Pivot Point Analysis/Trends

[james6848]

Hello all,
Can someone please give me a detailed run down of pivot point analysis (the method where you end up with a pivot point, R1, R2, S1, S2 etc.).
I'm a bit confused with these specific issues:
1. What exactly is a pivot point? (sorry!)
2. Why TWO support and resistance points?
3. What is the best timeframe for using these calcs (hourly? daily? weekly? etc.)?
4. Are these calcs just a more complicated way of deriving trend lines?
5. Does anyone actually find them an effective forecasting tool?

I found this pivot point calculator if anyone is interested: http://www.futurestrade.com/

I'd welcome any general comments on the subject.
Thanks.

James

 

[rossored]

Hi.

1. A pivot point is a calculation of relative value based on the price activity of the prior day. It gives you a "balance point", as well as potential s/r levels.

Generally, the calculations for pivots and s/r levels are:

Pivot = H + L + C (High, Low, Close prices) / 3
R1 = (Pivot * 2) - L
S1 = (Pivot * 2) - H
R2 = (Pivot - S1) + R1
S2 = Pivot - (R1 - S1)
R3 = (Pivot - S1) + R2
S3 = Pivot - (R2 - S1)

Personally I think they're a bit subjective but they work better on some contracts than others - ie emini S&P futs, because pivots are generally used by pit traders who do the actual buying and selling of futs contracts in the pits.

2. Why not? How about 3?

3. Daily pivots work best on a daily basis. I dont think they're used on other timeframes at all.

4. Nope, nothing to do with TL. Pivots are just a calculation for estimated s/r levels.

5. Sometimes. Like everything else, some of the time they're real accurate, some of the time they ain't, and the market will motor right through PP, R1, R2, S1 or S2 without so much as a blink.

HTH.

Matt

 

[james6848]

Thanks Matt.
Regarding point 4: I thought trend lines were used to derive potential s/r points. Why is this different?

James

 

[dbphoenix]

Hi.

1. . . . as well as potential s/r levels.


5. Sometimes. Like everything else, some of the time they're real accurate, some of the time they ain't, and the market will motor right through PP, R1, R2, S1 or S2 without so much as a blink.

HTH.

Matt

Key word for "1" being "potential". As for "5", not that I've done an exhaustive statistical analysis, but pivot levels, like all such mathematically-derived S/R levels, "work" most often when they coincide with the previous day's high, low, and/or close. Otherwise, there can be an enormous fudge factor, though when one applies multiple pivot levels including half-stops along with Fib and so forth, price is bound to hit something somewhere.

As for the trendline thing, trendlines are used to indicate trend. Whatever value they may have as S/R is likely to be incidental to whatever S/R is provided by price S/R. But the post isn't about trendlines, so I'll leave it at this.

 

[Roberto]

5. Does anyone actually find them an effective forecasting tool?

The honest and short answer is that that doesn't matter.

What might matter is whether a significant number of people find them a significantly _more_ effective forecasting tool, over a significantly large selection of data, than randomly selected levels are over the same data.

To put it another way: yes; many people do believe that they're an effective forecasting tool. And when medical researchers do a double-blind controlled clinical trial, the "placebo group" (patients who have taken something like powdered milk tablets with no pharmacological activity) _always_ shows significant improvements in their condition, when compared with the "no treatment" group.

Why do you think that is? The reason might be the same reason that many people believe that pivot levels are an effective forecasting tool. They are actually using them and making a profit, so they "must" be, no? Well, no ... not necessarily: the reality is that if you compare the placebo with no treatment (compare pivots with no pivots), it's very easy to "prove by results that it works".

What you actually need to be able to do to prove that "Product X works" is to prove (statistically and with valid methodology and all the rest of it) that it works better than the powdered milk tablets. And that's where a wheel so often comes off.

I'd love to know if someone has ever proved convincingly in that sort of way that pivots work better as effective forecasting tools than randomly drawn lines. Call me a cynic, but I seriously doubt it.

 

[commanderco]

Hello all,
5. Does anyone actually find them an effective forecasting tool?

James,
There is a very simple way to find out.

Plot the Pivot Points on a chart or Turning Points, Breakout points,
cross overs, or any other well known method of establishing points.

Bear in mind, that if you enter a stop into the market at these points,
it can be seen.

Watch the number of times that the market reaches these points only to turn
(fade)

If the number of fades is reasonable then the tool is effective, but not
necessarily in the way that you were thinking.

 

[blackcab]

I'd love to know if someone has ever proved convincingly in that sort of way that pivots work better as effective forecasting tools than randomly drawn lines. Call me a cynic, but I seriously doubt it.

Well let's do an experiment then, on YM. It shouldn't be too difficult but we'll need to agree some things first:

1. What time of day shall we take as the open and close?

2. What shall we take as the high and the low - the absolute H/L or a pip or two either way if there was 'obviously' more action there, if the absolute H/L was a single tick for example?

