- Joined
- 28 December 2013
- Posts
- 6,352
- Reactions
- 24,221
Skate.
Dump it Here
- Thread starter Skate
- Start date
Skate.
- Joined
- 28 December 2013
- Posts
- 6,352
- Reactions
- 24,221
Skate.
- Joined
- 4 December 2008
- Posts
- 486
- Reactions
- 281
Backtest and explore are looking for different things. Explore is looking for signals generated in that period. It will include signals for the following week. (trades to be taken outside your window) Backtest will look for any trades opened during that period. The signals for those backtest trades occurred in the week prior to your window.
Or maybe more simply - The backtest results for that period should match the exploration results seen at the end of the previous week.
Below is an example, I used Jan this year as 2019 didn't give me many results.
Explore for Sun 19th Jan to Fri 24th Jan:
Backtest for the same date (does not match the above):
Exploration run at the end of the week prior to the backtest window - Fri 17th Jan:
(These explore results from the week before match the backtest results above)
@Skate I was testing CAM and GMMA, not WTT. Anyway, I was planning to look on WTT, so I hope I will post results this week.
- Joined
- 28 December 2013
- Posts
- 6,352
- Reactions
- 24,221
Putting out a call
There must be a member who trades the genuine "Radge WTT" who is willing to post up a 6 month backtest.
This can't be correct
I'm sure the turnkey "WTT Strategy" would perform better than the backtest I've posted below.
The backtest period is from 1st January 2020 to 30th June 2020
This was a period that caused most traders real grief.
Skate.
- Joined
- 28 December 2013
- Posts
- 6,352
- Reactions
- 24,221
Skate.
@Lone Wolf @Skate Thank you gents, you were both right. a. i had the settings incorrect and b. The backtest vs exploration matched the previous week. ill post the settings up in the next post just to make sure they are correct. I am also looking for someone to do a 10 yr back test on the code with and without historical constituents. I have made some modifications I am happy to pay someone to do it as well, I just don't want to purchase the platinum just yet...I need to see value in it. whos keen?
The universe that i tested on was the ASX current ( 2528 symbols ) vs current & delisted ( 4615 symbols ).
So around 2000 extra companies don't exist today that did exist in the past.
The 2 main reasons that we're interested in are.
1 - It does really really badly and dies. Most of the times this is a slow death and trend systems will be well out of these. Sudden deaths can and do happen and can't be avoided.
2 - It goes really really well and they get merged, takeovers etc. This is where Trend systems should already be in a well established trend.
Looking back further than 2-3 years ( especially on the low end of town ) and you're missing out on all the companies that did so exceptionally well that they don't exist anymore.
So around 2000 extra companies don't exist today that did exist in the past.
The 2 main reasons that we're interested in are.
1 - It does really really badly and dies. Most of the times this is a slow death and trend systems will be well out of these. Sudden deaths can and do happen and can't be avoided.
2 - It goes really really well and they get merged, takeovers etc. This is where Trend systems should already be in a well established trend.
Looking back further than 2-3 years ( especially on the low end of town ) and you're missing out on all the companies that did so exceptionally well that they don't exist anymore.
- Joined
- 12 January 2008
- Posts
- 7,169
- Reactions
- 17,743
Amibroker back testing question related to current topic.
Our system starts a trend trade in a stock eg Bellamy's and the system holds it until it's taken over by China.
In real life the position is sold before Bellamy's gets delisted. How does the Ami back tester handle this situation?
Our system starts a trend trade in a stock eg Bellamy's and the system holds it until it's taken over by China.
In real life the position is sold before Bellamy's gets delisted. How does the Ami back tester handle this situation?
you can tell Amibroker to sell any stocks before they delist in your backtests https://norgatedata.com/amibroker-faq.php#exitpriortodelisting
- Joined
- 13 September 2013
- Posts
- 988
- Reactions
- 530
the WTT book from memory was written for the russel 3000. Hence the high(er) price filters.
- Joined
- 28 December 2013
- Posts
- 6,352
- Reactions
- 24,221
Risk & Reward
Trading is more difficult than most realise because it's a trade-off between the risk you are willing to take versus the reward associated with that risk.
Backtest
The backtest results below are irrelevant.
Visualisation of the Risk versus reward
I've posted the backtest so you can visualise the Risk versus Rewards of trading.
The STOCH Strategy
This would be the perfect strategy for a new trader on their "training wheels" as the strategy has a high win rate with half-decent returns (rewards). Trading infrequently also somewhat goes to elevating the emotional side of trading.
Disclaimer
1. The Flying Pelican Strategy (currently being traded)
2. The ORB Strategy (currently being traded)
3. The STOCH Strategy ("never traded" - as it falls well short of my risk tolerance)
Skate.
- Joined
- 28 December 2013
- Posts
- 6,352
- Reactions
- 24,221
Skate.
- Joined
- 8 June 2008
- Posts
- 12,420
- Reactions
- 18,071
just realised I made a mistake in my spreadsheet: and a positive result I did not loose 11k, "just" 6k which is not my worst day ever :-! and then in line with expected loss based on the fall yesterday...hopefully will catch up later on, but we are in a weird market not helped by the coming Xmas/NY break
yup. does seem to be a weird market. i lost a lot of profits yesterday but rebounded a bit today. i seem to be following closely to you and skate with my fluctuations.
- Joined
- 8 June 2008
- Posts
- 12,420
- Reactions
- 18,071
yes small rebound today.All good but no real direction at this stage
my weekly is doing good. its my monthly and daily that are suffering right now tbh.
- Joined
- 4 December 2008
- Posts
- 486
- Reactions
- 281
So in your test, you had significantly better results when you used the historical constituents list? Is this consistent with other systems you've tested? My findings have always been the opposite. I don't think I've ever had a system perform better with historic constituents considered.
Are you just doing as recommended by Norgate and adding the historial constituent check to your Buy condition?
"AND NorgateIndexConstituentTimeSeries("$XAO")"
Similar threads
- Poll
