Reducing Entry Signals

[LeeDanello]

Julian Cohen wrote:

What I heard, and I’m happy to be corrected, is that he trawls through the backtest and manually checks each trade to see if it was hit in the order he expected. If that’s tru then he is either not doing many trades (which is the case if his RSI is set to under 5), or he’s testing over a shrt period such as a year. Either way it is a truck load of work. So much easier to check to selection bias the way we do it.

I’d rather do a truckload of work and be sure about it rather than relying on some dodgy code and hope that works. If you have any doubt about your system then you probably shouldn’t trade it.

[TrentRothall]

That’s how i heard it too guys, i think he mentioned he has systems on sp100 and some of the universes with a small number of stocks. Couple that with tight parameters, there might not be many trades.

[JulianCohen]

Maurice Petterlin wrote:

Julian Cohen wrote:

What I heard, and I’m happy to be corrected, is that he trawls through the backtest and manually checks each trade to see if it was hit in the order he expected. If that’s tru then he is either not doing many trades (which is the case if his RSI is set to under 5), or he’s testing over a shrt period such as a year. Either way it is a truck load of work. So much easier to check to selection bias the way we do it.

I’d rather do a truckload of work and be sure about it rather than relying on some dodgy code and hope that works. If you have any doubt about your system then you probably shouldn’t trade it.

Thankfully the dodgy code I’m relying on was written by me!

[RobGiles]

Scott,

I look at my MOC system from a diversification of systems perspective. The CAGR is important to me, but I ultimately want a portfolio of about 5 – 6 systems that work in varying market types and are not all closely correlated. I trade mine in a company owned by a trust, not in my SMSF.

[RobGiles]

Len Zir wrote:

Just finished listening to part 3 of Caesar Alvarez’s presentation on mean reversion trading on Better System Trader where he answered listener questions. Somebody by the name of “Rob” (?this forum’s Rob?) asked what he thought of trading an MOC system with 4:1 leverage. His answer was that it was “suicidal” and guaranteed to wipe out your account. He cited what would have happened to an MOC system during the flash crash and the 1987 crash.
Of some interest my MOC system actually made money the day of the flash crash. Nevetheless interesting perspective on leverage. It doesn’t sound like he uses leverage on his mean reversion systems based on the CAGR’s quoted on his site.

Guilty as charged…that was me!
I’m still trading it at 4:1 but only have small % of my total investment capital allocated to it. I intent to build another MR strategy next year that carries positions overnight, but uses max 50% leverage, maybe none. If I can get > 15 CAGR at zero leverage in a volatile market regime, then I’d like to add that system to my portfolio.

[ScottMcNab]

Just spent time over the weekend doing that too Rob…now have a MRV that holds for a max of 7 days and does not use margin….CAGR in the low 20’s for period (2000-present) but drops high teens for (2010-current)…tried on RUA as suggested in BST podcast but didn’t work for me… I also plan to keep the 4:1 MOC but will allocate a lower amount …maybe 10% for it and perhaps 20 % for MRV.

[SaidBitar]

The only problem with moc 4:1 if the stocks dropped 25% intraday then we are zero i think even if they recovered at the close i am not sure it will help due to automatic liquidation i know it is a risk and it may happen but honestly i dont think about it much
But good excersise will be checking stoploss at 10% from the entry i know this will kill the system but if it will make you sleep better then be it knowing in the worst case you will lose 40% of the account

[SaidBitar]

Another idea that i think is better than the stop loss is testing if you have a certain percent loss on a day exit all trades instantly so assume your account is 100k and you decided that if the account drops by more than 30k to exit all positions with market orders
It can be tested with amibroker but i think it is easier manually just take the backtest result calculate the daily profit and then calculate the daily loss and from there you can test

I wanted to test the strategies for the 1987 crash i am sure i have data back to 1985 but somehow they are showing me from 90s i will check during the week where are the data

[LEONARDZIR]

The problem with 1987 is at one point there were no bids under the market. So in that situation stop loss doesn’t really work..

