I want to try algorithmic day trading...

The firm I worked at used literally zero TA. I don't think many people there believed in it. I can't use their strategies, so I figured I might as well ask for thoughts about alternatives that I can use

Have a look at the strategy I mentioned. If you have access to tick data, you could try applying the 200 MA on it. Combine reading the order book with your entries, to try add another layer of edge. Watch correlated markets too. For example, The Dow tends to follow the Nasdaq.

I'm sure it's not quite THAT easy... If you could make money using only mechanical entry rules based on moving averages, 98% of retail traders wouldn't be losing money. How do you tell a trend that's actually likely to keep going in that direction from a temporary artifact of random price movements?

There are a lot of HFT firms, and a decent number of people working at them. Is working at an HFT firm really so amazing that it must be larping?

I encourage you simply apply these to a chart, and look at it. It's not "overfitting", it's literally what a lot of people are using, which is why it works. Retail traders lose for a variety of reasons, not necessarily because there are no easy approaches to trading. One of those reasons being we often don't bet with trend.

Markets certainly don't always trend, and you must always be aware of news, and be disciplined with your stop placements, and your draw downs. But I promise you this is a viable way to make money, in the long run. Moving averages don't actually require other people to use them for them to be helpful filters (even though a large amount of market participants use them).

I thought about making models based on simulations of market participants, but it seems like it only extends your problems from overfitting to fitting to completely made-up data

Simple models with only a small number of parameters already easily overfit, a simulation will always have a huge number of parameters and is basically guaranteed to deviate significantly from reality in some unexpected way, all the time

Well, here's what I can tell based on what I know. Outside of cryptocurrencies or some other exotic thing, what you have available as an individual trader is basically equities, futures, options, and forex. Fixed income is another thing but doesn't seem to be very suitable for short term trading or individuals. Stocks require you to post a lot of margin and commissions seem to be pretty expensive. Futures commissions seem to be much lower as a fraction of both contract size and tick size, and margin is very favorable (50:1 in some contracts compared to 4:1 at best for stocks). The more popular futures are also very liquid and have tight spreads. Options offer comparable leverage to futures but commissions and spreads are higher. Also, they are complicated and there is a lot to learn compared to futures. Forex is liquid and leverageable like futures, but it's an OTC market unlike trading on a futures exchange, and your broker makes money by quoting you some arbitrary spread they make up, can reprice your trades arbitrarily, don't guarantee fills even if the price crosses your limit price, all kinds of stuff. The thing that happened with the Swiss franc a few years ago spooks me because there are no guarantees whatsoever about getting fills on stop orders in an extreme market event, again unlike say CME where at least you're guaranteed that your stop orders are queued at the exchange and will fill if at all possible. So overall, futures seem to be the best deal at least in my case.

There isn't much of a futures market for forex, unfortunately.

But if have concerns like this:
> broker makes money by quoting you some arbitrary spread they make up, can reprice your trades arbitrarily
then you could consider going through an ECN for forex: dailyforex.com/forex-brokers/ecn-brokers

there are downsides to that too though, and if you're wanting to trade on low time frames, frequently, then thicc futures markets (like the SP500) will suit you well.

Thanks for the advice. If I decide to spend some time on this I'll start with simple methods like moving averages and see if I can get a machine learning model to show some predictive power with just those. If not then I'll add more complex features to the model.
Honestly my biggest concern with forex is that the volatility is typically low enough that you really need a lot of leverage to make any money, and then you're exposed to getting well and truly fucked by black swan events like the Swiss franc unpeg. I don't want my broker coming after my savings the next time something stupid like that happens and I'm leveraged out the ass

Out of curiosity, what's your opinion on making money in a market, coming from an HFT background?

Do you believe the only way to make money in the market is to identify situations where other market participants have made a "mistake" allowing you to profit off of that, or for there to be arbitrage opportunities that you focus on trying to make money off of?

Is the idea of trends/momentum foreign to you?

With a background at just one HFT firm, I don't have much knowledge of general trading strategies. HFT is very different from other types of trading. At my firm at least it was basically about scalping for just one or two ticks based on really rock-solid arbitrage relationships or microsecond level market microstructure. I definitely don't think that's the only way to make money in the market, but it's the only way I know much about. Trends and momentum are foreign to me in the sense that I've never seriously tried to use them or even looked at how well they work with backtests.

