admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
How to survive draw-downs
A did a word document on this in Dec 2020, and much of it applies.
Likely I will re-write the info in this thread.
Here are my thoughts.
Ive been trading now for 23 years and survived. The last year was by far my most profitable in 23 years, was enjoying new highs for many months in a
row, but was sure a decent draw-down will come. Sorry for the bad news but its like taxation and death. You cant escape them, but lot you can do to
minimize the pain & or delay it.
Draw-downs are painful, but much less so if they come out of profits.
If you cant handle the stress, use smaller position size , re-evaluate what you are doing or put your money else where
In my mind system failure is the greatest risk in trading. No matter how good the algorithm, developer, track record. This risk is unseen by portfolio
software and by traders without years of experience.
1) Don't over trade. Ive never seen any aggressive trader survive long time.
If your worried about exchange margin, my guess is your twice as aggressive as you should be.
I allow 4 * max previous draw-down for capital allocation. That applies to each system, not the portfolio. Portfolio dd will be much loser.
2) Don't put all your trading capital in one system. GSBsys1es had 4 years outstanding track record. I have a strong suspicion this became a magnet
for capital without diversification in other systems. I also believe that the version 2 variants of this system are better than the version 1
variants, but I trade all variants in roughly equal weighting of contracts.
I add systems well before I add contracts.
Look at the negative period performance table between systems.
see tables down the page here.
https://trademaid.info/gsbhelp/Systemsforsale.html
Observation is that even the same system on different markets gives significant diversification, though different systems are much better.
3) your future DD will be greater than your last and or in sample DD.
I will post updated graphs as an example on this.
4) When is a system busted?
No one I know claims to have got this nailed. I can say look at what similar systems are doing on the same and different markets. Look also how
similar systems are doing by completely different authors. From 2004 to 2006 you are not going to make money day trading ES regardless of what system
you are trading. There was simply no range or trend.
5) Understand the current market conditions.
This is a bit subjective and only easy with the condition of hind-site.
2 years ago or so, long only systems went ballistic. I was late to scale up my long only systems and late to decrease position size. Hence very low
profits during that period. (much better than losses though)
On the plus, we then had a bear market with superb conditions for ES NQ day trading. I significantly changed weighting of systems to these markets,
with glorious profits. However I was deliberately slow to increase position size with a much bigger account size, as I didn't want the next major
draw-down to be really painful.
Now I have a DD coming out of fat profits.
I still kept trading some of the other markets that were not going great, but just at low leverage / weighting.
Of course this is not helpful to new traders who start trading straight into a DD
5b. Beware of tight stops. Risk averse traders often put really tight stops in times of extreme volatility. This tends to have disastrous results.
Being risk averse is fine, but trade a micro contract if you cant afford getting stopped out multiple times in a how.
6) How often do you re-optimize.
Experts are divided on this topic. I can say GSB systems dont need frequent optimization and normally work better on anchored, rather than rolling
walk forward.
you can read this topic for more detail.
https://trademaid.info/gsbhelp/ArticlesPart1Howoftendoyounee...
I will add to this significantly in time, and when I look at end of month results
Here is a graph of gsbsys1es from a week ago.
my comments were
Look at the hypothetical graph. $20 rt commission and $25 round turn slippage used.
Unusual in that the OOS last few years was better than IS- but i put that down to extremely good market conditions
note also how the OOS curve is choppier than the IS curve - again normal that oos is in some ways worse.
The previous oos draw-down from in sample results was about twice the later draw-down. Again this is normal. (as of a month ago)
Attachment: Login to view the details
As for what to trade.
This is a good guide, but hopefully will be refined a lot with the next build of PA pro showing alternative similar options.
Thanks received (4):
+1 REMO755 at 2023-04-21 18:36:51 +1 getty002 at 2023-03-19 19:57:36 +1 autotrader2023 at 2023-02-25 09:30:13 +1 tecra10 at 2023-02-24 07:38:55
|
|
|
gurmehar9
Newbie
Posts: 1
Registered: 30-12-2022
Member Is Offline
|
|
Well written. Your experience speaks for itself, its not only GSBsys1ES that tripped today but many other systems by other developers realized some
unexpected losses. Keep up the good work, sir.!
