OmniTrader Professional Forum - OmniLanguage Discussion
Hull MA calcs vs Classic for Wma & Ema

^ Top
Jim Dean

Posts: 3433

Joined: 3/13/2006
Location: L'ville, GA

User Profile
 
Subject : Hull MA calcs vs Classic for Wma & Ema
Posted : 2/6/2019 11:30 AM
Post #30287

Salim emailed me an interesting article that does a "good" study of Lag & Smoothness for a wide variety of Moving Average calc-types. Here's the link for it (pdf is attached) ...
https://pdfs.semanticscholar.org/257b/837649d8b50662b3fe2c21fce825a1...

His purpose for bringing it to my attention was in relation to some TradeTight routines (MTV, CVW, etc) whose algorithms are closely tied to slope and stack "voting", using a family of moving averages. I've typically used Wma, occasionally Sma, and fairly often a "bounded" Ema (see info in link below re bounding) ... presumably this was a suggestion that I consider using alternative MA calc-types to improve performance.

Click Here for my evaluation of how this can be implemented in OLang, and a bunch of snapshots of comparisons on OT's charts, and OLang code you can use to duplicate it.
Attached file : Moving Average Analysis.pdf (272KB - 362 downloads)

^ Top
Mel

Posts: 235

Joined: 3/18/2006

User Profile
 
Subject : RE: Hull MA calcs vs Classic for Wma & Ema
Posted : 2/6/2019 2:58 PM
Post #30288 - In reply to #30287

He does not evaluate overshoot or lag for price impulses. Depending on what you are doing, the biggest problem comes from how MAs respond weirdly to price impulses. On relatively smooth charts, they are all the same, on wavey ones like your QCOM, the analysis is good.

^ Top
Jim Dean

Posts: 3433

Joined: 3/13/2006
Location: L'ville, GA

User Profile
 
Subject : RE: Hull MA calcs vs Classic for Wma & Ema
Posted : 2/6/2019 3:10 PM
Post #30289 - In reply to #30288

Yup, good point. It's definitely not as fancy as some of the MESA stuff. But it's fairly easy to understand. And, for many purposes, overshoot doesn't "logically" matter (if you're comparing MA to MA, not to underlying). Also, the solutions I arrived at are eminently OLang-able, and bound-able.


https://www.omnitrader.com/currentclients/proforum/thread-view.asp?threadid=7665&posts=3