Add mode to analyze data to determin signals

[Aphexus]

You are using a fixed table possible signals. It would be better to analyze the data and determine the most likely signals. Why? Because they could easily change the signals. Ideally you would use AI but to analyze such data you would essentially assign probabilities to all the order sizes(and possibly prices, price differentials, etc. These could even depend on time(ideally) because maybe they use different signals at different times or different days, etc.

The idea isn't that complex as it might sound. Essentially if a signal isn't a signal it will act as noise within the data so you are essentially creating a filter. If a signal shows up with a high probability with downward price movement only then it is likely that could be a signal. You could then view the signal data sorta like a spectrogram and if there are clear patterns then they are likely using signals.

[mlebjerg]

I like the idea! I just don't know enough about AI to make something like that

[Aphexus]

What you could do is simply figure out how the orders clump around price. What you effectively have is a bunch of data that gives you orders sizes and price movement and various other things. (all the data you can get) What you want is to try and find any correlations between the various data and the price movement.

So lets take the simple data of price and sell order size all parameterized by time. That is, you have p(t) and sos(t).

You want to see if there is some correlation between them.

The most basic thing you can do is the cross correlation. That is, you just compute sum(p'(t)*sos(l - t)). This tells you the correlation between p'(t) and sos(t) at different lags. If the price and order changes a lot together then near l = 0 there will be the largest value. If it happens, say, 10 seconds later then large changes in price will correspond to large change orders 10 seconds later.

One would expect there to be a correlation between price change and order size but it all depends(and if the graph has any anomalous patterns it could reveal something deeper going on).

You could filter out only order sizes of X and then use that in the cross-correlation. If there is then strong correlation movement between price changes and those fixed order sizes it would suggest they move the price strongly. E.g., if 123 order size moves the price the most(as compared to all others) then it might suggest that 123 is a signal to move the price a lot.

The basic tool in such analysis is the cross correlation which is just computing the convolution(multiplying arrays together, essentially). But by using it in various combinations and plotting the graphs and looking for things that stick out would reveal patterns.

The main issue I think is that they can put in orders than cancel them(spoofing) and those could be signals and I don't think we can get this data(you can watch it in the trading programs but you can't download the entire data steam). One steal might be able to find patterns though for actual orders that may be used to move the price if they exist.

My guess though is that they are not using orders to do signals(except possibly by spoofing). They can just send text messages(probably encrypted). No need to use trades and such to tell their buddies what to do.

I think what we are seeing, mainly, is automated algorithms that are fighting to beat the others due to large amounts of TRS, ETF, shorting manipulation and retail are not big players in it. It's institutions that are short and long where one will be screwed by the other so they are constantly fighting to either tank the price or spike it trying to destroy the other side. It is likely some shorters have been shorting GME for years(MC, Archegos, and a bunch of others) and using various means such as TRS to get institutions involved. It's a huge mess but I feel that retail is just a small part and most sold out like paper handed bitches because they saw the $$$ and cashed out and sold out(sold their soul because now they have no leverage against the system that will ultimately ruin their lives).

The fact is, the only way to stop this is for enough retail to buy the shares and DRS. If there are more retail than I think or if institutions have accumulated very large positions of long and short and something causes the shorts to cover and those long institutions don't sell then we could see a huge run. The shorters goal is to drag this out and get more paper hands. It's as simple as that, nothing else means anything. What I can say though is that it is(or possibly was) a very huge thing that was going on and "apes" don't have as big a play in it as some think. Us benefiting will depend on other big players rather than us. I doubt the split will do anything but might add a little pressure. Hopefully RC has everything under control and didn't get in over his head.