Thanks received (10):
+1 cico at 2023-03-01 00:24:21 +1 jeffreed44 at 2023-02-27 02:33:36 +1 engtraderfx at 2023-02-24 19:29:24 +1 BlackBox at 2023-02-24 13:33:42 +1 meldinman at 2023-02-24 08:22:55 +1 tecra10 at 2023-02-24 07:34:26 +1 JozefSusko at 2023-02-24 07:32:06 +1 SwedenTrader at 2023-02-24 02:34:09 +1 Piet at 2023-02-23 23:57:12 +1 timemeantnothing at 2023-02-23 23:48:17
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
more thoughts
Ive claims that market regime and when a system is busted is somewhat subjective. While this is still true, other groups of traders have made some of
this very objective.
Thats beyond my pay grade - but I just wanted to acknowledge these facts.
Others have a system that rotates out so you get the best performing systems from the last few months. I mentioned that to a very seasoned trader and
he said he does the opposite.
What im saying is I know little on these topics, and opinions would vary wildly.
Not a good analogy, but we all know how unreliable political surveys are. There are some parallels that both have a large degree of UN-predictabilty.
Certainty is not always good. A few years ago I spoke to a Canadian couple jailed in another country for 'espionage' . Fascinating story but the
country concerned had a 99% conviction rate for that crime.
Thanks received (1):
+1 SwedenTrader at 2023-02-24 02:38:33
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
One thing to remember, is that sys 1 etc, have a very low max dd, and it does not take much randomness and stops in a row to break this... So I would
not say that even doubling the existing dd, is a metric to define the system as broken.. I agree with Peter, that looking at all systems in a market
combined, gives you a very good view if a market perhaps has changed, and needs a different weight.. knowing if an individual system had failed or
the market in general, makes your decision making much easier.
In regards to system portfolio rotation, picking the worst systems, every n time, actually provides you with best end result most often (we backtested
this ).. but that only is the result of testing on systems that you have, meaning that all those are picked with the wisdom of hindsight.
Picking the best systems, can very often lead you into picking system that have peaked, and are about to head into dd, and the systems you rotate out
of are the ones that are about to head into highs.. meaning that you could risk ending up with a double dd..
However, picking the best ones, enables you to be sure, to never be in broken systems,
You make the choice 
My choice is to rotate into into top n systems from each lookback period, using , 3m 6m and 12m.. so you get a variety of lookback periods,
hopefully reducing the risks above.
Avoiding the bad, is more often more beneficial than picking the best... imho
Thanks received (2):
+1 getty002 at 2023-03-19 20:01:25 +1 admin at 2023-02-24 15:24:18
|
|
|
cheese_man
Junior Member
Posts: 11
Registered: 2-12-2022
Member Is Offline
|
|
I feel like I can just reduce trade size for GSBsys1 at this moment and let the sys2 and the NQ systems do their thing without much more thought. Yes
- the drawdown should be larger than previous ones, yes - we are already closer to the middle of the drawdown than the beginning. But there's no need
to turn it off, just reduce it's weight and ride out any potential dd to it's conclusion. Then re weight it to the original weight near or at new
highs. It's definitely not optimizing profits but it is conservative portfolio weighting.
|
|
|
timemeantnothing
Junior Member
Posts: 22
Registered: 7-7-2022
Member Is Offline
|
|
Hi Peter, I have found a big difference in results trading the micro versions of the systems. The mini version will have positive months but the micro
will be negative making it hard to be in them. I see it with the newer NQ systems. It makes sense to diversify but it's hard when the micro doesn't
match with the mini results. Would it be worth working a micro-only version of the systems?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@timemeantnothing. This is interesting and no suprise
Are you talking about the live fills or the hypothetical. I have strongly stated that you should be using symbol substitution if at all possible
ie chart on @nq and trade ngh23 or @nq.d and trade nqh23 etc
TS can do this with tma
https://trademaid.info/gsbhelp/TrademaidAutomationforTS.html
@Daniel,
I appreciate your comments about system rotation. Its a topic much research could go into. There could also be variations where you change the
position size rather than totally kill systems that are not going well
but dont appear busted
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
SO while we are in a significant draw-down now, i got the results of a few users accounts for last 12 months. Note however other users, often trading
the same (but less) systems PL is really bad. This comes down to system choice, execution errors etc.