So if you were to see an image like this, do you think it's extremely weird market participants actually, at times, apparently actively co-ordinate with each other to impact price? As is demonstrated by The Dow Jones and 200 day moving average.

Or would you just refer to this as overfitting?

Attached: Dow_Jones_Daily_time_frame.png (1274x623, 49K)

(Takes place on the weekly too)

Attached: Dow_Jones_Weekly_time_frame.png (1274x623, 46K)

Trading alts is almost impossible imo. BTC is possible, but you need to read a bit about it's history and big resistance/support lines, then you might be on something

A chart like that could be explained just by long term investors wanting to buy at that time, without needing any kind of active coordination. If that's the case and it's predictable in some way, there would certainly be trading opportunities in following the trend.

My concern with this is that trends like that happen easily even if you just generate prices with a random number generator. If you want to make money, the question is whether you can look at that chart with the right half of it covered up and say that the trend you've seen so far will continue. There might be just as many charts with a similar first half that reverse in the second half.

>the question is whether you can look at that chart with the right half of it covered up and say that the trend you've seen so far will continue
I think if someone "knew" this was going to happen, then there'd be absolutely no risk involved, and it would essentially be something like low-risk arbitrage opportunity.

Technical based trading incorporates risk management through stops to try and reduce damage from getting it wrong. Technical trading also tries to increase the likely of an expected event to take place by using something called "confirmation".

So for example, if something shot up in price, then had a slight pull back, you wouldn't necessarily immediately buy that price pullback. You'd wait for some large green candle to show up, to indicate that the market is now more likely to keep trending upward.

Another way to put the odds in your favour is a concept called Time Frame Confluence (take a look into it). So for example, it you're trying to go long long on something, you'd be wanting to see as many bullish on reasons on as many different time frames as possible.

As for the Dow Jones right now I certainly don't know what's going to happen. But I have to be bullish, so long as it's above the 200 day sma, and up-trending above the 200 sma.

>1 post by this ID

Useless

Algo(bot) trader reporting in

Attached: bot.png (1917x1040, 85K)

If you're willing to share, what do you trade and how much size? Do you make a living trading?
Same questions

Can you help me to get general idea of HFT or being a quant.

I'm going to comp sci college this year and currently work on some simple Python GUI bot (sentdex tutorial on youtube). I'm a complete noob, but I do swing trade stocks/forex/crypto and have a very good understanding of the market. However, the problem comes in from the programming side of doing things.

What would you advise me to learn during the next 4 years, if I want to become a quant? I've looked the Joma interview with his brother that worked at Two Sigma and he said to focus on computer science, statistics and some general finance/economy.

Currently I wanna do a custom bot that will extract data from GUI program and then paste that data on a website that will automatically do a data visualization and calculate probability of successful trades based on the module that I would use. I'm currently looking to just write a basic module that will scan assets and notify me when they are overbought/sold and calculate win% ratio for given module and find out which is the most successful. I don't think focusing on arbitrage is really worthy, since institutions have supercomputers doing that every milisecond. Maybe in crypto market and smaller exchanges, but let's leave this out.

Any suggestions, as said, I have 4-6 years to do something and am willing to go down this path.

Attached: 1522609028155.jpg (741x568, 73K)

How does that work in general? Do you write your own bot? Do you implement your own algorithm / strategies? Do you just leave it running on or just for some periods of time?

>I want to try algorithmic day trading. Basically just turn price/volume charts into features using basic technical analysis concepts and feed them into a machine learning model.
It may be done. You probably don't need machine learning at all. (I x4'ed my BTC stack with just TA since Jan 1)
Obviously, nobody is going to give you accurate instructions about how you should be applying it, kek

This geezer blogs about exactly that : financial-hacker.com

You are actually wrong, this guy has all the free information on how to get started:

youtube.com/watch?v=HjNHATw6XgY&t=1s

pythonprogramming.net/data-analysis-tutorials/

Well I have never actually used the bot to trade on the stock market , just the crypto ones and the size traded is around 5000(levered).
Is the sentdex tutorial the dash one or the one that he uses tkinter with? Also make sure you have a soild understanding of python before looking into making guis. You want to be very fond of asyncio,multiprocessing,numpy and threads so you can make a nice bot that can crunch large amounts of numbers and can fetch lots of data while having a gui that does not hang.