3. How shall we distinguish between significant & insignificant action at each pivot level? You could see if all the day's prices that were within say 1 point of a pivot level were consistently more frequent than other prices within the day's range. Or you could see if the reversals of the direction of a short moving average that occured within say 1 point of a pivot level were consistently more frequent than reversals at other levels.

I'd like to see the results of this so will do it anyway. Any thoughts?

 

[aarie5]

Hi James

What time frame will you trade ?

 

[Roberto]

Well let's do an experiment then, on YM.

It's not something I have access to, but it sounds interesting.

1. What time of day shall we take as the open and close?

Why would we need to change those?

2. What shall we take as the high and the low - the absolute H/L or a pip or two either way if there was 'obviously' more action there, if the absolute H/L was a single tick for example?

I think probably it wouldn't matter too much either way, as long as you're consistent?

3. How shall we distinguish between significant & insignificant action at each pivot level? You could see if all the day's prices that were within say 1 point of a pivot level were consistently more frequent than other prices within the day's range.

That would certainly be an interesting finding. You might still have to explain to me how it would necessarily make for a useful forecasting tool ...

Or you could see if the reversals of the direction of a short moving average that occured within say 1 point of a pivot level were consistently more frequent than reversals at other levels.

Again, I can see that that would certainly be interesting ...

I'd like to see the results of this so will do it anyway.

Ok, well, I for one will certainly be interested in the results.

Any thoughts?

I suspect that to prove anything, you might really have to have a large amount of data and have it independently assessed in something simulating real-time (i.e. something with a "right-hand edge") by quite a large number of "traders". I realise, obviously that this isn't going to be feasible or practicable for you to do, and I'm in no way trying to prepare the ground for argumentatively saying "inadequate methodology" afterwards, because I really _will_ be interested in looking at what you can come up with, and I really _am_ willing to be wrong about this (that's never a big deal for me, as you can see from plenty of my posts), but I'd really be grateful if you could explain just a little more about how what you're proposing to test will lend credence to the "pivots make effective forecasting tools" theory.

What I'm inarticulately trying to drive at here is the possibility that (so many "indicator-based" styles of trading being either very slightly self-fulfilling or even very slightly self-defeating) you might prove relatively easily that there are real "statistical anomalies" around certain pre-set levels (which perhaps many traders work out in broadly speaking the same way), but not at the end of it all be able to show that what you've demonstrated has anything to do with being an effective forecasting tool. Does this make sense, or rather, do you see what I mean?

 

[SOCRATES]

May I compliment you on puttting it so very succincly and tastefully.

Kind Regards As Usual.

 

[nine]

Seeing we are not talking about trendlines I will stick to pivot points. Personally I have found them to be completely useless on some contracts. On others I have found them to be interestingly useful and will give one example where they "work."

In one index future you will frequently get a fast break upwards. When this occurs, if there is a pivot point 15-20 points from (my definition of) the initiation of the break it is likely to stop and significantly reverse at that point. Stop means for 3 points before until about 7 points after the pivot. As with all resistance a stop loss is required for those times where it is "wrong".

There is another interesting characteristic of these pivots .... when it doesn't stop it will often pause at the pivot point (on a short timeframe chart) and then on its second attempt break through with a gap that results in a large amount of slippage --- the worst I have personally experienced being 23 points beyond my stop.

Both the reversals and the big gaps suggest to me that this is "true" resistance for this contract.

While you are looking at pivots its worth looking at the Camarillo (spelling ?) pivots which also seem useful in some cases:


All my pivot formulae as coded in C for SierraChart
bottom ones are camarilo
sh & sl are just the low and high

{pp=(sh+sl+sc)/3;} else {pp=(sh+sl+so+so)/4;}

r1=2*pp-sl; s1=2*pp-sh; r2=pp+r1-s1; s2=pp+s1-r1; r3=pp+r1-s2; s3=pp-r2+s1;
r4=pp+r2-s2; s4=pp-r2+s2; r5=pp+r2-s2; s5=pp-r2+s2;
r05=(pp+r1)/2; r15=(r2+r1)/2; r25=(r3+r2)/2; s05=(pp+s1)/2; s15=(s2+s1)/2; s25=(s3+s2)/2;
clb=((sh-sl)*(1.1/2))+sc; cs=((sh-sl)*(1.1/4))+sc; cl=sc-((sh-sl)*(1.1/4)); csb=sc-((sh-sl)*(1.1/2));

 

[james6848]

Thanks for all that everyone. It would indeed be interesting to apply some calculations to 'live' days and see what happens (I've noticed how easy it is to notice patterns AFTER the event! ;-) ).

aarie5 - in answer to your question about what timeframe I want to trade: in all honesty I don't know, I'm just trying to learn about things at the mo. In an ideal world I would like to deal with shorter time periods (but possibly not daytrading - too stressful!). I guess I am looking for small but consistent and regular profits (me and a million others!).

I must say, it's easy to have information overload in this business, ain't it?

