Number of Bars to use for Elliott Wave counts

[fhirsch1]

Hi -

I'm relatively new to using GET. It's been really useful for helping me to identify and enter trades, but I'm confused about something. My question is, how many bars should one use to obtain accurate Elliott Wave counts, and are there conditions under which one should use different numbers of bars? I posted this question to another bulletin board recently and an experienced person said it's a matter of preference, and that he uses 3,000 bars. However, if I use 3,000 bars, for example, and then look at some potential Type 1 sells that I just scanned for recently, the Wave counts I see placed on the screen are wildly different that what I get if I use 300 bars. So, it really doesn't seem to be a matter of preference, but a crucial issue for trade identification and trade entry. The help screen on esignal.com for 'Elliott Wave' states that 'To obtain the optimum wave count, we recommend using between 300-600 bars of data.' Can anyone offer any insight on this issue?

[gsp100677]

Hi fhirsch1,

I had many of these same frustrations when I started using AGET earlier this year. I have come to the conclusion that the reason nobody can give you a satisfactory answer on this question, is because nobody on these forums really knows how to get the kind of accuracy from a wave count (I suspect you're looking at intraday charts) that most people expect when they start using the software - hence the reason for all the additional tools for confirmation, such as GET Osc, MOB, Eclipse, etc....


Basically, The reason people say it's a matter of preference, is because, you won't get a perfect count of every price move, so the number of bars you choose to use, depends on the timescale over which you prefer to trade. The shorter you make the lookback period of bars, the more the focus of the count will be on shorter term moves, & longer term moves outside the lookback period of bars will not be labelled at all, as the software won't look any further back than that number of bars to come up with it's count.

Of course, there is also a school of belief that says the shorter the wave count, the less the accuracy. e.g. if you make the lookback period 100bars - the software only has the last 100bars of info to decide the wave count from & thus may frequently renumber the wave count.

On the other hand, if you make the wave count 3000 like the guy you mentioned in your post, you may discover that the inevitable miscounts that occur, don't get renumbered till much later, & the much shorter term price moves don't get labelled at all.

Therefore, I recommend you view what the wave count looks like when you have an extremely short lookback period (e.g. 50bars) versus an extremely long lookback period (e.g. 3000 bars) to get a feel for how this makes a difference to the count, then test the wave counts for lookback periods in between to determine the best/most frequently accurate count for the price moves that occur over the time period that suits your style of trading the best.

Two things to bear in mind while doing this are:
1. When using long lookback periods, make sure you actually have enough bars loaded in the chart (otherwise it can only look at what's loaded)
2. When testing these wave counts, bear in mind that the wave count you see on the chart only tells you what the software thinks the wave count is now (& from this you can judge if it labels price moves large or small enough, accurately enough for you), but it doesn't tell you how often it renumbered the wave count as those (now historic) prices unfolded - to view that you might want to consider the bar replay feature.

Also consider changing other parameters fo the Wave Count study in conjunction with lookback period, such as Wave1-3 Ratio, to obtain a more suitable/accurate wave count for you.
You might want to see the post "Q. Problems with Elliott Wave renumbering anyone?" for a few further tips on this (it's a couple of months old now so it might take a bit of searching).

Easy,
Gsp

[MR]

Hi,

I am an original employee of Trading Techniques, the company that developed the Advanced GET Elliott Wave (EW) algorithm. Often when there was foundational research to check out, Tom Joseph would assign me the project to test. We found out statistically the best consistent wave structures are identified within 300 bars. 300 has been the benchmark, standard, EW default setting since 1988.

We discovered up to 600 data bars are allowable in a calculation in some situations, and it would not skew results significantly over time. One example-- when trading an S$P contract on a 3 minute time frame, at times, it can be helpful to extend the wave count default to up to 600 data bars if the intra-day market is trending that day. Another exception usually has to do with checking something in a long trend. If you think previous pivot action could be a possible Wave 1 and 2, you apply the localize feature to check out those relationships.

Having used AGET since 1988, I have never changed the AGET Elliott Wave default setting. It is important to get in a habit of consistency.

I do use the EW "Localize" feature to experiment with ideas. This is where a “preference” becomes an issue. The more sophisticated you are in understanding Elliott Wave theory the better you can apply your own judgment in fine-tuning your understanding of a wave count.

No matter what, I still would always double-check the 300-default setting to make sure the wave relationship has not dramatically changed. If it does, it can be a concern.

If you use less than 100 data bars for your EW calculation, you might as well not use the algorithm. Personally, I contend if you use less than 180 data bars, you will generate inconsistent results.

As a futures trader I would double check a new front month with a good continuous contract file, primarily because it reduces confusion. Some new contracts have very limited data. The comparison can really help.

I am not sure where people get the idea loading up 3,000 data bars for the Elliott Wave calculation is good. The only thing I can think of is maybe there is confusion between loading up more data bars for doing better technical analysis, than loading up more data bars for checking out a Elliott Wave logic. There really is a difference. When doing AGET Education support, I use to tell people I personally load up to 3,000 data bars for my own analysis. But I thought I made it clear it was for better analyzing highs and lows, pivots, trends, and other patterns. It is purely for better standard technical analysis purposes. It had little to do with finding a better Elliott Wave logic.

The process of “cross-referencing” is much more important. It is a key ingredient to better success.

When Elliott Wave questions arise, we shouldn’t necessarily focus on changing the EW calculation start point, but we should be trying to compare and contrast the alternate wave counts and higher time frames with each other. We should be referencing the Elliott Oscillators with other time frames to see if there are degrees of strength building or weakening. We should be looking at the higher time frame for similar wave count synergy with the next two lower level time frames. These things are more important than playing with the Elliott Wave default setting.

Think of the Elliott Wave algorithm as a guide. Keep your setting to a consistent 300.

If you really want to experiment, take your database, put it in Training Mode. If you take the time to go through stock after stock, futures contract after contract. Start at the beginning of the data and advance it slowly using different settings, 300 or 600. Add some AGET tools you are trying to learn while doing this. You will be amazed how doing this enough will eventually help future trading, and you will grow more confident in the settings you are using.