How to analyse results – Detailed report – Gain Vs Drawdown?

SpreadBetHub · Wed, 04 Oct 2017 10:25:16 +0000

Dear Coders/Traders I have been testing and coding for a while now and have around 25 systems which I am running live. Looking through the detailed reports I am wondering what to take from the results to determine the likely success of a particular system. So for example there seems little point ‘just’ looking at the gain, if the drawdown is big. I know that you can do a walk forward or a Monte Carlo to really put the strategy through its paces but before considering it for these tests I am trying to decide if it is worth it? So what are your opinions on using these calculations as a guide??? A) Gain Vs the drawdown to give a multiplier (Thinking being that the drawdown is the risk of a strategy which should be weighed against the possible gain) B) No. of units that a system is up Vs the worst loss to date (Thinking being that the Gain is only as good as it’s worst loss multiplied by however many units) C) No. of units that a system is up Vs the average loss (The average loss is more representative and so should be used to calculate how many ‘real’ units the strategy is, in relation to the gain) D) The average loss in units Vs the Drawdown (To establish the makeup of the drawdown – How many ‘real’ units the strategy was down during the drawdown) So for example – In the attached picture here would be my calculations for A,B,C,D A = £1,140 / £107 = 10.6 x multiplier B = £1,140 / £35 = 32.5 units (32.5x that the system can have its worst loss before the expected gain is wiped out) C = £1,140 / £16 = 71 units (71x being the average amount of losses the system is UP in units) D = £107 / £16 = 6.5 units (The most number of units that the system was ever down during its worst drawdown) Before people critisize this post and say that I should just run a Monte Carlo or Walk forward, it all takes time and to be honest if something has a really good GAIN and a really small drawdown then the system will surely always do well on those extra tests – Especially if it is backed up visually by a smooth and steady equity curve. Obviously this question is assuming that the results already meet the following minimum requirements: a) The spread is correct and factored into the backtest b) The Tick by Tick mode is ticked c) There is a good number of results over a reasonable time period ( not 100 trades over 5 years – maybe more like over 100 per year?) Please let me know your thoughts as to which method you think would be the best or if you know of any other? I have heard people talk about ‘Sharpe Ratio’ but I’m not sure how to work that out from these results. Thank you in advance Mark

This topic has 6 replies, 3 voices, and was last updated 7 years ago by Despair.

Viewing 7 posts - 1 through 7 (of 7 total)

10/04/2017 at 11:25 AM #48179

SpreadBetHub

Dear Coders/Traders

I have been testing and coding for a while now and have around 25 systems which I am running live.

Looking through the detailed reports I am wondering what to take from the results to determine the likely success of a particular system.

So for example there seems little point ‘just’ looking at the gain, if the drawdown is big.

I know that you can do a walk forward or a Monte Carlo to really put the strategy through its paces but before considering it for these tests I am trying to decide if it is worth it?

So what are your opinions on using these calculations as a guide???

A) Gain Vs the drawdown to give a multiplier (Thinking being that the drawdown is the risk of a strategy which should be weighed against the possible gain)

B) No. of units that a system is up Vs the worst loss to date (Thinking being that the Gain is only as good as it’s worst loss multiplied by however many units)

C) No. of units that a system is up Vs the average loss (The average loss is more representative and so should be used to calculate how many ‘real’ units the strategy is, in relation to the gain)

D) The average loss in units Vs the Drawdown (To establish the makeup of the drawdown – How many ‘real’ units the strategy was down during the drawdown)

So for example – In the attached picture here would be my calculations for A,B,C,D

A = £1,140 / £107 = 10.6 x multiplier

B = £1,140 / £35 = 32.5 units (32.5x that the system can have its worst loss before the expected gain is wiped out)

C = £1,140 / £16 = 71 units (71x being the average amount of losses the system is UP in units)

D = £107 / £16 = 6.5 units (The most number of units that the system was ever down during its worst drawdown)

Before people critisize this post and say that I should just run a Monte Carlo or Walk forward, it all takes time and to be honest if something has a really good GAIN and a really small drawdown then the system will surely always do well on those extra tests – Especially if it is backed up visually by a smooth and steady equity curve.

