Please don't ever stop pumping out content lol.
Literally just fed this into claude, works on all JPY pairs wtf. Haven't tested it on index yet but this is insane.
No parameter, ATR based SL.
More importantly, I'm looking for more ideas that are similar to this ( market effect, no PA etc )
Claude funnel with an outstanding build again today.
This has had zero modifications outside of the original source that my funnel used.
A specific edge for yen pairs (long & short).
Profitable on USDJPY, EURJPY, GBPJPY, AUDJPY and CADJPY.
Strategy clue: research Gotobi Days.
@MarginCall4 yep precisely - i’m still doing more research but i effectively 8xd the risk after the last month , without taking in consideration the lowest expected DD was much larger than my account
Lesson learnt on risk management - if unsure, size DOWN rather than size UP.
I'll do a full breakdown of this a bit later, but needless to say, I've learnt a lot about how to better manage risk from a systematic perspective.
Keep in mind, I've only been in the algo trading space for about a year now so I'm glad to have learnt this lesson early.
And yes, this is an actual portfolio I ran, not a random screenshot :
https://t.co/0Bv8BBlcvi
@_EldarKurtic "Wow! almost as if one package is a general purpose inference and training stack which supports multiple models whereas one is a specialized inference stack for one model."
@Blinklebloop Could be building a user base/ dependance/ increased adoption/awareness of DS products. Once DS stops open sourcing their models, then we know who the sole provider is
"It's kind of complicated. Each day, you get 30 "hours" in the US or 21 hours in the other listed country.
DeepSeek costs 3 hours for a one hour session of unlimited usage. However: each country gets a further discount of 2 or 3 hours if no other sessions are also running."
Effectively unlimited - insane news, found my alternative to recreate strategy quant overnight lol
We've updated limits to be even more generous.
DeepSeek Flash is now 100% free and unlimited in these countries:
US, CA, GB, AU, NZ, IE, NO, SE, DK, FI, NL, AT, LU, IS, DE, CH
Try it here: https://t.co/uCHw0PuzKO
Does anyone know if @cognition Devin's plan still gives unlimited SWE-2 usage?
I unsubscribed since I thought it was over, but it still says free use offline?
If thats the case iamma resubscribe asap lol, I've been running out of usage overnight with the stuff I'm trying to develop.
To give everyone a sense of proportions - this was something that ran in the background, took about 5-6 hours for the builder I believe?
Kinda expected this result :
10,000 generated in builder ->
55 passed 2x slippage and spread ->
0 passed Out of sample testing
No better test than OOS ( aside from running live )
But learning this is giving me a greater appreciation of the different strategies/filters which are possible, just gotta make it work with some custom logic that I currently utilize
Late to the party for this one - but another avenue for strategy generation.
Basically my main takeaways from this was
1. In builder, choose whatever entries/indicators/exits you want to use ( it doesn't really matter ) - but apply a filter on stability ( pf and Sharpe in SQX don't really matter )
ii ) Also make sure to use lot sizing based on % risk ( never flat lots ! )
2. Choose a relatively short IS test period ( I'm using 3 years 2018-2021 )
3. After a few thousand strategies are generated, feed them through the retester function ( effectively, this just re-runs those top strategies from builder.
4. Retester stages were :
i ) Double the spread
ii ) Double the slippage
iii ) Out of sample testing
iv ) Parameter sensitivity ( my addition )
Only those strategies that make it to out of sample ( and it will be very few, 1% at best I imagine ) - make adjustments to it to
1. Strip away SQX boilerplate code
2. Add in your own risk management/rules as part of a wider portfolio ( e.g. for me that is going to be going flat before high risk news, single entry a day etc )
By no means am I an SQX expert, but I do love the addition of something I can run in the background to mine more strategies.