I myself do not work in quant field , just a comp sci person who thought I could make some nice money putting my skills to use so I cant for sure say what those people will want. But I know that AI(neural networks) are probably going to be very in demand in finance. So learning tensorflow or something like that is probably a good idea as well.
I wrote the bot myself and it trades off my own strats which are pretty successful and I just leave it to trade. If for some reason it throws and error it will send me an email but I have error handled pretty much everything so thats never happend.

I'm primarily a longer time frame trader, so I don't make a living solely off of it, no. I believe it can be done though. As you mentioned before most retail traders fail. Trading, like any other profession or business, isn't easy. I don't know why this shocks people. Most businesses fail. You have to put in time and effort, and even then, it's not like The Market is fucking obliged to provide you with a Salary and Dental.

The reason why I don't make a living from it is mainly due to me having smaller sizes - I'm not a multi-lot forex trader. I have experience in using CFDs with 500:1 leverage. I've traded index futures and index CFDs, usually on shorter time frames than my forex positions.

I'm going to have an interview with a prop trading firm next year, no guarantees I'll get an actual trading job though.

In terms of my trading history, I've been actively trading since early last year. Had huge losses to begin.Then, when I realized what I was continually doing wrong (i.e. betting against trend), I managed to change my approach entirely, make back my losses, and then more.

When there was a volatility spike at the start of this year, in the US indices, I was too complacent in my inexperience, and ended getting hammered. While I haven't lost everything, it hit me bad. I have been recovering though.

Ironically, it actually forced me to improve my filtering processes, and risk management techniques. It's why I'm so big on moving averages, because had I focused on them, it would've gotten me out of a lot of trouble.

Even though I haven't been trading for years and years, and certainly don't have experience in anything like HFT (although those trading strategies really interest me), I believe I've been through a lot. Not just from a trading perspective, but psychologically too.

You can make good profits market making, directional trading will be very difficult.

The tkinter one, but he has dash also. The thing is it's outdated and I got stuck on a problem and kind of focused on learning how to make other things (website/data analysis) and will then return to that.

Here's my problem:

stackoverflow.com/questions/51310289/how-do-i-pass-in-an-index-while-extracting-hundreds-of-columns-per-second-from-a

youtube.com/watch?v=PPRfRV_yIBI&feature=youtu.be

I would take those (((materials)) with a grain of salt, when I have a profitable strategy, I am not selling it, I am not explaining it to anyone, and I probably not even telling anyone I have it, kek

I got my HFT job out of college with just a comp sci background so I can't say what quant companies would want in general, but you probably can't go wrong with statistics, machine learning, and time series analysis. To pass the interviews you need to be pretty good at mental math and puzzles, order of magnitude estimates, things like that. You can find sample interview questions with google. If you want to trade for your own account in the meantime you're right not to bother with arbitrage at all, it is completely impossible without institutional capital and risk management.

Market making from a retail account? It's hard to believe you can make money in that with retail transaction costs. What do you think you can make a market in and make money as an individual trader?

Something is going wrong with your pandas code. I would assume that when you pull data from the api(please tell me you are using websockets) and you add it to the frame its not the same. For example does the api somes times group the data? Or you might actually be hitting an api request limit? Try changing the pull interval to every 5 seconds.

Looking at the vid you are just spaming them with requests , my advice would be to use websockets to get trades pushed to you and then you can handle them. In the mean time just try adding a

time.sleep(5)

after the data had been added to the dataframe.

I'm not saying he is handing out free profitable strategies, I'm just saying that he shows for free how to build a setup from which you build your own custom strategies.

Of course nobody is gonna show a profitable strategy for free. He's just showing how to set up an environment.

ty

Do you have time or discord to talk about this? I'm short on time currently, but as said, I'm going to college in two months, currently working at a company trying to get internship for PLC programming so I have so extra skills in industrial industry.. however, after that and in mean time I'd like to focus on finance also.

If you have discord or skype, I'd appreciate if you could help me or guide me a little bit from the start and give me a reality slap here and there. I don't know anyone going into my direction and I tried few forums, but they are mostly dead.

