Previous close = Always better results?

Research
1

Hi guys,

I have build a few fundamental simulations (nothing special, I value a stock at some point in the future at 150 and it is sold now at 100) and can’t help but notice that every time I change the transaction price setting to previous close prices the sim automatically has a higher average return. In almost all years.

The returns of next opening, next close and average of next high, low and 2x close are almost the same.

What is so special about previous close prices? I can’t figure it out. It doesn’t make sense to me. Some previous close prices of stocks will be higher, some lower. Why does it effect future returns so much?

Hopefully someone can give me some insight.

Thanks

2

Because there is no lag between signal and supposed execution of trade. It often takes investors and traders the weekend to assimilate the data and trade it on the next possible day. But ‘previous close’ assumes you can trade at the exact moment the signal is given which is not really possible in theory.

Your factor models or whatever include the data of the closing price. So to trade off that price isn’t really possible. Unless you had ranking systems that updated in real time and you could get all the shares you need in the last second of the day with normal slippage. Otherwise, you wait until Monday and trade alongside everyone else. Trading Friday’s close is kind of like a slight front-run.

3

I am sorry but I still don’t get it. Why are next days open and close prices (in most cases) (I looked at the past 15 years) higher then the previous day close?

A buy according to the porto123 simulations is always done better yesterday at close.

Even in down markets such as 2008. Shouldn’t buy at next open and next close return better results then a buy at previous close price, since markets were in a general downtrend so you the longer you waited, the cheaper you could buy.

4

If you have a strategy that has any kind of edge, then you will want to buy as soon as possible to avoid alpha decay. Even 1 day can make a difference. Furthermore, your actual formulas likely include the last close. You run a simulation that includes all kinds of weekend data. Maybe even earnings were released. But by using ‘previous close’ you essentially build a time machine and trade on data that you didn’t have on Friday’s close. So even look ahead bias is introduced.

5

Thank you for your reply’s hemmerling.

I get your point about alpha decay. But that only happens if people trade the edge right, so that alpha is less or even traded away?

So the fact that previous close prices are on average more profitable then next open or next close prices, is prove that people are trading this system, edge, factor or whatever you call it, right?

6

Maybe. But not necessarily. You could rank over-reaction. People might not necessarily be trading in the opposite direction and profiting from it. But if prices go too far too fast, human emotion kicks in and fear, price drop as people sell to lock in a profit and so forth. If you have a 14 day RSI signal and you wait a couple days, chances are you lose out on most of the action.

Also note the look-ahead bias. In simulations you are using all weekend data up to (but not including) Monday’s open. But you are trading Friday’s close. There is a lot of information there which is being traded on which you didn’t have on Friday’s close.

7

There are a couple of other angles you could look at this phenomenon from.

  1. If you’re trading Monday’s open or Monday’s high-low average, that will introduce a lot of noise into a system based on the pricing on Friday’s close because there are often huge gaps in price over the weekend. Noise tends to lower returns.

  2. When I place my trades in the morning based on last night’s prices, I want the price I’m buying or selling at to be the same or better than last night’s because that will maximize my profit. Stocks that are ranked highly are more likely to have increased in price from last night, which will make them less profitable to buy; stocks that are ranked badly are more likely to have decreased in price from last night, which will make them less profitable to sell.