[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / r / s / t / u / v / vg / vr / w / wg] [i / ic] [r9k / s4s / vip / qa] [cm / hm / lgbt / y] [3 / aco / adv / an / asp / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / qst / sci / soc / sp / tg / toy / trv / tv / vp / wsg / wsr / x] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/sci/ - Science & Math


Thread archived.
You cannot reply anymore.



File: ture_1.png.full.png (256 KB, 1272x793)
256 KB
256 KB PNG
Hi /sci/
Asked on /biz/ but nobody gave a fuck about it.
QUESTION: does technical analysis worht comething? it looks kinda esoteric to me.

I suck at math, but as far as I know, a graphic is about the information you have. you can predict certain things if they fit into certain pattern, but I don't see how stock market would do something like that. so this forecasting because "here it looks like a triangle and that means it will break the resistance and go up"... well...
>>
value is not a consideration for investments post-2008, so you're only interested in where the line is and where it is going next. technical analysis is a great tool for this
>>
>>11854241
TA is widely discredited. Read some academic literature on the subject. Or if you want something more mainstream, try "a random walk down wall street".

>>11854249
>/sci/ is a retard when discussing business
lel
>>
>>11854253
>Thinks the market reflects actual value rather than investor sentiment
You really got us!
>>
>>11854241
The quant firms that use algorithmic trading pay the exchanges to house their servers literally in the building
They aren't trading on technical analysis, they are scalping arbitrage opportunities that arise from latency delays
TA and actual algo are scams
>>
>>11854278
>market price doesn't have anything to do with value
Look, some opinions are so retarded they can be dismissed offhand. You appear to believe one of them.

If value is irrelevant, why don't you go buy a stock with a 150% dividend yield. Stocks like that should exist if value has NOTHING to do with market price.
>>
>>11854241
Texas sharpshooter fallacy
>>
>>11854309
well, this is one of the biggest problems now. markets don't care too much about real world and the economy makes no sense.
not saying that market prices don't have nothing in common with value butI don't think this is the best argument.
>>11854433
so, post-hoc bias?
>>
>>11854241
Technical analysis is worthless and you should steer clear of /biz/ if you're looking to discuss economics.
t. econ grad
>>
>>11854953
Stay in your lane good goy econ grad. Academics know nothing about trading.

None of these people know anything OP. Theu never studied TA for years. that's why their studies fail to be reproducible these days.
>>
>>11854253
Imagine getting rid of a useful tool that takes into account that market trends also go through their own rhythmic cycles just like everything else in nature both short-term and long-term. Go back to finishing your freshman year of undergrad, kid.
>>
>>11854241
Trade long term on fundamentals. Anything else will result in tears if you're a normal human being.

There's algos that abuse technical analysis concepts.

There's algos that analyzes all available active positions and orders and do market-moving trades just to crash out your orders on their stop loss.

They are there to pillage your wallet. To take the money of newcomers to the stock market. It's big business to pillage noobs.

Every day hundreds of people with a bit cash to spare enters the stock market for short term trading after reading on google for 45 minutes.

Every day a lot of them are robbed by algos. Don't be one of them. Trade small in the beginning and be patient.

There's no get rich quick scheme in stocks, only get poor quick.
>>
>>11854892
>post-hoc
yep. everything eventually fits into a "pattern," but you never know where in the pattern you are
>>
>>11854980
He's qualified, you're not. So sit down bitch boy.
>>
>>11855010
Appeal to false authority, fucko. Econ grads can't even use math as a tool properly and think consumer preference is something that you run through a function. Not only stay in your lane but try and actually repair it.
>>
>>11854241

We had a trading robot competition on my computer class.

We first need to understand why we are using some tools like TA and what they are for.

In fact, a simple logic would be to buy an increasing stock and sell it as soon as it start to decrease.
Problem: how can you know if a stock is decreasing (if it decrease by 0.01$ but only for 1second, you may not want to sell)?
Solution: use a lagging indicattor to determinate the tendency of the market (MACD?)

Or you may buy an action and hold like 6-7 months: the market have a positive tendency on a long run (Buy and Hold Strategy can hold stock for years).
Problem: what if the market crash?
Solution: use an indicator to determinate an exit strategy (Chandelier Exit?)