This account has $40 rt brokerage, and hence why its only trade emini, not micros
It peaked about 105k profit, down to 95k
This account $254k PL, and note how much was made on mnq vs nq
Sorry the font is so small, but couldn't get it all on the screen otherwise
Profit $230,231
This is another account that started under 100k
$127.85% annual return. The monthly bar graph is good to see the size of draw downs / run ups
|
|
|
cico
Junior Member
Posts: 26
Registered: 4-1-2020
Member Is Offline
|
|
Dears, I'm very glad to see this topic. In the past I was trying to re-evaluate the weights of the systems based on Monte Carlo DD, f.ex. on 15 years
history. But it has made just a small changes and was not "fast" enough. I was afraid to use shorter history because of the phenomena which Daniel is
mentioning above - to give more weights to the systems on their highs, which are just in front of DD, and to give a smaller weights to the systems in
the DD, which are just about to recover..
So actually I do not change the weights too often (almost do not change them at all) and my thinking is that I do allocate into the single system
capital, which is not significant from portfolio point of view, so even in case system would crash, it will not be crucial. So have the weights set
the way allowing the working systems to take care of loosing ones.. But who knows if it is a good attitude or not? Me not.. :-)
What I'm thinking of are two ways. First one is quite simple - cut the size of the system to the half once it goes through the 1/2 of historical max
MC DD and restore the original weight once it recovers. The cons is you loose profits on the way from DD to the recovery. Second one is to have for
example 30 slots for systems for live trading and to have ready 50 systems for trading or more. And every month (or week?) choose the best performing
30 systems out of those 50.. It can be best performing in P3M, P6M or P12M or the combination (100% weight for P3M, 66% weight for P6M and 33% weight
for P12M..). But based on which metric? NP/DD? Net Profit? Some ratio? Should we force some systems to be in the shortlist - f.ex. to avoid all
systems to be in the same ticker etc.?
To wrap it up, I am at the crossroad with this topic at the moment and I would be more than happy to cooperate with the community here to find the
most suitable solution.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Ive updated reports, but no time to digest. Stopping work for today. One factor in your decision is how much of a profit buffer you have.
I like the idea of reducing position size a bit, and perhaps a bit more weighting to longer term systems. (I have low weight on over night / long
term)
Observation is ES systems had much harder time than NQ, but the recent DD is not big relative to past performance in the last year. - But its big
compared to past dd for ES systems
These situations dont have clear best thing to do until you have hind sight
|
|
|
timemeantnothing
Junior Member
Posts: 22
Registered: 7-7-2022
Member Is Offline
|
|
Hi Peter, in regards to the micro results being different than the emini results. If you take a look at Strikers results of GSBsys3MNQ and
GSBsys1MNQV2, you will see a big difference in what the emini did compared to the micro version. Anything Striker can do about it or the code itself?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@timemeantnothing, Striker use symbol substitution (great) so I would expect long term results to be similar
Its when you key of the micro you will get bigger discrepancies.
I run a number of accounts on my server, and fills vary a fair bit even though orders send to my app (IBLINK) at the same time
|
|
|
cico
Junior Member
Posts: 26
Registered: 4-1-2020
Member Is Offline
|
|
Quote: Originally posted by Daniel UK1  | One thing to remember, is that sys 1 etc, have a very low max dd, and it does not take much randomness and stops in a row to break this... So I would
not say that even doubling the existing dd, is a metric to define the system as broken.. I agree with Peter, that looking at all systems in a market
combined, gives you a very good view if a market perhaps has changed, and needs a different weight.. knowing if an individual system had failed or
the market in general, makes your decision making much easier.
In regards to system portfolio rotation, picking the worst systems, every n time, actually provides you with best end result most often (we backtested
this ).. but that only is the result of testing on systems that you have, meaning that all those are picked with the wisdom of hindsight.
Picking the best systems, can very often lead you into picking system that have peaked, and are about to head into dd, and the systems you rotate out
of are the ones that are about to head into highs.. meaning that you could risk ending up with a double dd..
However, picking the best ones, enables you to be sure, to never be in broken systems,
You make the choice
My choice is to rotate into into top n systems from each lookback period, using , 3m 6m and 12m.. so you get a variety of lookback periods,
hopefully reducing the risks above.
