Return Over Maximum Draw Down – Code Snippet

Vonasi · Tue, 16 Apr 2019 16:16:03 +0000

I coded a little snippet that can be added to the bottom of any strategy that you are back testing that returns a score for ‘Returns Over Maximum Draw Down’ and thought that I would share it here in case it is of any use to anyone. Return Over Maximum Draw Down (RoMaD) is simply a ratio calculated by dividing the current profit by the biggest draw down that a strategy has ever seen (so the biggest all time fall in equity). This simple ratio is a good indicator of strategy quality. The higher the number the better the strategy. In the first attached image it can be seen how RoMaD has changed over the history of the strategy and how the big drawdown in 2001 has resulted in a pretty poor score of just 1.8106. (Return is only 1.8106 times the size of the biggest draw down). In the second attached image it can be seen that with the addition of an extra condition to close positions the RoMaD score has increased to 5.4483. Here is the code to add to your strategies: maxstratprofit = max(maxstratprofit, strategyprofit) if longonmarket then runningequity = strategyprofit - ((positionprice - low) * countofposition) MaD = max(MaD, maxstratprofit - runningequity) endif if shortonmarket then runningequity = strategyprofit - ((high - positionprice) * abs(countofposition)) MaD = max(MaD, maxstratprofit - runningequity) endif RoMaD = strategyprofit / MaD graph RoMaD Note: I’ve just coded this and not road tested it at all so if you find any issues please let me know!

This topic has 21 replies, 3 voices, and was last updated 5 years ago by GraHal.

Viewing 7 posts - 16 through 22 (of 22 total)

← 1 2

04/17/2019 at 10:56 AM #96678

Vonasi

STOP PRESS!!! I just spotted an error in my RoMaD codes in the above posts. When short I needed to use ABS(COUNTOFPOSITION) instead of COUNTOFPOSITION to make it a positive value. I was testing on a long only strategy so missed this important point! I have edited the code in the previous posts so it is now correct.

1 user thanked author for this post.

GraHal

Vonasi

Moderator

Topics: 192

Replies: 5088

Been thanked: 1810 times

04/17/2019 at 11:19 AM #96682

Vonasi

Here is another way of scoring a strategy. This is Return over Average Draw Down – RoAvD. This calculates the average of all draw downs and then divides the strategy profit by that average.

This code only really works accurately on strategies that only open and close trades at the open of a new bar as without using tick by tick it is not possible to know whether a low or high happened before or after a trade that is opened mid bar.

Once again it works on strategies that trade both long and short or both. A high score is a good score.

All these strategy rating codes have just been coded and posted and not thoroughly road tested at all so please bear this in mind when using them. If you spot any errors or possible improvements then please do let me know.

if longonmarket then
maxdd = max(maxdd, (positionprice - low) * countofposition)
endif

if shortonmarket then
maxdd = max(maxdd, (high - positionprice) * abs(countofposition))
endif

if strategyprofit <> strategyprofit[1] then
ddtotal = ddtotal + maxdd[1]
ddcount = ddcount + 1
maxdd = 0
endif

averagedd = ddtotal / ddcount
RoAvD = strategyprofit / averagedd

graph RoAvD

if longonmarket then

maxdd = max(maxdd, (positionprice - low) * countofposition)

endif

if shortonmarket then

maxdd = max(maxdd, (high - positionprice) * abs(countofposition))

endif

if strategyprofit <> strategyprofit[1] then

ddtotal = ddtotal + maxdd[1]

ddcount = ddcount + 1

maxdd = 0

endif

averagedd = ddtotal / ddcount

RoAvD = strategyprofit / averagedd

graph RoAvD

Vonasi

Moderator

Topics: 192

Replies: 5088

Been thanked: 1810 times

04/17/2019 at 12:03 PM #96687

Vonasi

I was just in the middle of coding something that calculated return on average draw down for buy and hold but my PRT platform just crash closed and I’m guessing I have lost everything just as I had finished it. There may be a short delay while I thump the living day lights out of something!

Vonasi

Moderator

Topics: 192

Replies: 5088

Been thanked: 1810 times

04/17/2019 at 12:51 PM #96696

GraHal

Vonasi wrote:

There may be a short delay while I thump the living day lights out of something!

Damn that’s so flippin annoying!