So what is the purpuse of TA?
Imagine, you give me 1$.
I roll a dice.
If it is 1-2, I give you 1.20$.
If it is 3-4-5-6, I give you back 0.20$.
This situation is not worthit at all, you have 1/3 chance to gain 20% of your money and 2/3 to loose 80%.
Better go to casino.
With TA, you can evaluate the increasing/decreasing range, the risk of a reversing tendency, etc... (they are many kinds of indicators, just play with them and find the better mix for a random stock)
TA cannot predict the futur, it is just a mathematical tool to be on a statistically winning position.
Imagine, you give me 1$.
If it is 1-2, I give you 0.80$.
If it is 3-4-5-6, I give you back 1.80$.
With this situation, you better play many times.

The reals problems with TA is to determinate the tendency of the market, the trading strategy (long or short run - generally linked to the tendency of the market) and find which indicators to use for which stock (MFI+KC? RSI+OBV+BB? etc).
>>
>>11855020
>Appeal to false authority
Unless his degree is false, it's a valid authority,
>>
I notice that in my previous post:
>>11855059
I explain the purpuse of TA but I did not give any exemple.

So here we go.
Let take the Bollinger Band and use it as a momentum indicator.
You made an average of the price and then you draw 2 lines to covers 95% of the prices (or more, your choice) of a PAST period of N days.
If the price crose the upper line, that means statistacally the prices have a high change to decrease and go back to the normal (the 95% case).
Same if the price cross the lower line, it should go up, you better buy.
You may even use thouse lines to figure out the volatility of a stock (if they are far, the prices change a lot).
But same as always, one indicator is not enough... A trend indicator is generally needed to confirm the tendency of the price. Exemple: if you are lucky, the price can cross the upper line and then the upper line increase and cross back the price (the new price become the new 95% case - you should not sell).
>>
>>11855103
That's cool and all but monkeys throwing shit at a finviz print get better returns than TA.

In terms of your specific example, monkeys beat Bollinger bands because Bollinger bands make you close out winning trades too soon. All you're really doing is churning commissions for your platform.
>>
>>11855069
Also all economics degrees are false, so none are a "valid authority"
>>
Continuity of (and ending i guess):
>>11855103

Unfortunatly, I can read, but my english writting suck hard (cliché about French's language skill seams to be true for me).
Anyway, if you are really interesting about TA, I recommend you this paper:
https://mpra.ub.uni-muenchen.de/64521/1/MPRA_paper_64521.pdf
Bonus: starting from the page 49 of the pdf, it show some tables to benchmark the return of different indicators, which is pretty cool.
>>
>>11855121
This is why I specify the need of a trend indicator on top of BB.
(Not sure about the monkeys, they may be too random... Not BB, even used alone)
>>
>>11855069
It's still an appeal to authority fallacy, you fucking retard.
>>
>>11855130
>trend indicator on top of BB
BB aussi indique la tendance, les deux s'incestent
>>
If you want to know if TA works, don't ask some econ student or people repeating the same tired lines from A Random Walk Down Wall Street, ask a professional trader or someone who has back tested a strategy based solely on TA and compared it to a control without it.

To state that historic support and resistance has no influence on price is absurd.
>>
>>11855147
Complètement d'accord, je me suis mal exprimé.

Avec BB tu sais si l'évolution des prix est une irrégularité statistique.
Maintenant, si tu ajoutes un MACD par dessus, tu pourras confirmer cette tendance, mais ça vient surtout du fait que MACD a du lag ^^'
>>
>>11855147

Après, pour commenter:
>>11855164

Je ne suis pas étudiant en économie, mais en informatique. (ÉPITECH en plus, on n'a pas de professeur...)
C'est un robot qui a corrigé nos robots en leur envoyant de vrais cours d'action (historique Yahoo).
Du coup, si on a un retour sur investissement positif, ça veut dire qu'on aurait eu un retour sur investissement positif dans la réalité.
Note importante : la simulation ne prenait pas en compte les stocks. On pouvait acheter et vendre à tout moment.
Mais j'en ai conclu que le TA, c'était puissant.
>>
>>11855201
Non. Tu réfléchis comme un typique francais philosophe et reveur. Il te manque le pragmatisme américain.

Les performances en backtest ne reflètent pas la réalité puisque tes ordres n'ont pas influencés le marché. Ce sont des résultats purement hypothétique.