Attached: BXGV6Ru.jpg (1000x800, 56K)

I also like to talk about trading forex/stocks in general and finding opportunities and volatility.

Bitmex has a maker fee of -0.025 regardless of volume, I'd try this first. If you have at least high five digits you can also go to bitfinex. They offer 0% maker fee, if you have a monthly volume of 7.5M$+.

FH5rGx
I will be uploading some tutorials on making bots soon.

Interesting, thanks. I want to try trading more regulated securities with a longer history than crypto first, but it seems like there should be a lot of opportunity in coins too since they're less examined by big players at least for now
Thanks for sharing your trading experience. Sounds like you're doing well, good luck with the prop firm interview

I need four digit tag to add you.

Please don't say you are gone and forgot to add four digit tag. Fuck sake.

The real REEEEE#9022
I linked the bot trading server before

Not bad, thanks

Not total bullshit. test preliminary strategy here:
tradingview.com (pine editor->strategy tester)
possible to get ~55-60% profitable trading on RSI/sRSI alone (not so much in the last few days due to the hyper bear market but meh). I think alg. testing probably works much better in a bull market though.

>I think alg. testing probably works much better in a bull market though.
Why's that? If bear markets behave very differently from bull markets, you could at least have separate strategies configured for them backtested specifically using previous bear markets?

your looking at doing statistic arbitrage
read up on it
me working on this sort of bot too

there deff am room for little guys

Why did you leave your HFT gigs ? I thought you would make a lot of money there as a dev.

I am in a quant fund that do directional trading. Well they dont have 1 indicator but thousands, with some basic TA thrown into the mix. I do not think its much easier for a retail than arbitrage because they have a ton of expensive data sources and the power to process and do simulation on that data every day is not cheap either.

Maybe you could try quantopian.com/ or other similar site which gives you the material at the cost of some of your alpha.

Personally I think you can make more money as a retail by investing instead of trading. Say investing in google 15 years ago and hold. Investing in crypto/tesla 5 years ago and hold. I bought some biotechs with promising treatment for cancer and I am holding them...

a TA thread in Jow Forums is always a remarkable thing, kek

I left because I made enough money that the hours and stress weren't worth it for me anymore.

I'm not going to try to compete with serious funds with big complicated models... I'm just thinking of doing it on the basis that no amount of quant funds will ever be able to arb away every opportunity to make say $100 a trading day on average if the models are only 60% or less reliable. Everyone's models are always misconfigured in some way, and if they aren't they become misconfigured within weeks just from changing market conditions.

For my savings I just buy and hold broad stock index funds. I don't think I have any particular talent for stock picking, and I'd rather try strategies where I can immediately tell if they're losing money and just turn them off for a while before I lose more than a few thousand dollars, as opposed to sitting on positions that'll randomly draw down 30% or more a few years down the line if I happen to be wrong about my picks

yep, you are right. (probably) don't expect to be able to use the same algorithm for bear vs. bull though (imo)

Trading views back tester is terrible , it often makes trades more than 2 candles late.

Guys, seriously, before saying anything just state how many bots and scripts you wrote and how much time you have been doing it. Otherwise it's bunch of clueless people making statements that reflect their wishes more than objective reality.

My journey started with MQL and static (not ai) strategies - wrote around 370 expert advisors (bots), custom indicators and scripts. Strategies came from all over the internet, books (read a lot on economy and trading before starting to write bots) myself and others.
Than I got into ai, finished multiple courses and got a good overview of the field, tools and methods - implemented many algorithms myself - starting from backprop NN to forests of decision trees working on discretized data using entropy theory, all to extrapolate.

In the long run nothing based on TA can work. And even though you use AI it's still TA.
Sorry, but if you waste enough time - and it will take you multiple years to learn it all and become convicted, you will see.

Only possible way is if you have what James Simons has - all the data and people, check out his interviews on YouTube.
Also recently talked with a guy who worked for 10 years in banking industry with AI and told me that it was all for their clients, they knew it didn't work. Only thing slightly working was scanning news just before a company released financial statements.

Wasted a lot of life on these trials, it almost made a black hole out of me - you can only lose as much money and fail before you realise the truth, and if you are a persistent fuck you will only journey downwards that AI TA rabbit hole before you test it all.

