Using ATR incorrectly during back-testing :( | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

trailing stop ea using ATR 29 replies

Correct number of weeks for ATR and hour ATR 1 reply

Please help - my EA incorrectly opened positions at market opening after weekend 4 replies

ATR using O/Close 4 replies

Creating bands using ATR 11 replies

Using ATR incorrectly during back-testing :(

  • #1
  • First Post: Aug 9, 2018 9:38pm Aug 9, 2018 9:38pm
I think I just made a rookie mistake so hopefully this thread is in the appropriate place. *sigh*

So I just finished a very comprehensive manual backtest and recently started trading it live. To my horror when comparing my notes, I noticed that the ATR figure I was using on a daily (live) basis was different than the one I used in my backtesting. It appears the ATR indicator included in everybody's basic copy of MT4, continually updates its average in real time - even on a daily chart. I assumed I was getting numbers on the previous X amount of days, not the previous X+today. So now my meticulous, beautiful back-tested strategy is suspect. Essentially I was using a volatility crystal ball on the last day of every averaged period. This would be no big deal if I was using 50 periods, but my sample size is much smaller.

What should I do? (besides kill myself)

On a shorter ATR daily time frame, is it standard practice to not include the present day in your system or do people just include it and pay no attention to the fact that at the start of the U.S. trading period, only 1/3 of the current day's volatility is being averaged into your ATR sample.

Thanks for listening.

I need a drink.......

SD
  • #2
  • Aug 10, 2018 7:42pm Aug 10, 2018 7:42pm
Yes ATR includes the current period. If I want to use ATR in a strategy, I use the previous day’s.

Also, there are several different ways to calculate ATR, some use the simple moving average of the true range (which is the simplest), others use an exponential moving average of the true range, which gives greater weight to more recent periods.

Annoyingly, platforms don’t usually show which method they use to calculate ATR, you can only tell by working backward and calculating it for yourself. (Although I can tell you that the CTRADER platform uses a simple moving average calculation.

More about the three accepted methods here:
http://www.macroption.com/atr-excel/
  • #3
  • Aug 13, 2018 4:05pm Aug 13, 2018 4:05pm
Using Sunday's 3 hour minuscule volatility as the end point for Monday's numbers is madness. On a 7 period ATR you can see the average dip on every Sunday and then "correct" on Monday. Are people really using the artificially low Sunday number for Monday when testing? It seriously underestimates Monday's range almost every time.
  • #4
  • Aug 13, 2018 4:29pm Aug 13, 2018 4:29pm
Maybe just go off Friday's bar?
  • #5
  • Aug 14, 2018 11:55am Aug 14, 2018 11:55am
Yep bonkers I know. if you want to program it, yes go off Friday but I'd focus on improving ATR as it's very very blunt.

Some things you might want to test - weighting day of the week (Mon/Fri always break more ranges than tue/wed/thu), weighting sessions (smaller asian sessions), and weighting recency (yesterday is more important than 2 days, 3 days etc), weighting big news (NFP's for eg) and finally ignoring sunday, unless it gaps.
On an intraday basis, there's some fairly stable volatility calculations hour by hour you could use for weighting.

Finally, if you're getting very fancy, incorporate how volatility clusters and how corresponding breakouts or breakdowns flip volatility to it's opposite (crazy big range or virtual standstill).
  • #6
  • Aug 14, 2018 2:41pm Aug 14, 2018 2:41pm
Quoting triphop
Yep bonkers I know. if you want to program it, yes go off Friday but I'd focus on improving ATR as it's very very blunt. Some things you might want to test - weighting day of the week (Mon/Fri always break more ranges than tue/wed/thu), weighting sessions (smaller asian sessions), and weighting recency (yesterday is more important than 2 days, 3 days etc), weighting big news (NFP's for eg) and finally ignoring sunday, unless it gaps. On an intraday basis, there's some fairly stable volatility calculations hour by hour you could use for weighting....
I think I'm going to maintain an ATR calculation via spreadsheet and just stop using the MT4 indicators for now. Combining the Sunday bars with Monday seems natural as "Sunday's range" is normally inside of Monday and/or usually jibes with Monday's flow. What do ya think?
  • #7
  • Aug 14, 2018 8:02pm Aug 14, 2018 8:02pm
Quoting VEEFX
{quote} Glad you are finding out anomalies in MT4. I am a non-believer in ATR, ADR or anything that uses Highs and the Lows including trend definition. The highs and the Lows are the areas on charts that is the most dependent on liquidity and varying market conditions that CANNOT be predicted or understood or reflected accurately in any history of time series. It will ALWAYS result in in-consistent results either in backtesting or in live testing. The ONLY exception to this is the HL of WEEKLY candle because this is the only duration when the entire...
You sound like a conspiracy theorist.

ATR works great for me and is a fairly accepted technical discipline amongst successful retail traders. While fake breakouts are the norm in Forex, faking daily volatility ranges would be a much more difficult trick.
  • #8
  • Last Post: Aug 15, 2018 9:48am Aug 15, 2018 9:48am
Quoting STD-Deviant
{quote} I think I'm going to maintain an ATR calculation via spreadsheet and just stop using the MT4 indicators for now. Combining the Sunday bars with Monday seems natural as "Sunday's range" is normally inside of Monday and/or usually jibes with Monday's flow. What do ya think?
I think that'll do fine -or alternatively with Friday's flow. Think VEE has got mixed up into pin-point accuracy rather than probability but either way, facts don't lie. Good luck
0 traders viewing now