Daru

Independent quantitative research organization

[Daru @DaruTrading]

DaruTrading → Daru Finance.

Quant research. Crypto. Algo trading education. Memes.

Expect strategy backtesting, walk-forward validation, signal research, and why 90% of "edge" is noise (and scams)

This is for the people who seek to understand the markets and stop losing.

[Daru @DaruTrading]

If you use Monte Carlo permutation testing to validate your trading strategies, your test is uninformative.

Across 437K strategies and 6.6B permutations, MC filtering adds zero selection signal.

This paper documents why and what passes the audit.
https://t.co/ceKlkwOyBO

[Daru @DaruTrading]

@gem_ilm gaming chair cuz u can laydown

[Daru @DaruTrading]

A false-positive bound holds against the test's null (trade-sequence exchangeability), not against 'this strategy has live edge.' On the metrics practitioners rank with (PF/ROI/Sharpe) the test is degenerate under fixed notional sizing, so it bounds nothing, its rank is a constant. And the false positives that matter live are between-strategy data-mining artefacts, which a within-strategy shuffle can't see by construction. That's what White/Hansen/Romano–Wolf are for.

[Daru @DaruTrading]

If you use Monte Carlo permutation testing to validate your trading strategies, your test is uninformative.

Across 437K strategies and 6.6B permutations, MC filtering adds zero selection signal.

This paper documents why and what passes the audit.
https://t.co/ceKlkwOyBO

[Daru @DaruTrading]

@perplievm Wrong, a lot of the strategies in the study had a profitable run, and Monte Carlo did not inform anything about it.

[Daru @DaruTrading]

@x1cryptopulse Nothing related to Monte Carlo in the paper showed any meaningful forward selection.

I got much better prediction power from TDA and portfolio creation. You can find more in https://t.co/lql8F7Ihq7 and https://t.co/CKqBSAnTqK (TDA)

[Daru @DaruTrading]

@x1cryptopulse Nothing showed any signal, only bar level monte carlo permutation but it only barely beat noise, its not worth the several hours of computation power required

[Daru @DaruTrading]

@ChaletMocha @TheCryptoDaddi @grok Thats only about losing weight, and whatever you lost in those 2 months you could've lost naturally with about the same effort

[Daru @DaruTrading]

@SolShadowCat @daraofsoll Dara glazes me every other day, i mean he even copied my name… He’s just jealous

[Daru @DaruTrading]

@SolShadowCat thats crazy, ur on another game you clearly give money away to help and you dont even mind the publicity, definitely one of the few people in CT that do that

[Daru @DaruTrading]

@SolShadowCat ur right. Depends on what you do them for, me personally when I did giveaways I would not want to pay people who just started following and never engaged with me before

[Daru @DaruTrading]

@TheCryptoDaddi @grok Guess you fell for the propaganda. Also HGH is slow acting and very subtle in comparison to steroids.

But if you are this offended then you could go ahead and take them for 6 months and I'm sure you will get nothing but a "better genetics" natural transformation

[Daru @DaruTrading]

@RonPull I'll never forget, you got me my first 1000 followers

[Daru @DaruTrading]

@perplievm Sure, I'm not saying they are useless for all cases, but for strategy equity curve permutation across 400K+ uncorrelated strategies, it has proven useless for forward selection.

[Daru @DaruTrading]

@fivo_dot_com This is just the beginning, my next paper will have at least 2 million strategies using different strategy groups like market making, funding carry, statistical arbitrage and more!!

[Daru @DaruTrading]

Detection isn't prediction.

[Daru @DaruTrading]

The premise everyone gets wrong: "find a strategy that works."

Across 1.6M tested walk-forward with costs on, almost none do.

Selection is the alpha, not the strategies.
Done right, 8 of 12 crypto markets cross break-even.

https://t.co/8r2yfIIykU