This post wouldn't stop me when I started.

Attached: Screenshot_20180718-190620.png (1080x1920, 327K)

What do you mean specifically by "in the long run nothing can work"? You tried all that and nothing worked consistently even in backtests and paper trading? Or did they look good in testing but fail when actually deployed? Or did they work for some time but eventually stop working?

What assets did you trade?

I mean that you can tweak it so that it seemingly works but it's a game of chance.
Backteating won't help here either, you might have some luck with picking the right periods and parameters that 'work' for a couple of trades but the next time or the time after that will even things out.
Honestly when you see people writing tutorials teaching all these approaches in books etc. It's either that the person doesn't have a clue about putting real money to work and wants to show how a specific AI tool can be applied or that he knows it's shit and wants to get some money back or be recognized for writing such and such in hope of being employed or is in denial phase.

Abitrage algorithms or HFT algos are the only ones worth learning about in terms of profit not based on blind luck.
Also methods rooted in reality, such as what a guy I know does, he/his company buys undervalued stock in a company that he sees potential in, goes to shareholders meetings, proposes to introduce Blockchain token into the company and waits, as far he is very successful with this, but also has large background in CS, trading and blockchain.

Attached: Screenshot_20180713-091849.png (1080x1920, 294K)

Stocks, everything on forex and crypto.

*Backtesting

Hm, thanks, this is interesting feedback.

I'm still not sure I understand what the specific cause of your strategies not working was, though. Naturally if you're data mining for periods and parameters that happen to work you're guaranteed to overfit and the strategy will immediately fail when you deploy it. But if you use good machine learning practices and at a minimum separate out a validation data set from your training set there is no reason I can see that backtesting shouldn't tell you whether your strategy is going to be profitable before you place your first trade. Are you saying NOTHING you tried ever gave you even a percent or two odds over 50% on your trades? What kind of timescale were you trading on?

See, the problem is not overfitting as most commonly shown (overfitting in the light of prediction accuracy on the training data), it's that you have to pick a period for backtesting.
Then you divide that period into two, this gives you data for learning and for testing.
The question is whether you would get the same parameters for the AI if you would change this period, shift it a round a bit and the parameters differ.

All timescales for Forex, even custom ones, renko charts, periods based on prime numbers etc.
For stock mostly daily.
You can even choose periods or seeming time cycles through genetic algorithms but this won't give you much insight.

Attached: Screenshot_20180709-171427.png (1080x1920, 900K)

About that 50%, no, some scalping methods could of worked but I didn't inquire further when I saw that all profits would be eaten up by commission.

Why does it matter that changing the backtesting period changes the parameters? Let's say you use your half and half split method, you learn some parameters from the training half, and the validation half shows that the strategy has a 55% win rate with the selected parameters. Did that just never happen for you in practice, or did the 55% win rate in validation just never materialize when you actually ran the strategy? Are you sure you weren't doing something like "testing" 100 different strategies and picking the winners from the validation set, which would contaminate the validation data and still cause overfitting?

Another simple, yet not know by all insight: when you reverse a losing scalping strategy, you will almost certainly end up with another losing startegy, due to spread.

This is one reason I'm looking at futures, commissions are low enough that scalping for even 1 tick would be profitable, and that's already deep HFT territory that I wouldn't bother attempting as a retail trader. Scalping for 5-10 ticks would make commissions negligible

(That is, unless your stop loss amount is greater than or equal to your profit target, in which case a 50-something win rate would get eaten by commissions)

What are you doing when backtesting, you try out the algorithm, you want to see if it would perform well in the past.

Therefore you pick a period. Let's say last 200 days, starting date, ending date and split ratio.
Now, yey, it works with 55% accuracy.

What do you do now?
Do you choose another period, another split ratio, another starting and ending date or is it that the last data is most accurate for you and testing it differently would not yield meaningful results?

History is finite. If the backtesting periods will overlap, this would cause anomalies. You can only backtest as much, which is nearly none.

To answer your question, it didn't materialize for me, e.g. I could be up in the first month than it would reverse on me.

I think most quant funds use AI and it works (beats index funds over years) but like you said, you need ridiculous amount of expensive data + process simulation everyday with smart people understanding it.

