Jay Vijay

Trying to become a professional cricketer.
Now look at them yo-yos, that’s the way you do it.

[Jay Vijay @jayvijayrising]

My training approach -

How will I become a professional cricketer?

I've started thinking about athletic training as a system-identification problem.

At time t, let:

X_t = state of the body / movement features
Y_t = observed performance

So initially the problem looks like:

Y_t = g(X_t)

But that's not the interesting part.

The more interesting problem is:

X_(t+1) = f(X_t, A_t)

where A_t is whatever was done between the two states.

In other words, don't only model what produces good output.

Model how the system moves from one state to another.

With repeated video, you accumulate:

(X_1,Y_1), (X_2,Y_2), ... , (X_t,Y_t)

But there is another wrinkle.

My ability to perceive movement is itself changing.

Let V_t be the original video and Phi_k be my feature-extraction model at some later time k.

Then:

X_t^(k) = Phi_k(V_t)

The video from six months ago hasn't changed.

Phi has.

So a feature I discover at time 200 can potentially be extracted retrospectively from videos 1 through 199.

The dataset therefore becomes richer not only forward in time, but backward.

Eventually the question becomes less:

What does excellent movement look like?

and more:

What sequence of state transitions produces it?

X_1 -> X_2 -> X_3 -> ... -> X*

If you can learn that transition function well enough, training starts looking less like imitation and more like adaptive control:

observe -> estimate state -> intervene -> observe next state -> update model

Repeat.

I'm leaving out what I'm actually using for A_t and how I'm constructing Phi.

That's the experiment.

[Jay Vijay @jayvijayrising]

The match-winner project (left). Take 23. Sep 17 26

[Jay Vijay @jayvijayrising]

Today’s gonna be an especially hard day to make it to the gym.

But if I can make it in today, all other hard days will become easier. I badly don’t wanna go but this is too big a steal so I have no choice.

[Jay Vijay @jayvijayrising]

One ball closer (left). Take 29. Sep 17 26

[Jay Vijay @jayvijayrising]

@HRH_SHP That’s her, you seem jealous

[Jay Vijay @jayvijayrising]

@saylor @benjamincowen fckn insulting lol because of the politeness

[Jay Vijay @jayvijayrising]

@Read_Meee People want comfort and not the truth. A responsible dad reveals the truth so the kid’s life can be better and real. “Honey, I’m sorry you had a bad day. You want an ice cream?” produces losers.

[Jay Vijay @jayvijayrising]

One shot closer. Take 28. Sep 17 26

[Jay Vijay @jayvijayrising]

rewatched sep 20 26 1057pm, The trap region looks tight, I mean there's not much range of movement there which probably hurts bowling a lot.

[Jay Vijay @jayvijayrising]

One swing closer (right). Take 22. Sep 16 26

[Jay Vijay @jayvijayrising]

@_The_Prophet__ At least 50, left handed, divorced or seriously injured in a long term relationship, a couple of jobs in elite institutions, a few entrepreneurial stints and just now coming into serious money, That’s who you are sight! Tell me I am wrong!

[Jay Vijay @jayvijayrising]

One swing closer (right). Take 29. Sep 16 26

[Jay Vijay @jayvijayrising]

One stop closer. Take 1. Sep 11 26

[Jay Vijay @jayvijayrising]

One stop closer. Take 1. Sep 11 26

[Jay Vijay @jayvijayrising]

One shot closer. Take 27. Sep 11 26

[Jay Vijay @jayvijayrising]

THE MATCH-WINNER PROJECT (left). Take 21. Sep 10 26

[Jay Vijay @jayvijayrising]

@rand_longevity next year you will be posting this nonsense

[Jay Vijay @jayvijayrising]

One ball closer (left). Take 27. Sep 10 26

[Jay Vijay @jayvijayrising]

One swing closer (right). Take 21. Sep 10 26

[Jay Vijay @jayvijayrising]

One ball closer (right). Take 27. Sep 10 26

[Jay Vijay @jayvijayrising]

The videos are making clear the gap between how I believed I moved and how I actually do. The project is harder than I thought but the path is clearer. Overall higher probability now but on a longer horizon