Avoiding the bad, is more often more beneficial than picking the best... imho |
Daniel, thank you very much for sharing your point of view and even backtested experience. Have you ever think to look back even on 1m period? For
example mentioned Sys1 from 6m period is not so bad, but from 1m lookback period it is a full loss.. And if I can have a second question, how often do
you rotate? I'm thinking if a week period can bring any advantage or if it would just bring an work overload / overfit.. Thank you.
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
| Quote: Originally posted by cico | | Quote: Originally posted by Daniel UK1 | One thing to remember, is that sys 1 etc, have a very low max dd, and it does not take much randomness and stops in a row to break this... So I would
not say that even doubling the existing dd, is a metric to define the system as broken.. I agree with Peter, that looking at all systems in a market
combined, gives you a very good view if a market perhaps has changed, and needs a different weight.. knowing if an individual system had failed or
the market in general, makes your decision making much easier.
In regards to system portfolio rotation, picking the worst systems, every n time, actually provides you with best end result most often (we backtested
this ).. but that only is the result of testing on systems that you have, meaning that all those are picked with the wisdom of hindsight.
Picking the best systems, can very often lead you into picking system that have peaked, and are about to head into dd, and the systems you rotate out
of are the ones that are about to head into highs.. meaning that you could risk ending up with a double dd..
However, picking the best ones, enables you to be sure, to never be in broken systems,
You make the choice
My choice is to rotate into into top n systems from each lookback period, using , 3m 6m and 12m.. so you get a variety of lookback periods,
hopefully reducing the risks above.
Avoiding the bad, is more often more beneficial than picking the best... imho |
Daniel, thank you very much for sharing your point of view and even backtested experience. Have you ever think to look back even on 1m period? For
example mentioned Sys1 from 6m period is not so bad, but from 1m lookback period it is a full loss.. And if I can have a second question, how often do
you rotate? I'm thinking if a week period can bring any advantage or if it would just bring an work overload / overfit.. Thank you.
|
Hi there, no problem.. I think a good way of thinking here is... what will 1-20 trades tell you ? In a single month you most likely have less
trades... Imho this is just noise.. so in order for it to have any value, you need an aok amount of trades.. if not.. its just random noise..
I am in the opinion that i dont want to rotate at all if i can avoid it .. because risk is high, that I am rotating out when in dd, and rotating in
to something heading towards dd...
Might be something to think about..
Thanks received (3):
+1 admin at 2023-03-03 02:17:20 +1 SwedenTrader at 2023-03-02 15:18:29 +1 cico at 2023-03-02 13:14:30
|
|
|
timemeantnothing
Junior Member
Posts: 22
Registered: 7-7-2022
Member Is Offline
|
|
Current drawdown hurts. Is anyone else feeling the pain or am I the only one? lol
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
DD always hurt in proportion to size of DD. For those with big profit buffer its not at all bad, for those with out profit, those who start at
begining of dd, or those who are too aggressive in money management, then that's really bad.
I have been trading for 22+ years and survived them all, but few people last that long let only come out with good profits.
There is merit to reduce position size in DD according to your account size. If you stop completely you have risk of missing out on run up. But then
its also good to be slow to increase position size as the reduces the next dd from being as harsh.
Trading is much more difficult than people think.
Even for long term stock investors, there is a bias for retail buy and hold investors tend to have poor returns for reasons of psychology. They sell
in market panics and buy market tops.
Whats hard about trading is we do not have the wisdom of hind-site.
|
|
|
cheese_man
Junior Member
Posts: 11
Registered: 2-12-2022
Member Is Offline
|
|
Hey mate! If it hurts that means the size is too big. The only thing to do to stop the pain is to reduce the size and Peter’s recommendation of 4
times the max historical drawdown per system is a fine and necessary rule to go by. That means if your current drawdown is a new max DD then you
adjust accordingly and it also means you should be readjusting as your account balance falls.
This is all about survival rates. The only way to be successful is to stay in the game over the long term (years and years not tomorrow or the next
day).
Thanks received (1):
+1 admin at 2023-03-15 14:42:01
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
Nice to have a good day today.
Someone made this comment to me.
1) On current drawdowns / system performance
yes, widespread with many systems - everyone is feeling this.
Our take is this:
GSBsys1ES has a life return of +$60,410 at Striker since 2018, that beats Gold, Bitcoin, Grains, Stock Market, Bonds - what hasn't it beat?
It's just a lousy drawdown when something goes 5 years you can have an awful month.
