Home › Forums › Trading System Mentor Course Community › Trading System Brainstorming › All Weather Discussion
- This topic is empty.
-
AuthorPosts
-
October 8, 2023 at 7:08 am #115725ScottMcNabParticipantJulian Cohen post=14186 userid=5314 wrote:I think
Volatility: stddev(S.NetPct,20) * sqr (252)
I think that gives you 20 day annualised std dev
Hoping someone with better maths than me can help with my limited understanding of this metric please. I suspect I am misinterpreting it. The only statistics course I did was almost 30 years ago which puts me firmly into the “a little knowledge is dangerous” camp. If the distribution of returns was normal, then a 2 SD move would contain approx 95% of observations….but I believe that distributions of returns are not normally distributed and have fat tails such that a 3 (or even 4) SD move would not be unexpected. I am trying to use this metric 2 ways.
1. in backtesting of systems…eg… i have a long term trend following system with 15 cagr and -30% DD with 15% vola….i then change or optimise a metric and the DD drops to -24%….is this a real improvement of have I just curve fitted ? I find it so difficult to get a sufficient number of trades in one specific market that I don’t tend to reserve a small 5-10 year time frame for OOS as it doesn’t really provide a sufficient sample size. In the past I tried to answer this by taking it over to the AUS and TSX data sets and seeing if the change in the metric resulted in the same overall trend to backtest results. I have now also tried looking at the change in vola. If the improvement of DD from -30 to -24 resulted in no change to the vola (ie remained at 15% in this example) then I am less likely to think I have made a meaningful change than in the case where the lower DD was also accompanied by a drop in vola (eg to 12 or 13%)
2. deciding if a system is broken or needs to be shut down….along the same lines, using vola may be helpful in predicting future DD…rather than using the simple assumption that the system is broken if it reaches 2x the drawdown of the backtest (which is obviously only 1 or many possible outcomes), is it possible to extrapolate vola using a 3 or 4 SD move for a more meaningful estimate ?…so for a 3SD move it would be a case of doubling the vola and making it negative to get a likely future DD expectation….my concern is obviously that my lack of understanding regarding the distribution of the returns is leading to incorrect conclusions
Thanks
October 8, 2023 at 9:40 pm #115844Nick RadgeKeymasterScott,
1. My thought process with this is that markets evolve and parameter settings should not be static. They too need to evolve as well. I will test on a full data set to ensure long-term objectivity. I will then test over the last 5-years to see if ‘recent’ conditions warrant a slightly different parameter setting. That said, any adjustment to parameters needs to be limited, ideally to a singular one, and within that it too must also show robustness.2. I’m not sure I’m the guy. I would suggest using 3 – 4 std dev to calculate a possible future drawdown would be useful.
October 9, 2023 at 8:13 am #115845ScottMcNabParticipantThanks Nick…as much as it hurts I think 3-4 SD move is what I will go with at this stage….which directs me more towards the “ETF index camp” for a long term system than individual stocks
October 14, 2023 at 7:17 am #115846Nick RadgeKeymasterJust a heads-up for anyone using the DBC ETF for commodity exposure. I have just found out its a PTP which means that they can hold 10% of profit as withholding tax for non-US residents.
I have switched to PDBC without much difference to performance metrics
November 1, 2023 at 9:27 pm #115847Nick RadgeKeymasterAnother heads up…(learning new things!)
If you use the IBKR Rebalance Tool to rebalance the portfolio each month, it places LMT orders. You may need to transmit those then amend to MKT. Due to a gap up I missed a buy adjustment in QQQ last night, which of course rallied +1.7%.
November 12, 2023 at 1:13 am #115869PaulBaileyMemberI have a question for those using Self Wealth not worthy of a new topic:
Where do you park your money when not invested in the ETF strategy?
When I’m running my Australian strategy I put it in HBRD.
Is there anything equivalent in the US market? I’ve looked and cannot find it.My US ETF strategy generates daily signals and the lag between selling HBRD and transferring the $ to my US account makes my previous work practice inappropriate.
Thanks for any ideas.
November 12, 2023 at 5:44 am #115877Nick RadgeKeymasterI’d be inclined to use AAA.ax for the ASX cash element and SHV or BIL for the US side of things.
December 2, 2023 at 5:17 am #115879PaulBaileyMemberI have a question about the mechanics of the ETF market that is going to reflect some naivety on my part.
When we buy or sell an individual share, our ability to do so relates to the market depth – and large orders on thinly traded stocks either wouldn’t get filled, or you would move the market substantially should you really want to do so.
My understanding is that the price of an individual ETF is based on the price of the underlying basket of shares (or whatever they are invested in).
Would it be true to say, therefore, that a large order eg $200k, would be met by the market maker and not move the price of the underlying ETF – for example something like QQQ where daily activity is huge.
Thanks
PB
December 2, 2023 at 5:17 am #115880PaulBaileyMemberI have a question about the mechanics of the ETF market that is going to reflect some naivety on my part.
When we buy or sell an individual share, our ability to do so relates to the market depth – and large orders on thinly traded stocks either wouldn’t get filled, or you would move the market substantially should you really want to do so.
My understanding is that the price of an individual ETF is based on the price of the underlying basket of shares (or whatever they are invested in).
Would it be true to say, therefore, that a large order eg $200k, would be met by the market maker and not move the price of the underlying ETF – for example something like QQQ where daily activity is huge.
Thanks
PB
December 2, 2023 at 5:17 am #115881PaulBaileyMemberI have a question about the mechanics of the ETF market that is going to reflect some naivety on my part.
When we buy or sell an individual share, our ability to do so relates to the market depth – and large orders on thinly traded stocks either wouldn’t get filled, or you would move the market substantially should you really want to do so.
My understanding is that the price of an individual ETF is based on the price of the underlying basket of shares (or whatever they are invested in).
Would it be true to say, therefore, that a large order eg $200k, would be met by the market maker and not move the price of the underlying ETF – for example something like QQQ where daily activity is huge.
Thanks
PB
December 2, 2023 at 9:25 am #115902JulianCohenParticipantin something as liquid as QQQ I doubt that even a $10,000,000 order would affect the price much. But in a smaller volume ETF, then maybe, but even $200K shouldn’t make too much difference.
December 2, 2023 at 10:13 am #115904PaulBaileyMemberThanks a lot Julian, that was my assumption – just making sure my thinking wasn’t deranged!
Thanks for taking the time to answer I really appreciate it.December 4, 2023 at 12:12 am #115906Nick RadgeKeymasterJust a heads up on the meaning of All Weather.
All Weather is NOT a strategy that trades ETFs. It’s a strategy that uses ETFs to trade a mix of uncorrelated asset classes to ‘weather’ economic storms.
Shouldn’t be confused with an ETF strategy that is used for purely growth.
December 4, 2023 at 3:12 am #115909GlenPeakeParticipantThe latest Chat With Traders with Tom Basso… Episode #271
He speaks a little bit about the All Weather strategy and hedging.
271 · The Sweet Balance Between Risk and Reward w/ Tom Basso
December 4, 2023 at 5:21 am #115910Nick RadgeKeymasterThanks. I tried reading his book, The All Weather Trader. Got halfway through and had to stop. Incredibly convoluted and overly complex solution.
-
AuthorPosts
- You must be logged in to reply to this topic.