N'oublions pas que même si les français sont les meilleurs en math au monde, ce sont les américains qui sont partis sur la lune.
>>
>>11855242
>même si les français sont les meilleurs en math au monde
ok boomerbaki
>>
>>11855242
The act of ordering doesn't necessarily change the price depending on the size of your order and the volume being traded. For example, most brokers offer fractional shares so it's possible to by just a few cents worth of stock at a time which would have an imperceptible effect on the price and back testing being hypothetical doesn't disprove the validity of it or TA. As a small fish you can't really change anything in the market regardless of what you do unless you're trading stocks with a very small market cap.
>>
>>11854241
There isnt a market in existance that isnt controlled by the powers and principalities. Save yourself a lot of time and dont even bother. If you have to be told this you are not of the proper bloodline to make money off the scam.
>>
>>11855242
jesus fucking christ—and they say american education is bad. You can’t even write your own language right, frogtard.
>>
>>11855242 n
Ça reste du backtest.
Ici, on ne s'intéresse pas l'impact du robot sur le marché, mais à sa capacité a prendre des décisions.

Vu le flot d'argent qui transite en bourse, je ne pense pas qu'ajouter ou retirer un acteur puisse influencer le marché sur le court terme - à prouver.
Pour ce qui est du long terme, on ne veut que tester si notre robot prend toujours les bonnes décisions malgré l'influence des autres acteurs.
>>
>>11855242

Suite de:
>>11855316

Le thread m'inspire.
Je viens de ressortir mon vieux code (ça va m'occuper), je me permets donc une petite ouverture...

On pourrait ajouter quelques règles du style, si notre ordre d'acheter/vendre X actions inverse un indicateur, alors on réduit X (ou pourquoi pas l'augmenter si on veut brain les autres :p)

On pourrait même ajouter du machine learning *_*
Je vois :

I) Prise de décision en fonction de différentes combinaisons d'indicateur
- on utilise tous les indicateurs qui existent.
- le ML choisi quand acheter/vendre en fonction des conseils des indicateurs (si le ML voit par exemple qu'il est rentable d'acheter quand RSI et BB sont d'accord, mais moins rentable quand RSI et MACD le sont)

II) Prise de décision en fonction de la prédiction du prix
- on utilise tous les indicateurs qui existent.
- on s'en sert comme feature pour que notre ML puisse prédire l'évolution des prix.
- on achète et on vend quand on prédit un changement de tendance durable.

II) Deep-Learning, OSEF
- comme le grand II, mais on utilise aucun indicateur, que l'historique brute des prix/stocks.

L'OP parle du TA?
Là, on l'utilise, mais en complément du ML.
>>
>>11854241
Finance PhD student here. The evidence for TA is weak and it's mostly BS, but longer term trend following work in the sense that it reduces drawdowns and gives you a similar return to the market long term, but it will underperform the market for as long a 5+ year periods. That is it arguably gives you higher risk adjusted returns. Shorter term trend following used to work but it's been arbitraged away since the 80s.
>>
>>11855329
Petite précision pour le grand I.
On utilise du ML, car une combinaison pourrait être rentable à un moment t, mais pas forcément à t'.
Le ML pourrait trouver des patterns pour appliquer les meilleures combinaisons à un moment donné.
>>
>>11855303
Individual traders can have the highest probability of beating the market because you're practically invisible to market makers and you can almost instantaneously enter and exit positions at the market price. Efficient-market hypothesis is a joke, no system can possible understand all the information that will influence asset pricing. Wall Street has an average performance because they are almost all of the market and any outperforming strategy gets picked apart by other people on Wall Street. Contrary to what people think, they're not that focused on beating the market, managing money is inherently profitable because of the upward bias of the US market and most of their customers won't notice a small change in fund performance while their management fees stay the same regardless.

Anecdotally, my lifetime performance is 100x my starting capital. Of course you can easily write this off as whatever bias and think I'm going to lose it all but rational people with a good ability to control their emotions have a huge advantage in the market. Like the house in a casino, even a small edge is enough providing you're not running into gambler's ruin from over sizing your potions. Wall Street has some intelligent people with tools far greater than mine but it also has a lot of idiots that greatly reduce their performance. There are still people on Wall Street who firmly believe EVs will never replace the majority of ICE vehicles. Most people will always be relatively irrational.
>>11855332
You should get more practical experience in the market. It's trivial to see that TA is advantageous, as a simple example just look at something like the 300 level on SPY which is major support/resistance and watch how much more likely the price is to bounce off of it than to break through it. A lot of TA indicators are bullshit but as a whole it's very useful.
>>
>>11855329
Try trading that way in real life.
You win if other people pay you a constant fee to trade, but you lose if you pay other people a constant fee to trade.
Si tu fais pas le marché, tu es le marché
math: “ant on a rubber rope”
>>
>>11855386
et le marché gagne toujours... cest quoi ton point ?
>>
>>11855329
complétement tarré. en quoi ca prédit les tendances ou les conditions futures ? ca se content d'overfit le data.
>>
>>11855379
>there are still people on Wall Street who firmly believe EVs will never replace the majority of ICE vehicles. Most people will always be relatively irrational.

