Machine learning in trading: theory, models, practice and algo-trading - page 3101
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Well, this is how the histogram looks like (according to excel version)
You can see that there are months where the patterns did not work.... and they should be explained by other splits, but preferably by removing them where they accumulate.
And, what I want - if not to train the model so hard, then at least to detect in advance, perhaps probabilistically, the section of change from a positive shift in the quantum segment of the probability of a favourable outcome to a negative one.
Here's thinking about the target one here, the sample is already forming.
If we represent the percentage of delta of positive and negative outcomes in the quantum segment for a month as +1/-1, the graph has already this picture - and it already looks more interesting.
What's the question, that's the answer, don't be sorry
There is some quantum crap cut that is performing poorly somewhere, what to do. average its signals given new data where it performed poorly, so that it performs not so poorly there, but not so well on past data either.
What's the question, that's the answer, don't be sorry
Average, subtract and divide :)
Anyway, as I understand it, are you suggesting to change the target at the site where the "bad" signal is?
Alexei Nikolaev in blogs on R implemented a model of the game Cafe, or the victory of the minority, similar in terms of the market, if the position of the player is in a society with fewer participants, he wins (in the cafe, according to the date, players who came on the day with the smallest number of visitors win, and with a large number of visitors lose), but this is too simple model, in real life there are still a lot of types of players, ranging from the state and other large players and small players, which are a large number. The model is not even roughly created yet)
But the graphs there are even very similar to tick wandering.
The SB model in pricing is a basic, limiting variant, which apparently never happens in reality, as an analogue from physics - an ideal gas. This model is obtained under two conditions - a) a large number of participants; b) absolute independence of their trading strategies from other participants. It is clear that the second condition is difficult to fulfil, so we can investigate how the deviations from the SB model will be influenced if, for example, there are several groups (clusters) of participants with different strategies on the market. Or some part of participants has insider information. It is impossible to make money on pure SB by definition, you can make money only on deviations from it.
Personally, I don't see any use for the SB model at all.
It does not give anything, it does not emphasise good properties, it does not suppress bad properties, it does not simplify.
Yeah, the graph looks like prices, so what?
The SB model itself is of course of little use, it is only useful when deviations from the model are diagnosed.
I call stationarity the usual econometric term: constancy of the mean and variance. The markets naturally do not have such, they are not a "monument". Heteroskedasticity is removed, the rest is close to SB.
In general, the type of distribution says little about predictability. It is such a mathematical game, distant from trading. Add some fluffiness to the quote that covers the spread. Or a steady return to the mean at certain times of the day. The spread won't change, and it will be possible to make money. Roughly this and stuff like this can be called inefficiency. To do this, write algorithms taking into account the fact that you can't predict everything, and you don't need to. I would not say that there is such a curse there, just that there are really efficient tools from which you can't get anything out.
Imho of course, but if you build pricing models with understandable deviations from the SB, then you can then, for example, generate artificial quotes on this basis, even for a thousand years. Then on this quotation learn with the help of MO to determine those places where there were deviations, and then try to do the same on real quotations. Alternatively.
The SB model itself is of course of little use, it is useful only when deviations from this model are diagnosed (found).
Regarding market models - I think that we should initially create some kind of online game, like an RPG with an advanced clan-state, where each clan has its own currency, where there is a real market, and watch how prices change depending on events in the world. Previously, such a portatype could be the second ruler, where there was a market and at least two currencies. You can at least make monitoring of changes in average prices of resources on its basis and then add there the emulation of trades - with the history of quotes. In general, according to my feelings, the laws of economics work there, there is inflation, there are illiquid expensive assets, there is a need for quick sale, there is a need for labour, and conversion operations. I.e. the point is to investigate the economics and the influence of an organised exchange market on price changes (several servers will be needed - where there is and is not an exchange). I think that there you can identify patterns and make predictions based on analysing the development of the game world.
And, everything will come to the importance of indicators such as the balance of trade, money supply, labour cost, GDP per capita, interest rate (there is in the game and this mechanism), orders for the production of equipment....
The topic is interesting, but financially costly.
Simplified models, without researching global ones, will not give an understanding of the market - so you have to go from complex to simple.