 

[SOCRATES]

You're not kidding, but in the end analysis, all of it is very simple.

 

[james6848]

You're not kidding, but in the end analysis, all of it is very simple.

It is simple...




...once you know it!

 

[blackcab]

Roberto - good points. At this stage I'd just be looking for the presence of any outstanding behaviour at pivot levels. If there's none, that's that. If there is, I'd just try and characterise it simply (eg "congestion often happens at them but reversals don't") to feed into my general knowledge. I wouldn't aim for great precision in the results for various reasons, but you have to be precise in the measuring process even so. I wouldn't be interested in trying to use them to forecast price, but that's a different k of f.

 

[dumbowl]

...

{pp=(sh+sl+sc)/3;} else {pp=(sh+sl+so+so)/4;}

r1=2*pp-sl; s1=2*pp-sh; r2=pp+r1-s1; s2=pp+s1-r1; r3=pp+r1-s2; s3=pp-r2+s1;
r4=pp+r2-s2; s4=pp-r2+s2; r5=pp+r2-s2; s5=pp-r2+s2;
r05=(pp+r1)/2; r15=(r2+r1)/2; r25=(r3+r2)/2; s05=(pp+s1)/2; s15=(s2+s1)/2; s25=(s3+s2)/2;
clb=((sh-sl)*(1.1/2))+sc; cs=((sh-sl)*(1.1/4))+sc; cl=sc-((sh-sl)*(1.1/4)); csb=sc-((sh-sl)*(1.1/2));

:?: :?: :?:

r4 = r5

and

s4 = s5

 

[Roberto]

Roberto - good points. At this stage I'd just be looking for the presence of any outstanding behaviour at pivot levels.

IMHO doing that is just comparing the "placebo group" with the "no treatment group". With the added complications of anything self-fulfilling and/or self-defeating.

The placebo group _always_ does better, but this honestly doesn't prove anything at all about the new drug being tested!

Sorry if my analogy is not the easiest one to follow. But honestly, it's not possible to prove that they can be an efficient forecasting tool by doing that sort of "research". It's only possible to convince the great majority of people who haven't really thought it through and don't quite understand evidence, probability and proof.

 

[blackcab]

IMHO doing that is just comparing the "placebo group" with the "no treatment group". With the added complications of anything self-fulfilling and/or self-defeating.

The placebo group _always_ does better, but this honestly doesn't prove anything at all about the new drug being tested!

Sorry if my analogy is not the easiest one to follow. But honestly, it's not possible to prove that they can be an efficient forecasting tool by doing that sort of "research". It's only possible to convince the great majority of people who haven't really thought it through and don't quite understand evidence, probability and proof.

I'm not sure what you mean. Let's forget forecasting - my idea is simply to see if price has historically done anything consistently statistically "unusual" at pivot levels. Given the relative ease of doing that analysis, I can't see that one would be anything but better informed having done it, regardless of the results, therefore it's worth doing. It might be turn out to be useless, and IMHO it's nothing to do with forecasting, but there are other reasons for doing it.

I probably won't get time to do it anyway!

 

[Roberto]

I'm not sure what you mean.

That's my fault, sorry. Sometimes I ramble a bit incoherently ...

my idea is simply to see if price has historically done anything consistently statistically "unusual" at pivot levels. Given the relative ease of doing that analysis, I can't see that one would be anything but better informed having done it ...

I understand that it's relatively easy and that that's what you were proposing to do.

My contention is that on the question of whether these things can make effective forecasting tools (which was how the thread started) you will actually be _NO_ better informed from it at all, just _feel_ better informed because you'll perhaps know something you didn't know in the first place.

I probably won't get time to do it anyway!

That's often the way with these things! Sometimes these things can be like a drug that's never actually been proven to work at all, though millions of people swear by it and it seems "impossible" that it could actually be completely ineffective. At the end of the day, all they can do is endless, futile clinical trials none of which shows any therapeutic benefit at all, but as even doing that doesn't ever disabuse the believers, there's never much point in doing it anyway!

(I recognise that your contention is that they can be be valuable in other ways without necessarily being effective forecasting tools, and am not trying to argue that point!).

 

[nine]

:?: :?: :?:

r4 = r5

and

s4 = s5

Yes. I dont use them so I must have just put dummies in for r5 and s5 when I was coding the pivot dll --- then not noticed that they were never updated. According to linnsoft the last two pivots are:

S4: (HI.1 + LO.1 + CL.1)/3 – (2 * HI.1 – 2 * LO.1)
S5: 2 * (HI.1 + LO.1 + CL.1)/3 – (3 * HI.1 – 2 * LO.1)
R4: (HI.1 + LO.1 + CL.1)/3 + (2 * HI.1 – 2 * LO.1)
R5: 2 * (HI.1 + LO.1 + CL.1)/3 + (2 * HI.1 – 3 * LO.1)

which I am pretty sure gives:
s5 = 2s3-s1
r5 = 2r3-r1