it with CEF)

This is beneficial to me, since I’m only selling gold when it’s doing well, and buying it when it’s doing poorly, so I make money both on metal price and also on the discount. For example. I rebalanced into it at an 8% discount about 6 months ago and it’s now trading at a 2% discount as of today. All else equal with the underlying metal, I made 6% profit just on the NAV. This only works if you plan to buy/hold long term and rebalance specifically out of it when it’s doing well and into it when it’s doing poorly.

No idea what we’re doing with GLD specifically though. I’m highly skeptical anyone can predict the underlying gold metal price using any statistical models. If we’re doing some kind of NAV bid/ask arbitrage then I imagine that’s definitely possible but question why a huge hedge fund isn’t already doing it and extracting all of the economic rents for themselves as professional arbigeurs.