And I second the field work with small stocks. Sure its not sexy but its the best way to make money in finance as a retail imo.In 4 years of trading I never made money consistently with TA but I did not try with futures.

It's also that you can try nearly infinite combinations, when you will retrain the network, will you teach it or build a new one.
What are the parameters, the kind of network used, maybe you should use something different than a network, maybe an octopus network talking with other network, maybe recurrent one?
Maybe some Boltzmann machine in order? Some Deep Belief Network?

And the things you can do to data is only limited by your imagination.

So I cannot say that I tried it all, but you get the picture and after a while you are convinced about the limitations of TA and AI.

Attached: Screenshot_20180708-164041.png (1080x1920, 513K)

So you train on the first 100 days, and now you have parameters that are known to give a 55% win rate on the last 100 days. Why would you need to try a different period? You have your parameters, and you know that they would have performed at 55% over those last 100 days. Doing the calculation again with a different period and picking the "best" set of parameters of the two already starts to cause the exact validation data contamination problem I mentioned.

Thanks for sharing your experience though, it gives me more reason not to bother even entering any trades unless I have something that's paper traded profitably for at least a few months. What's the longest period of time a strategy would look like it was working before it went south on you? Did you ever implement some kind of statistical testing that would give you the probability that your strategy is still performing at 55% but is just having a bad run, vs the probability that it's no longer performing at 55% and should be shut off and replaced with a new strategy/parameter set?

Seems like you can try a thousand different network configurations, train them all on your training data, pick the top few based on the training data, evaluate ONLY those few on the validation data, and either keep or throw out the whole set based on their aggregate performance in validation? If you do this there should be no consequences to trying as many models and parameter combinations as you want, other than cost of computational resources?

Agree, right now the keyword is infrastructure, HFT or data mining on the scale of Simons requires vast funds, knowledge and smart people (who without that infrastructure wouldn't be of much use)

It's just brutal seeing many people being lost in the false knowledge, and the smarter ones milking them. (TA)

You don't need to try another periods. The reason behind what I wrote was that backtesting should check the strength of the method/algorithm not the data.
If you would have more relevant history than you would see if it really has the strength it shows on the first period. But history is limited.

Prices are like shadows, you have 2 great boxers fighting and you try to predict the outcome of their next fight by looking at their recorded foot positions during last fight.
The understanding of price and what it represents is missing.

It always went south depending on the periods and parameters, I think around 14 trades, you might feel it's all going good but it somehow gets you in the end.
Now you think, what about SL on capital etc. Just don't ;)

Attached: Screenshot_20180708-161503.png (1080x1920, 868K)

> It's just brutal seeing many people being lost in the false knowledge, and the smarter ones milking them. (TA)

I have an acquittance working for a famous market maker who told me they have specific models to milk retail flow. And i have seen it happen, sudden surge of volatility on a stock just to chase stop loss orders. The stock would go back up right after touching some limit of a famous TA indicator.

Agree again. You can breed the NNs in a genetic algorithm and get the greatest by your function of greatness from the population over many generations.

But it doesn't insulate you from the assumption that there exist patterns in historical data that might show in the future.

I don't doubt you, but I definitely feel like something fishy is going on if you have a strategy that performs well on 100 days of sight-unseen trading data, and then 14 trades later of exactly the same kind of sight-unseen price action it always doesn't work anymore.

What does SL stand for? Stock loan? I don't intend to trade equities at all, way too many costs and funky complications. Futures have no short sale restrictions, don't require borrowing anything, allow posting insanely low margin, and don't charge any interest on the rest of the value of the contract. Spreads are tight in the liquid ones and unlike in forex there is no need to cross the spread at all, depending on your strategy.

> But it doesn't insulate you from the assumption that there exist patterns in historical data that might show in the future.
Yes, this is a key point. Do you think there simply aren't any patterns like this to be found, like with the efficient markets hypothesis?

Sounds like alot of limit orders and trailing stops.

Market makers are out to get you. It's right there in the agreement when you sign up for a account that there exists conflict of interests and you must accept it. (At least in Europe) Also many stories about this on the internet.
Always used ECN for mine accounts.

But there are many levels of price manipulation, e.g. many times the small stock indexes in European countries are being supressed just because some whale predicts the reversal on larger indexes (like SP500/DAX) and is selling into apparent strength or buying into apparent weakness, learned about it from a guy who did this in his daily job.