are you this autistic not to spot a liar ?
>>
get >>11855555
>>
>>11855536
que le TA n’est rien qu’une tapetterie financière, construite pour enculer les amateurs
>>
>>11855548
They're usually not lying. A number of analysts, traders, and even billionaires like Jim Chanos have made this their hill to die on. Retail traders have been always over represented in EV stocks and now their going to force mass institutional buying as they included in the S&P 500.

I could have picked from thousands of examples, these types of events play out in the market constantly disproving rationality of market participants.
>>
>>11855591
the end game had always been to fill the indexes with overpriced shitstocks in order to dump the bags on pension funds
>>
>>11855620
Overpriced for sure but historically large caps stocks have outperformed small and medium caps. The institutional traders allow the idiot retail traders and more speculative prop and hedge funds to filter out the signal from the noise leading to decent returns without a lot of risk or effort. Smart retail traders and investors are making bank by being early on the trend of new industry and buying stocks like TSLA years before they're discussed by everyone with a investment account. That's how I beat the market at least, along with the powers of autism giving me a dissociation with money and emotion, allowing me better money management.
>>
>>11855672
thats a worthless statement. stop spewing bullshit you've read elsewhere and think.

its obvious the big company has grown more than the small company... how do you think it happened to get big ?
>>
>>11855672
whats your average CAGR on your leveraged capital?
>>
>>11855700
Just say what you mean. For every small cap that succeeds, there are dozens that fail. You don't get rich just buying small caps or else you could just be in a small cap index ETF.
>>11855705
I've never sat down and calculated it. Since I opened my most recent account a year and two months ago, I've gone from 11k to 98.5k as of today. My return will drop off as I become more conservative in my speculation.
>>
>>11855736
barely beating real inflation lol
>>
>>11855539
> ca se content d'overfit le data.
Pas forcément... Outch, ça ne va pas être facile à résumer.

Je comprends que l'idée d'ajouter du ML (qui est un outil de probabilité) sur des indicateurs (qui sont des outils de probabilité) est terrible.
Je reste bien évidemment ouvert à la discussion, surtout si tu vois une faille.

Ici, j'ai présenté 2 grandes idées. (que je n'ai pas encore testé)
I) Du ML sur les décisions
II) Du ML sur les prédictions

Vite fait, la 3e idée (mineure) ne peut pas avoir d'overfit vu qu'on supprime les indicateurs.
Il va falloir sévèrement tester le modèle, puis le robot trader utilisant le modèle.
Mais c'est du deep learning pur et dur.

Bref...

Pour la première idée, lorsque l'on fait du TA, on utilise des indicateurs qui nous conseillent quand acheter/vendre.
N'en prendre qu'un seul est peu fiable, on va donc faire un cocktail : MFI+CCI+KC+CE par exemple.
On peut acheter quand notre sélection d'indicateurs est entièrement verte. (elle doit donc être assez courte)
Pour la vente, quand un indicateur d'exit est allumé, on quitte. (CE)
Si on vend quand tous les indicateurs le disent, on prend de gros risques. (CE exclus)
Si on vend dès qu'un indicateur passe au rouge, on minimise nos gains.
L'idée est donc d'envoyer tous les indicateurs au ML et de le laisser choisir :
- une liste d'indicateurs à utiliser pour l'achat.
- une autre liste d'indicateurs pour la vente (certainement la même).
- pouvoir changer ces listes pour s'adapter au pattern de l'évolution des prix.
En gros, on ne choisit plus nos indicateurs, on les laisse se sélectionner automatiquement.
Ça doit permettre :
- d'avoir le maximum de vente bénéficiaire (on cherche à être supérieur à 50% ??)
- d'avoir le rendement maximum par vente (on cherche à être supérieur à 0% + taxe ??)
>>
>>11854241
I looked into it for a bit and it seems like schizo shit but not based-schizo shit.
>>
>>11855539
>>11855823