Obviously this question is assuming that the results already meet the following minimum requirements:

a) The spread is correct and factored into the backtest

b) The Tick by Tick mode is ticked

c) There is a good number of results over a reasonable time period ( not 100 trades over 5 years – maybe more like over 100 per year?)

Please let me know your thoughts as to which method you think would be the best or if you know of any other? I have heard people talk about ‘Sharpe Ratio’ but I’m not sure how to work that out from these results.

Thank you in advance

Mark

SpreadBetHub
Participant

Topics: 3
Replies: 5
Been thanked: 0 times

10/04/2017 at 11:56 AM #48183

victormork

Hi! I think you are making some good points here. I’ve been spending a lot of time this road and tried most of the methods that I’ve come across. Using the drawdown compared to gain is not a bad idea. It’s good to compare how many times your drawdown gets reproduced on the upside. I think you could also try to include the length of your drawdown to compare with the length of time above the water.

I calculate sharpe ratio but it’s not my favourite tool. I rather use average excess return/standard deviation > 0,2.

victormork
Participant

Topics: 19
Replies: 165
Been thanked: 57 times

10/04/2017 at 12:19 PM #48184

SpreadBetHub

Hi

Thank you for your response and for your idea to use that method in conjunction with time.

Please can you explain exactly how to work out the sharpe ratio using the detailed report I attached? And I am also interested in your preferred method you use “average excess return/standard deviation >0.2”

If you show me the full workings out of both I would be very grateful

Mark

SpreadBetHub
Participant

Topics: 3
Replies: 5
Been thanked: 0 times

10/04/2017 at 12:39 PM #48186
victormork
First you pull up the detail report and click on the tab “closed positions list”.

Click and hold with your mouse over the list and drag it on to an open excel document at release.

Then you calculate the excess return of your system compared to the risk free rate (T-bill) or if you want to set the bar high you could use the average return of the US500. The easiest way to learn this is by search on youtube. There are loads of clips where people show how to do it. (this is a part of how to calculate sharpe ratio but that’s a different story)

When you have the average excess return and the standard deviation of your closed positions you simply divide them with each other. If the result is above 0,2 you have a good system.

Be aware of that a lot of these calculations based on square deviation can (in my opinion) favour mean reversion. So if you have a breakout system that get acceptable result that’s great!

You can also check out system quality number which is also based expectancy and standard deviation. (Just google and you will see).

1 user thanked author for this post.

Derek
victormork
Participant

Topics: 19
Replies: 165
Been thanked: 57 times
10/05/2017 at 12:17 PM #48261

SpreadBetHub

Thank you for your reply

Can I ask you if you use the Absolute performance or the relative performance to calculate the average excess return and the standard deviation?

Thanks

Mark

SpreadBetHub
Participant

Topics: 3
Replies: 5
Been thanked: 0 times

10/05/2017 at 2:41 PM #48297

victormork

I use the absolute from each trade in a one or two year look back period. Depending on lookback period you need to adapt the risk free rate of return. To be frank I find it hard to explain completely in writing, that’s why I recommend that you search on YouTube where you can get step by step example in excel. The only thing you need from prt is the outcome of each trade in your lookback period. Try to calculate on one year first, because that’s the easiest task.

victormork
Participant

Topics: 19
Replies: 165
Been thanked: 57 times

10/06/2017 at 11:34 PM #48490

Despair

Another good measure is the ratio return/DD. If this is 2 or better you usually have found something.

Despair
Blocked

Topics: 97
Replies: 1034
Been thanked: 218 times
Author

Posts

Viewing 7 posts - 1 through 7 (of 7 total)

Find exclusive trading pro-tools on

How to analyse results – Detailed report – Gain Vs Drawdown?

1 user thanked author for this post.