Quote from dhpar:
Some good questions where answering may help to OP as well. Unfortunately I am out on yacht for the next two days so I can answer only when I get back - sorry about that.
One thing I see straight away - our wires may have got crossed with respect to how we use the word "prediction". You seem to use the verb "to predict" in similar way as "to forecast", i.e. to tell where the price is going to be in future (up/down/level) and bet money on it. In this meaning a typical quant does not predict. That's a work for analyst, research and most importantly for a trader.
For me it is true that when I read about the markets and review papers on the financial industry, often I see that predict and forecast are synonyms. The majority of quants are doing other things I hope to be finding out about.
Your view that predicting and/or forecasting is the job of three other groups is noteworthy. I do not list predicting and or forecasting as part of the work of colleagues or workers who do analysis, research or trading.
For making money I am interested in repeatability, reliability, effectiveness, efficiency and optimization all with respect to what the market s are offering.
Traders, I believe, achieve execution results by only giving consideration to the presnt, I call it NOW as well, during RTH's.
traders are equipped to recognize a finite number of possible complete data sets. For each they have a known and documented conclusion (which I call analysis). Analysis is recognizing a data set and pairing it with a concusion form a finite set of possible conclusions.
Trading is a canned process in real time where a stream of conclusions are generated in real time. This is C# programming for automated trading where the trader is a computer under automated running of capital.
The five associated platform signals are all prescribed by the conclusion with which the decision is paired. It is just a series of stacatto actions.
The analysis and research you speak of is not connected to this trading. Before and after RTH, lists are made available for equities that need to be traded. As RTH happens turns are done according to the cycle of price for stocks on the list. when done automatically, there is no trader participation once the list is loaded into the execution platform.
So it tuns out that the present prevails as the time of trading activity. there is no predicting or orecasting since the trading is done as a consequence of just taking data sets and acting accordingly by a prescription that is in place for making money.
However quant predicts where the price should be as of now based on an expected behavior of his model. He then recalibrates either the whole model or the model's inputs to observable market prices (if available).
I see from what you say that a model is operating and it uses data. The model behaves. At NOW it has a value(s) that may or may not differ from the real values in the present.
The quant does work as a consequence.
I will be very interested to see how the behaviour of the model is used by other workers doing other things besides what the quant does. Perhaps how analysts, et al, either singly or as a group, use the knowlege of what the price should be if the price isn't what it should be. Moreso, I would like to know what others do when the price is what it should be.
My orientation is to get the degrees of freedom offerredand use them to generate, ontinually another 70 degrees of freedom. the pertinent subset of binary outputs are monitored, analyzed (pairing), decisions made and actions taken. The pertinent subsets are determined automatically. All of this is, effectively streaming during RTH's. What shows is the present and an historical record is also created, all binary.
Then the trader (may) use this model to decide on hedging for the next time step.
My understanding is that the model output delivered after corrections provides infomation for traders. These traders deal from time step to time step. The corrected model information is used for decision making for the next time step. The trader uses it for hedging in the next time step.
I am not hedging at any time on the other hand and our traders do not do next time step decision making. They only do decisions in the present based upon present market binary data sets. This is the front running concept relative to your traders.
We are parasitic, sentiment oriented, speculators who front run informed traders (subject to the quant adjustment delay) that hedge as a strategy.
Of course quant is also able to forecast where the price is going to be given the set of parameters but that is trivial - it would be similar to forecasting value of your book tomorrow when you are told tomorrow's settlement prices.
I do see value in this activity you describe as trivial. It would not be a good idea to suggest the name first given to our equities paradigm. It just seems that it is natural to use leading indicators of price. As the leading indicator is time stamped and moves into the past, sooner or later, as the future comes into the present, the price condition, circumstance or situation occurs.
We regard this as a major advantage we have for front running the markets.
So in the real sense quant (usually) does not forecast.
So, If I understand you, the quant model is adjusted as required to provide current data (Now or after adjusted) that allows a trader to hedge in his next time step.
Our traders do Now by always bieng in the market; staying on the right side; and trading at the capacity of the market. This is a hold and reverse pool extraction paradigm that uses binary data elements in subsets of a finite set. Reversals in a timely manner keep use on the right side of the market and precipitate periodic pool extractions.
I will try to explain it better later.
(and disclosure: I left WallStreet some time ago so there may be some fresher insights - I am an independent trader).
.