Good job her indoors is out of doors > 1000 miles away! 🙂

GraHal

Participant

Topics: 203

Replies: 8374

Been thanked: 2024 times

04/17/2019 at 1:22 PM #96701

Vonasi

Right – second time lucky – here is RoAvDoBaHRoAvD! A catchy little title I think you will all agree. RoAvDoBaHRoAvD = Return over Average Draw Down OVER Buy and Hold Return over Average Draw Down. (RoAvD OVER BaHRoAvD)

Basically we calculate our strategy profit and divide this by the all time average draw down for the strategy and then we do the same calculation for buy and hold. Our ratio is achieved by then dividing the strategy result by the buy and hold result.

Once again high results are good and over 1 is better than buy and hold.

Once again – not fully road tested!

//RoAvD
if longonmarket then
maxdd = max(maxdd, (positionprice - low) * countofposition)
endif
 
if shortonmarket then
maxdd = max(maxdd, (high - positionprice) * abs(countofposition))
endif
 
if strategyprofit <> strategyprofit[1] then
ddtotal = ddtotal + maxdd[1]
ddcount = ddcount + 1
maxdd = 0
endif
 
averagedd = ddtotal / ddcount

RoAvD = strategyprofit / averagedd



//RoBaHAvD
buyholdpositionsize = 1
once firstprice = low
buyandholdequity = ((high - firstprice) * buyholdpositionsize)

if low < high[1] and not bhflag then
bhindex = barindex - 1
bhflag = 1
endif

if bhflag then
bhmdd = max(bhmdd, (high[barindex - bhindex] - low) * buyholdpositionsize)
endif

if buyandholdequity > buyandholdequity[barindex - bhindex] then
bhflag = 0
bhcount = bhcount + 1
bhtotal = bhtotal + bhmdd
bhmdd = 0
endif

bhaveragedd = bhtotal / bhcount

BuyHoldRoAvD = buyandholdequity / bhaveragedd


RoAvDoBaHRoAvD = RoAvD / BuyHoldRoAvD


graph RoAvD
graph BuyHoldRoAvD
graph RoAvDoBaHRoAvD

//RoAvD

if longonmarket then

maxdd = max(maxdd, (positionprice - low) * countofposition)

endif

if shortonmarket then

maxdd = max(maxdd, (high - positionprice) * abs(countofposition))

endif

if strategyprofit <> strategyprofit[1] then

ddtotal = ddtotal + maxdd[1]

ddcount = ddcount + 1

maxdd = 0

endif

averagedd = ddtotal / ddcount

RoAvD = strategyprofit / averagedd

//RoBaHAvD

buyholdpositionsize = 1

once firstprice = low

buyandholdequity = ((high - firstprice) * buyholdpositionsize)

if low < high[1] and not bhflag then

bhindex = barindex - 1

bhflag = 1

endif

if bhflag then

bhmdd = max(bhmdd, (high[barindex - bhindex] - low) * buyholdpositionsize)

endif

if buyandholdequity > buyandholdequity[barindex - bhindex] then

bhflag = 0

bhcount = bhcount + 1

bhtotal = bhtotal + bhmdd

bhmdd = 0

endif

bhaveragedd = bhtotal / bhcount

BuyHoldRoAvD = buyandholdequity / bhaveragedd

RoAvDoBaHRoAvD = RoAvD / BuyHoldRoAvD

graph RoAvD

graph BuyHoldRoAvD

graph RoAvDoBaHRoAvD

1 user thanked author for this post.

GraHal

Vonasi

Moderator

Topics: 192

Replies: 5088

Been thanked: 1810 times

04/17/2019 at 3:02 PM #96712

Vonasi

I’ve had a further idea for a strategy quality score. I’d like to multiply the total (or maybe average) draw down during the life of a strategy by the number of bars it is in draw down and then divide this by the return multiplied by the total number of bars. A sort of ‘Ulcer’ index or ‘Don’t panic’ index. Perhaps something for tomorrow.

Vonasi

Moderator

Topics: 192

Replies: 5088

Been thanked: 1810 times

04/17/2019 at 4:25 PM #96724

GraHal

Vonasi wrote:

A sort of ‘Ulcer’ index or ‘Don’t panic’ index. Perhaps something for tomorrow.

Yeah another great idea Vonasi!

GraHal

Participant

Topics: 203

Replies: 8374

Been thanked: 2024 times

Author

Posts

Viewing 7 posts - 16 through 22 (of 22 total)

← 1 2

Find exclusive trading pro-tools on