[ScottMcNab]

Think it comes back to the point in Nick’s post….if we have limited (eg 10% of) funds in leveraged MOC… even if goes to zero following 25% intra-day drop still have rest of funds (which have probably also been savaged of course !)

other option is to just ditch the leveraged MOC systems

[DanielBaeumler]

I’m also trading my US MOC with 4:1. In order to get a better feel for the risk, I verifyied worst case scenarios on all days selection bias would have kicked in (77 from 2000 to-date) and also ran a Forward analysis in AB on daily basis (took about 18 hrs). Quite a bit of work but it really helped me to get more comfortable with deciding how much money I want to allocate to this strategy.

Result: Worst days assuming worst case scenarios (all days were subjected to selection bias)

4/8/11: -12.4%

8/8/11: -11.2%

28/10/09: -9.6%

Worst months assuming worst case scenarios:

Aug 11: -11.1%

Jun 10: -4.2% (no selection bias cases in this month)

Jun 12: -4.2% (no selection bias cases in this month)

Unfortunately, I don’t get any data for 1987.

Conclusion:

Even though, trading 4:1 appears very risky, the system would have easily survived even the month of August 2011 with 2 pretty tough days. Nevertheless, we all know that future will be different to the past and the hardest days may be ahead of us. I will also reduce the capital allocated to this strategy but still stick to 4:1 for the time being.

[ScottMcNab]

Said Bitar wrote:

The only problem with moc 4:1 if the stocks dropped 25% intraday then we are zero i think even if they recovered at the close i am not sure it will help due to automatic liquidation i know it is a risk and it may happen but honestly i dont think about it much
But good excersise will be checking stoploss at 10% from the entry i know this will kill the system but if it will make you sleep better then be it knowing in the worst case you will lose 40% of the account

(Notwithstanding Len’s point and whether we could actually get a fill at SL 10% below entry ) did you find it made a big difference to your MOC systems Said (or anyone else that has tested this) ?…I tried it just then after reading your post and CAGR dropped 1-2% but still seems viable…guess not many days in backtest when individual stocks drops 10%….

I could imagine myself doing this and getting out of the bottom of the flash crash before it rebounded to normal..so perhaps sell half of position at 10% SL ?

out of theoretical interest,next issue would be if was possible in IB without changes to API ? …

[RobGiles]

Scott McNab wrote:

Just spent time over the weekend doing that too Rob…now have a MRV that holds for a max of 7 days and does not use margin….CAGR in the low 20’s for period (2000-present) but drops high teens for (2010-current)…tried on RUA as suggested in BST podcast but didn’t work for me… I also plan to keep the 4:1 MOC but will allocate a lower amount …maybe 10% for it and perhaps 20 % for MRV.

Thanks for sharing Scott. That’s exactly the type of system I’m looking for…I’d be totally happy with that (assuming the Max DD < high teens). Not surprising that performance drops when we're in a low vol bullish trend, but given it doesn't use leverage, that's system I'd be totally comfortable to use to compound away in market environments that might not suit trend following / rotational systems. I'd be really interested to know more about it if you're willing to share.

[ScottMcNab]

All my moc/mrv are pretty similar…based off price patterns similar to Nick’s second income…then added vola filter (trades in higher vola) …and other filter to keep selection bias down …uses adx and atr for this…doesn’t have index filter..close>ref(c,-1)….exits on 7th day if hasn’t closed higher…..interestingly in the BST podcast on mrv many were based on indicators (rsi etc) but not yet achieved success with those yet but will keep exploring

[SaidBitar]

Scott McNab wrote:

Said Bitar wrote:

The only problem with moc 4:1 if the stocks dropped 25% intraday then we are zero i think even if they recovered at the close i am not sure it will help due to automatic liquidation i know it is a risk and it may happen but honestly i dont think about it much
But good excersise will be checking stoploss at 10% from the entry i know this will kill the system but if it will make you sleep better then be it knowing in the worst case you will lose 40% of the account

(Notwithstanding Len’s point and whether we could actually get a fill at SL 10% below entry ) did you find it made a big difference to your MOC systems Said (or anyone else that has tested this) ?…I tried it just then after reading your post and CAGR dropped 1-2% but still seems viable…guess not many days in backtest when individual stocks drops 10%….

I could imagine myself doing this and getting out of the bottom of the flash crash before it rebounded to normal..so perhaps sell half of position at 10% SL ?

out of theoretical interest,next issue would be if was possible in IB without changes to API ? …

No I did not test with stop Loss on MOC.
But i think i will test with the account stop I mean exiting all positions at market after certain loss for a day such as 30%

[Author]

Posts