ML PPO bot training Talk

That is absolutely something I will take into consideration, I'm 100% back and forth on this whole volume thing as an indicator. Right now I'm testing. What works, what doesn't, will weed out what doesn't work.

The argument I've heard before in regards to volume being so broker specific is a group mentality type of argument.
If 90% of traders are putting in longs on NASDAQ cause of X and y and z Indicators, then usually that fits around the board on other brokers.

When I have two charts running side by side, one through topstepX and one through my personal tradeovate acct, volume and volume profile are pretty much identical. Volume amounts can vary significantly depending on the size of the broker and their client base. But the general consensus of more going long or more going short stays the same.

Actually a cool experiment might be to do volume based trades solely on either the broker with the absolute highest volume amounts and some relatively new broker(with less clients) that has less volume amounts.
Maybe even find what broker has the highest number of hedge funds as clients and see how their volume is being distributed since hedge funds tend to be on the opposite side of trades as a lot of us.

Curious if something like that could actually be a leading indicator.
Hedge fund broker > volume significantly rises(institutional activity)
Large retail broker > volume picks up following price action due to initial institutionial activity.
Small retail broker > slower reaction but the same none the less. Volume picks up due to price action.

Hmmm.. lol
There is also a chance that could be all nonsense as I don't fully understand exactly how institutions and hedges funds trade TBH.

For example I don't understand high frequency trading. How are these guys placing SO MANY TRADES in meer seconds without racking up an Exuberant amount of commissions and contract fees??? Anyways just a random side note question I've always wondered. Do they broker themselves and not have to pay the fees? I thought fees for futures contracts goes to CME or whoever provides those contracts. I dunno

This is what I use to interpret volume for my bots. I retrieved an explanation of it online cause I wouldn't be able to describe as simply and elegantly as this haha

... regarding volume you may have a look at ryuryu's posts here in FF - he developed some bots, reading price and volume from crypto ..

https://www.forexfactory.com/thread/...xaubtc-trading

Mucky

There is just always something. My first bot blew my practice account. So I had to get a new practice account, when I got a new practice account that also changed my account ID and I didn't realize that. So I had to figure out how to get my account ID again which on top steps not easy. You have to do it through the API and through python code. It's not just a copy and paste from the account page. Ridiculous if he asked me.
Anyways, I finally got the account ID found start of the bot and I changed the account ID.
Unfortunately, I did that in incorrect order so I waited 50 minutes before my bot started to make it decision only to realize that it was still trying to access my old practice account.
So therefore I had to stop the bot and go through all my codes to make sure that every account ID was updated. Reloaded the entire script and started it again.
Now I'm waiting patiently for another 50 minutes.
My bot has a 50-minute buffer to where it's reading the market before it starts to make decisions. After that 50-minute buffer, it makes a decision every minute

So far so good.

Obviously could be a simple coincidence. It placed sell trades and buy trades at the right time.
Am going to let this run 24/7 for a week. Retrain at the end of every week with the previous weeks data included.
Will analyze patterns in Time of day wins vs losses(Im assuming itll match up perfectly with the london and newyork market open->close.
Will see!

Stay tuned!
Feels good actually.

Update:
As the bot ran all night.
It lost 1000$
Still gonna let it run so I can accumulate time based data for best times to trade.
But regardless I have another 6 bots I want designed and trained and loaded into this API.
Will see what the weekend brings in terms of a new bot to run on the back burner. Gonna let this run for a few days after the weekend.
1 day isn't enough data

Okay so 100% back to the drawing board. Instead of aggregating all of my take-by-tick data into 1 minute bars. I'm just going to train a robot to learn how to trade with the tick by tick data. Tick by tick includes best bid best ask. Price. Time stamp. Volume.

I will list features that I use later when I come to a conclusion of what the features I want to use will be haha