Fishy because you asked about the longest. The average was a couple of trades. Statistically evens out to 50\50.

There is something called strategy tester in metatrader that shows you exactly when and where and how did you win or lose, it also tells you the longest losing and winning streak.

SL = stop loss, level at which you automatically close the trade

Yeah this is why I'm not going to bother with spot forex at all; trade directly on a regulated exchange and you don't have to submit your orders to dealers who are openly trading against you.

Price manipulation should be a non-issue though unless you are trading based on fundamentals

No, I meant it's fishy because 14 trades is way too short. A couple of trades is beyond fishy. How on earth does a strategy that's profitable for 100 days IMMEDIATELY revert to 50/50 within one or two trades when you put it into production, unless you're data mining on your validation set? I really don't get this at all

Oh, I guess my question may not have been clear. I wasn't asking about winning streaks, but rather how long your strategies remained profitable before they went negative overall. A 55% win rate isn't going to get you many winning streaks; I agree if you're looking for those with TA you're guaranteed to have a bad time

On the efficient market hypothesis you should turn to the 'only mathematician billionaire' I mentioned earlier.

I do think there are patterns, those are purely psychological and change over time, meaning you cannot rely on the ones reflected in the historical data.

The most common psychological patterns are already accounted for by big whales or liquidity provides who often know that there will be someone looking to place trade at this and that level because of TA and will put SL and TP (take profit) within predictable range.

Insight matters, if you know how e.g. ECB will react to certain drowdawn in EUR/USD you can make long term goals, seeing that after the day ends economy MUST be working, therefore less or more visible cycles do exist and given enough assets and time you can play it for it's money.

Your strategy just does not work anymore. Maybe other market participants found it, maybe they changed the way they trade ... Plenty of indicators work well only for a few months and you have to discard them. You constantly have to find new ones, especially if your time horizon is daily or less.

I can tell you with absolute certainty that if you will take a basic, shitty strategy like MA crossover and pick just the right parameters and starting date you will get more than 14 wins in a row, doesn't matter over how many days, it can be over many years in fact, or whatever really.

For mean reversion strategies it would be a problem if big whales see your pattern first and trade enough size to move the price before you, but for trend/momentum strategies their trading would only exacerbate the trend, no?

Sure, but assuming your test data includes the most recent previous trading days (which seems like a sensible method to me at least), then the only way for this to happen consistently would be for other market participants to randomly happen to find your pattern or change how they trade EXACTLY on the day you happen to start running your strategy. Otherwise wouldn't you already see the loss of effectiveness in your test data before you even start trading it?

You mean in backtesting, right? I don't doubt that at all, but again "picking just the right parameters" to get this result is not helpful. If you were getting 14 wins in a row in actual trading I'd think you'd be pretty profitable if you cut your losses quickly when your streak ends?

The momentum or ma are just stastical formulas you can check out in MQL code repository, algorithm's result names, they don't mean much over what you ascribe to them.

If the liquidity provider or a whale predicts that market is going one way and sees that if his move would bring price to certain level so as to make you think it's doing the opposite, he can get more cheap assets out of your pocket.

Attached: Screenshot_20180805-184744.png (1080x1920, 236K)

>doesnt use NapoleonX token's trading signals

Lmao their signals are basically free money.

The thing is, I suggested this cash SL earlier, that you don't know at the time and any strategy comes out as good as any other, you simply don't have the means to judge that.

Let's say you have a 'opposite' strategy that nets you 14 losses in a row.
When will you stop hoping for a win. The same goes for the winning one, over time, evens out.

I'm not too worried about big players predicting my trading desu... this is certainly possible if you're trading based on one or two simple TA indicators, but I doubt anyone makes money doing that even without any stop hunting. But as you said before, there are infinite varieties of settings and parameters you can apply to a strategy that's even marginally more complex than that, and if someone is trying to predict THAT trading behavior, well, good luck to them... Besides, if this behavior is happening in the market the pattern will already be visible in the price data, and any strategy that gets predicted and stop hunted this way will test poorly

it'll work in most cases, until those unexpected moments where it shits the bed and you lose everything

how much would you trust your code?