So be it.
Once the system turns around (and my opinion is that it will) then quite a few will return we do not see a panic, just a natural course.
So, our suggestion is to just keep on doing what you do - develop systems, monitor, prospect, correspond, etc and so on
|
|
|
timemeantnothing
Junior Member
Posts: 22
Registered: 7-7-2022
Member Is Offline
|
|
What's everyones thoughts on the continued drawdown of GSBsys1ES? Assuming you started the year with the original recommended account size which
around around 26k, you would be close to getting blown out now. And if you started with the new 40k recommended, you would be down 50%. Is this
normal?
|
|
|
cheese_man
Junior Member
Posts: 11
Registered: 2-12-2022
Member Is Offline
|
|
Boy oh boy… is there any system that’s profitable for the last 3 months of this horrible year?
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@cheese_man, i will review them as time permits.
I think ES day trading systems hammered. NQ not so bad.
Likely Vegas, the nq system I have not got around to putting up, and I have 2 GC that are at or close to new highs.
Currently much of my time is still spent on the bar interval project. This took way more time than I could have imagined.
The human brain does not focus well when you give it increasing tasks to do,
so im trying to finish the bar interval project and a swing GC system.
I have a habit of getting distracted, and it means that projects started first do not get finished. Hence I dont want that to happen
So updated results and publishing new systems will have to wait a bit.
@timemeantnothing. This system has had a superb track record, and now is getting hammered. A strong move happens, then reversal.
When looking at is a system busted, I also look at comparable systems - and note many are getting hammered too, but certainly not all systems.
Personally I think that this is more market conditions, but im now trading this 1/10 the leverage of in the past.
The NQ varient of this system years ago has spectacular draw-down, then took of much stronger than in sample.
As for is this normal. The reality is trading systems no matter how good they are, can and do break down. They also sometimes recover.
This must always be kept in mind. The future is always unknown. For traders who have been going for some time, many have really decent profit buffers
and the dd, while its a sting is still ok. For those who started in the beginning of a DD, its ugly.
|
|
|
Daniel UK1
Member
Posts: 470
Registered: 4-6-2019
Member Is Offline
|
|
Tricky times, January +, Feb -, March -, around 7% down i think for the year, trade around 0,2-0,3 risk per system on capital in risk and in total
around 0.8-1% per market.. have reduced risk in March.. I also believe as Peter, that its market conditions, rather than system failure
Thanks received (1):
+1 admin at 2023-04-11 18:45:35
|
|
|
goodoboy
Junior Member
Posts: 38
Registered: 27-7-2021
Member Is Offline
|
|
| Quote: Originally posted by admin | SO while we are in a significant draw-down now, i got the results of a few users accounts for last 12 months. Note however other users, often trading
the same (but less) systems PL is really bad. This comes down to system choice, execution errors etc.
This account has $40 rt brokerage, and hence why its only trade emini, not micros
It peaked about 105k profit, down to 95k
This account $254k PL, and note how much was made on mnq vs nq
Sorry the font is so small, but couldn't get it all on the screen otherwise
Profit $230,231
This is another account that started under 100k
$127.85% annual return. The monthly bar graph is good to see the size of draw downs / run ups
|
Good Morning Peter,
Thank you so much for sharing this knowledge. It helps for confidence.
Questions please:
1. Are these users account running many different GSB trading systems? Or just 1-2 GSB trading systems?
2. How are these users making such great returns when the GSB ES and NQ systems are not doing very well this year so far?
I am confused how these users are profitable this year, when the ES and NQ system I am running is currently drawdown.
Thank you so much for the help.
I am just learning as I go to get better knowledge of trading systems trader.
|
|
|
admin
Super Administrator
Posts: 5060
Registered: 7-4-2017
Member Is Offline
Mood: No Mood
|
|
@goodoboy
you mean the accounts that i publishing showing some very significant returns over a 12 month period, not profit year to-date. I think this is at the
heart of your confusion.
Personally my main concern is PL over the entire tax year, but dates for that depend on what dates your country uses. ie usa au nz all different dates
They are all different size, but running numerous gsb systems and my two accounts running all the non gsb systems I have for sale
Returns were correct at date published (a few months ago) but I would expect be down somewhat since then.
|
|
|
![[*]](images/davis/default_icon.gif){kind=icon}
