Algo-generated Support/Resistance lines - advanced

New comment
 

I'm a programmer, working on an algorithm based on a structure of SR lines. Would like to discuss it with somebody who analysed this topic, too.


I have a structure of 40 SR lines, generated on multiple timeframes, but it is not very obvious for me when to expire these lines. Some rules which I follow:

1. if SR line is retested from the other side, then it can be removed.

2. If SR line gets touched third time from the same side, it will most probably be broken.

Both rules work as everything - now always :)

(but much more often do than not)

But the most difficulty comes with SR lines which were created but then not touched again. I don't know when to delete them. I tried this combo: CREATED-CROSSED-CROSSED_BACK -> REMOVE. But this is failing too often.


Are there any books or articles answering this question?

Maybe somebody would like to exchange experiences?

 
Trading Connector:

I'm a programmer, working on an algorithm based on a structure of SR lines. Would like to discuss it with somebody who analysed this topic, too.


I have a structure of 40 SR lines, generated on multiple timeframes, but it is not very obvious for me when to expire these lines. Some rules which I follow:

1. if SR line is retested from the other side, then it can be removed.

2. If SR line gets touched third time from the same side, it will most probably be broken.

Both rules work as everything - now always :)

(but much more often do than not)

But the most difficulty comes with SR lines which were created but then not touched again. I don't know when to delete them. I tried this combo: CREATED-CROSSED-CROSSED_BACK -> REMOVE. But this is failing too often.


Are there any books or articles answering this question?

Maybe somebody would like to exchange experiences?

Hi, did you try count candles? for example, the first candle was touched, then set how many candles need to be, to touch again for example 10 if in this time price not touched support line then delete it. Regards Greg

 

been working on that for years

what I've learnt:

SR do NOT expire, price can and will react to SRs from years ago even if already crossed multiple times if there's no more recent price structure in the vicinity

SRs retested from the other side are stronger; i don't remove them

more recent price history is more important

if an SR is created (means at least 3 touches in the past for me) and not touched again it will stay until said touches are out of the past time window I use to determine SRs

I don't have a predefined number of SRs to look for, just a range around current price in which I'll look for SRs

hope this helps

 
luigisimoncini:

been working on that for years

what I've learnt:

SR do NOT expire, price can and will react to SRs from years ago even if already crossed multiple times if there's no more recent price structure in the vicinity

SRs retested from the other side are stronger; i don't remove them

more recent price history is more important

if an SR is created (means at least 3 touches in the past for me) and not touched again it will stay until said touches are out of the past time window I use to determine SRs

I don't have a predefined number of SRs to look for, just a range around current price in which I'll look for SRs

hope this helps

> SR do NOT expire, price can and will react to SRs from years ago even if already crossed multiple times if there's no more recent price structure in the vicinity

While I agree that SRs might not expire, I'd say there is more harm in keeping them around, than removing them. If price action goes through a jungle of earlier-created SRs, and every touch is considered a reaction you'd end-up overtrading for sure.

> SRs retested from the other side are stronger; i don't remove them

That is actually the rule I feel is quite opposite - after testing from the other side it is ok to remove SR. Of course not always, but statistically.

> more recent price history is more important

Of course. But how to translate this to the code?

> if an SR is created (means at least 3 touches in the past for me) and not touched again it will stay until said touches are out of the past time window I use to determine SRs

Hmmm, for me single pivot point is SR line already. And 3 touches from 1 side usually mean this level became a magnet and will be broken very soon. Triple tops/bottoms are very rare, at least on EURUSD, which I analyse mostly.


Thanks for your input Luigi!

 
Greg Pawlak:

Hi, did you try count candles? for example, the first candle was touched, then set how many candles need to be, to touch again for example 10 if in this time price not touched support line then delete it. Regards Greg

Thanks for your reply Greg. I think validity of SRs is not time-constrained.

 

the removal of old SRs is automatic if you recompute SR every so often and only consider recent history 

if a price range has no recent history then you go as far as necessary/possible to establish SRs in that area

also, and maybe the most important thing, forgot to mention last time: SRs are ZONES not exact price levels for me

but you may have a different definition, it's ok. me I see the main pairs bouncing at the same "level" several times from all directions

 
luigisimoncini:

the removal of old SRs is automatic if you recompute SR every so often and only consider recent history 

if a price range has no recent history then you go as far as necessary/possible to establish SRs in that area

also, and maybe the most important thing, forgot to mention last time: SRs are ZONES not exact price levels for me

but you may have a different definition, it's ok. me I see the main pairs bouncing at the same "level" several times from all directions

I've heard this "zones, not exact" many times before, but I tend to have slightly different opinion. If you look at charts, especially 1m timeframe, you'll notice, that price reacts to previous pivot points very often to the tick, with almost surgical precision. I have a concept on coding this. The only issue left for me is about removing those SRs, because there are just too many of them. Purely time-based rule for removal is not enough I think.

 
Trading Connector:

I've heard this "zones, not exact" many times before, but I tend to have slightly different opinion. If you look at charts, especially 1m timeframe, you'll notice, that price reacts to previous pivot points very often to the tick, with almost surgical precision. I have a concept on coding this. The only issue left for me is about removing those SRs, because there are just too many of them. Purely time-based rule for removal is not enough I think.

I see, we're chasing different animals...

I use a different concept elsewhere (for pivots) that may be useful for you: before using them I check if this special levels have "weared out" recently i. e. crossed by a (series of) candles with "significant" H-L range 

hope this helps, best with (y)our endeavour!

New comment