In the past, I met a lot of traders, the live results were completely different from the back test results. The cause was apparently futility – too little stop-loss. I’ll explain to you today why this can be a problem, what it should be, and how to avoid that risk. The next topic is about these strategies that are using the STOP order to open a position and at the same time are using too little stop-loss (this article is not about strategies that use the market order). What is too low a stop-loss? Well, it depends on the market and time. But in general, our average time frame is smaller than the size of an average bar. I will give you an example. If we are using an average 30 minute chart of $ 250 and stopping our $ 80 loss strategy, we are going to get into serious trouble. The live trading results (and in most cases will almost certainly be) can be very different from the ones we have from the completely back test. Let’s see why.
This problem occurs when the stop loss is so small that some trades have the same order of entry and stop loss in the same bar. Suppose we have a STOP order at price 100 as well as a stop loss of 99. Now, imagine that the bar opens at 98.7, goes to 100.1 and we open a long position – and the stop-loss is set to 99. And all of this happens on the same bar – that is, inside this bar, the input order is activated, the position is opened, and the stop loss is set.
It is now important to understand why this can be a dangerous problem. It’s very simple. There are many retrospective analysis platforms that are not able to detect whether the data is incorrectly configured or the data resolution is not adequate enough if the loss is stopped or not. In other words, there are certain situations where the stop-loss actually opens the position and occurs immediately after the entry order is activated because the market starts to move south. However, our back-testing platform is considered profitable trading (from now on I will write about TradeStation because it is mainly the platform I use). How is that possible?
Let’s continue with the demonstration of the situation described above. In this situation we can see a rising bar, which is a price close above the open price and at the same time close to the highest close. It is assumed that the bar was rising all the time and TradeStation assumes that the “inner” movement of the bar, i.e. the way the bar was created, was constantly rising, in a straight line.
TradeStation follows the logic, which is that when the high bar was closed the process of creating that bar was on the rise. In this situation, TradeStation assumes that the bar opened at 98.7 and the price steadily rose to 100.4. In the meantime, he activated our order for over $ 100.
However, the hypothesis is very vague and dangerous. What if the bar went up for the first time, our buy order was activated, but then it reversed and fell again, below our stop loss, and then began to rise again to approach the highest?
It’s a completely realistic scenario that happens every day, which would result in a clear loss (as soon as the position is opened) – and yet it is defined as if TradeStation (and possibly other software as well) doesn’t define the situation. any correction within the bar. So no stagnant loss has occurred and the trade has been profitable. This is the cause of serious problems. The back test clearly shows a lot of profitable trading, which would really be a loss, and this strategy starts trading directly and then everything starts to fall apart …
Fortunately, the situation is not as serious as it looks and the platforms behind the study generally take this risk into account.
The first protection against this threat is simple and effective on a certain level. TradeStation calls it LIBB (Look-Inside-Bar-Backtesting), while others call it different names, such as Bar Magnifier. The fact is that when activating this function the program looks inside the bar to the level of the best available data resolution (in most cases it is 1 minute), if there is no internal correction after activating the input command. , or if there has been a correction in the same bar when we have entered and hit the stop-loss.
While this may seem like a great solution (it’s now the standard part of most platforms), it shouldn’t be enough when it comes to small losses. Why? Imagine your state loss when it is 80 USD, but the average bar for your best LIBB resolution (i.e. especially 1 minute) is 150 USD. In this case you are having the same problem described above, when the platform cannot determine whether or not the loss has stopped inside the bar, and again, only the inaccurate approaches driven by the previous ones are made. if the closed logic bar closes closer to the lower or closer to the highest. In other words, you’re back at the beginning and with too little loss, LIBB won’t help you either and the problem continues.
So we’re getting to the point where we need to dig a little deeper to solve this problem. One solution would be to use an even finer data resolution – up to the markup level. But this is not as easy as it seems. First, the history of tick data is not so easily retrieved, or in a very short period of time. And if that data is available, it’s really expensive. But if you still buy tick data, you have to fix a number of technical issues – since tick data tends to be very large, most platforms won’t handle so much data, crashes or back tests are incredibly slow (I can confirm).
So we need to use a much easier solution – and that’s a pretty big stop-loss to use. And what is a relatively large stop-loss? Simply use a stop-loss that is 1.5-2 times greater than the largest one-minute bar on your board. It’s simple and you can avoid many problems. For example, if the largest one-minute bar in your data history is $ 300, use a stop-loss of at least $ 450. Time. It’s easier and safer to get used to bigger stop losses than to lie to ourselves and then wonder why a nice backtest equity is against the results of live trading.