La seconde idée tant qu'à elle ajoute un niveau d'autonomie au ML.
On n'envoie pas les données brutes, mais une liste de prix, volume ainsi que le résultat de tous les indicateurs.
Un peu comme quand on fait du ML sur une bande son... On préfère envoyer une Fast Fourrier Transform pour décomposer les fréquences et là le modèle peut isoler celles qui nous intéressent.
Les indicateurs servent donc de feature pour générer notre modèle.
On se sert uniquement du ML pour décider du poids de chacun.
Au final, on a un modèle censé nous retourner le prix et le volume de t+1 le plus probable possible.
Ensuite, cotée décision, ça devient super facile \o/ On n'a qu'un seul indicateur globale qui nous prédit un prix futur.
Modèle a testé, bien évidemment - sur 30% des données, basedons fous.
>>
>>11855555
Nice quints
>>
It's total bullshit, just buy the market and hold.
>>
>>11855831
basedons* pas "basedons"... Aucune idée de comment j'ai bien pu finir sur ces lettres.

Dernière réflexion après j'arrête vu le sujet n'était pas sur le ML.
Pour le II, on envoie bien tous les indicateurs, par uniquement les RSI/MFI&co capables de prédire un changement.
On doit pouvoir tracer un graphique des prix, les points ne se téléportent pas.
On laisse donc au modèle la possibilité de voir une tendance s'inverser lentement.
Là-dessus, un PPO serait certainement mieux qu'un MACD.

Pour les crashs boursiers, je n'ai pas d'outil mathématique à proposer.
On se mange les pertes et puis c'est tout.
>>
>>11855823
Tant mieux si tes convaincu. Je ne suis pas familier avec le KC ou le CE. Peut être que c'est ce qu'il me manquait.

Merci à toi d'avoir coupé ton potentiel de rendement de moitié sinon plus pour convaincre un étranger sur internet.

Ne valorise tu pas ton temps de recherche ou ta propriété intellectuelle ?

Manques-tu vraiment de validation personnelle au point de te tirer dans le pied à travers chacune de tes actions.

Tu es désolant, mais je remercie l'existence de bénévole en ton genre et ton accès Internet subventionné par l'État.
>>
>>11855870
Je suis pour le partage d'information ;)
>>
>>11855895
Merci ;) Au final une meilleure allocation des ressources planétaires bénéficie globalement à tous.
>>
>>11855905
Rooo, je vois la leçon que tu veux donner, mais je ne comprends pas pourquoi tu le prends temps à cœur x')

L'idée est partie d'un exercice scolaire:
>>11855059

J'ai tout appris sur internet, n'est-ce donc pas le meilleur endroit pour s'amuser à imaginer de nouveaux algorithmes ?
De plus je pensais que tu ne croyais pas au TA.
Mais si je t'ai convaincu, je suis heureux.
N'hésite pas à checker GitHub, tu y trouveras des centaines de robots Open-Source.
>>
I only hear about "technical analysis" in the realm of daytrading. I can give you this study: https://ideas.repec.org/p/spa/wpaper/2019wpecon47.html

If daytrading is so hard as the paper shows, it only means that technical analysis is just snake oil.
>>
File: original.jpg (29 KB, 430x320)
29 KB
29 KB JPG
>does technical analysis worht comething?
>>
>>11856105
>I search for post to make fun of them
>oh look! a typo!
>hahaha... tough look pal
>>
>>11855954
>Brazilian equity futures market
If anything, that paper is just proof that these guys were getting front run by their broker.
>>
>>11854241
it works about as well as a horoscope, or that voodoo magic where you read signs in the entrails of a dismembered rabbit or some shit
>>
bump
>>
>>11857790
I agree with this, but I'm confused. if it is useless, why are people putting loads of money on it?
is there any conclusive evidence that can tell if technical analysis is plain bullshit or it works somehow?
>>
>>11859599
If TA truly worked, it would be disadvantageous to publish the working strategies as the strategy would stop working if enough people to sway the market followed it. I'm not truly in on TA but I assume there are enough strategies that one of them ends up being true in any conceivable situation.

The real indicator of whether it works would be someone actually making lots of money using TA. As it is, TA is a ritual that online brokers etc are expected to offer, much like other market analysis.
>>
>>11859721
>As it is, TA is a ritual that online brokers etc are expected to offer

Ok, now I see TA as reading the entrails of a goat, but also some kind of ritual.
so 100% magic.
>mfw serious economy is driven by voodoo tier ideas.
>>
>>11854278
What do you think value is, retard?
>>
>>11859599
It's a scam to get plebs to pay them to invest their money or to rent out bloomberg terminals to boomers.



Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.