Back-testing of some candlestick patterns 🧵

I hv back-tested patterns shown below on 500 stocks (Nifty 500 stock group as on today) since inception on daily timeframe charts.

Below is some interesting information.

Let me tell you an amazing story...

In the 1990s, a maverick breast surgeon at @TataMemorial (fresh from his return from the UK) stepped up to do research. Now, to understand the situation, you should go back 30 years, when research was not as big as it is now, and certainly not from surgeons.

Surgeons, and especially cancer surgeons, were renowned for their technical prowess, and their sheer bravado – "wherever the cancer, however advanced, I will take it out". So, our surgeon-researcher was ridiculed for even attempting clinical research

For a surgeon, he couldn’t have chosen a worse topic to research on: early detection; nothing to do with surgery, or even treatment. Remember, this was the 1990s. Cowboy surgery was celebrated, and research ridiculed

Being a breast surgeon, he was troubled with women consistently coming with advanced cancers, and he set out to see if he could work on picking them up at an earlier stage. But community-based early detection needs money, and he just didn’t have it.

“Why are Tumblr and Facebook cracking down on sex so much? And why is Facebook in so over zealous about it? It’s all Apples fault!” Etc etc

I can explain all. It is largely the fault of a set of US laws called SESTA and FOSTA.

Important paper from Google on large batch optimization. They do impressively careful experiments measuring # iterations needed to achieve target validation error at various batch sizes. The main "surprise" is the lack of surprises. [thread]

https://t.co/7QIx5CFdfJ

The paper is a good example of lots of elements of good experimental design. They validate their metric by showing lots of variants give consistent results. They tune hyperparamters separately for each condition, check that optimum isn't at the endpoints, and measure sensitivity.

They have separate experiments where the hold fixed # iterations and # epochs, which (as they explain) measure very different things. They avoid confounds, such as batch norm's artificial dependence between batch size and regularization strength.

When the experiments are done carefully enough, the results are remarkably consistent between different datasets and architectures. Qualitatively, MNIST behaves just like ImageNet.

Importantly, they don't find any evidence for a "sharp/flat optima" effect whereby better optimization leads to worse final results. They have a good discussion of experimental artifacts/confounds in past papers where such effects were reported.

Shri Hanuman had boon to be immortal. And be present where ever there is Ram stuti.
I came across this beautiful story that I would like to share
#threadstories
#Hanuman ji is believed to live in the foothills of Himalayas. But once he spent a lot of time in Sri Lanka
1/9

Where Ravan's brother ruled for many years..
On his way back he came across tribal people from community called #Mathanga
He gave Darshan to them. The chief of the tribe asked him to keep coming back to guide their generations
Hanuman ji said that whoever recites this mantra
2/9

I will give them Darshan. But there are two conditions
1st he/she must know his relationship with me
2nd there must not be anyone around him in 950 mts radius. If present, that person must also know about my relationship with him /her.

He promised to come after every 41 yrs
3/9

The Mathangs also maintain a ‘logbook’ where they record all the words spoken and ‘leelas’ performed by Lord Hanuman.
They call him their Guru
Who teaches them how to live their lives.
A Sri Lankan organization called Setuu is studying this ‘logbook’ and also conduct
4/9

programmes for Hanuman devotees who can do ‘Sadhna‘ under the guidance of Setuu masters. Devotees who qualify all the criteria will be eligible for Hanuman’s ‘sakshaat darshan‘
The Setuu has certain conditions that need to be kept in mind which includes devoting one complete
5/9