programming profitable trading strategies in ninjatrader 8 w scott hodson

NickF

Aloha State All,

Sorry if I am request a question that may not belong to this radical.
I am currently doing some testing, trying to substantiate if there is whatsoever truth in some common strategies.
For example, I am checking to see if the next bar later a daily engulfing bar is raising and the results are non any better Beaver State worsened than 50%.
I found the same for pivots. In that respect is no edge in purchasing after a daily pivot is encountered soh, as far as I am concerned, seeing a pivot in a daily chart agency absolutely cipher.
Flush if a pivot follows later many falling parallel bars, it can non predict better than ergodic if the cost trend will change.
That is, if my code and thoughts are correct, there is no reward to debate engulfing bars or pivots as part ofanydannbsp;strategy, considering their success rate is very close to 50% when analysed severally.

Does anybody know any indicant that would provide an edge (even a tiny edge) when trading, that put up follow confirmed by spurting only that same indicator on a large number of trades?
My thoughts were that if we combine a few different indicators in a strategy, peradventur we can addition the odds.

I am working to test a strategy which says that most day-to-day gaps fade.
I found a pretty small correlation about the day of the week, basically it looks like a speculative idea to adjudicate to fade gaps connected a Monday. Thusly farther I didn't analyse the Pdanamp;L of col trading.
Is there any chance of finding a working strategy based on gap fading that could cater over 10% profit per annum?

Hawaii Nick,

Unenviable questions indeed, and whilst people are unlikely to share their attribute secrets, I can tell you that a decade past I was paid to research the lustiness of traditional technical trading. Information technology did not menu well.

Best wishes,

toggle quoted messageShow quoted textual matter

On Sat, 8 Dec 2022, 05:43 NickF danlt;nfieraru@... wrote:

Hi All,
Sorry if I am asking a question that may not belong to this group.
I am currently doing some testing, disagreeable to confirm if there is any truth in some common strategies.
For instance, I am checking to see if the incoming bar after a each day engulfing bar is raising and the results are non whatever better or worse than 50%.
I found the same for pivots. There is zero edge in buying after a daily pivot is encountered so, as far as I am concerned, seeing a pivot in a day-after-day chart means absolutely zipp.
Even if a pivot follows after many falling bars, it can not predict better than random if the price trend leave change.
That is, if my computer code and thoughts are correct, there is no vantage to consider engulfing bars or pivots as part ofanydannbsp;strategy, considering their success rate is very close to 50% when analysed independently.

Does anybody sleep with some index number that would provide an edge (even a small edge) when trading, that can live unchangeable by running only that one indicator on a magnanimous number of trades?
My thoughts were that if we combine a few different indicators in a strategy, peradventur we can increase the odds.

I am employed to test a strategy which says that most daily gaps fade.
I constitute a bad weensy correlativity about the daylight of the week, basically it looks like a bad idea to try to fade gaps on a Monday. So furthest I didn't examine the PdanA;L of gap trading.
Is on that point any chance of determination a temporary scheme based happening gap fading that could provide over 10% net per year?

In my experience it by all odds can but IT is non so easy to find a profitable strategy. Just trading a candle holder pattern or an indicator will most belik non turn.

Hi.

In my opinion, both recursive trading and discretionary trading could be profitable or they can be existent disasters. It's all depend.

Any profitable arbitrary dealer World Health Organization has objective rules can curriculum them and turn them into bankable recursive trading.

This is one of the best websites according to the CMT that has best characterized candle patterns til now.

Machine learning, walking in the cutting edge.

Several times, simplest is better

Although mastering the art of backtesting, it's not scarcely having a obedient handful of rules that have behaved well at whatever given time.

on-off switch quoted messageShow quoted text

El 8 dic 2022, a las 6:42, NickF danlt;nfieraru@...dangt; escribió:

Aloha State All,

Depressing if I am asking a question that may not belong to this group.
I am currently doing some testing, trying to affirm if there is some truth in close to grassroots strategies.
For example, I am checking to see if the next bar after a daily engulfing bar is raising and the results are not any better operating room worse than 50%.
I recovered the same for pivots. There is no edge in buying after a daily pivot is encountered indeed, equally far every bit I am concerned, seeing a pivot in a daily chart way absolutely nothing.
Even if a pivot follows after many falling bars, it can not predict better than stochastic if the damage trend will change.
That is, if my code and thoughts are correct, in that location is no advantage to consider engulfing bars or pivots as part ofanydannbsp;scheme, considering their success rate is very penny-pinching to 50% when analysed independently.

Does anybody know any indicator that would supply an edge (even a small edge) when trading, that can be confirmed by running merely that one indicator on a large number of trades?
My thoughts were that if we combine a a few distinct indicators in a strategy, maybe we can increase the betting odds.

I am working to test a strategy which says that all but daily gaps fade.
I found a pretty half-size correlational statistics near the mean solar day of the week, basically it looks ilk a bad idea to try to fade gaps on a Monday. And so far I didn't analyse the Pdanamp;L of gap trading.
Is on that point any chance of finding a working scheme based on gap fading that could provide over 10% profit per year?

NickF

Thanks to wholly who provided an answer and links, information technology is much comprehended.
After thinking a bit over my original question, at present I imagine it is more coordination compound than I thought.
Let's assume that I am correct in saying that a pivot and an engulfing candlestick are not able-bodied to provide by themselves some edge on time unit parallel bars and this proof is based on running these indicators on thousands of stocks over long-wool periods.
Is it manageable they may still be useful in compounding with different indicators? Or is it enough to check them singly and if they don't show any benefit, they can exist safely discarded from the toolbox?
I think I can now render an respond to that.
Let's say I rich person a hypothetical indicator, which is always right when the cost is above agitated average 200 and it is never right when the price is down the stairs simple moving average 200.
If I play my try for 20 years on thousands of stocks, and if at that place is an amount of time when the prices are in a higher place and below SMA200, it will come along that this indicator is worthless, having a success rate of 50%.
While the loser was that I did non ply the kosher conditions where this indicator really shines (which is to use it only when the price is to a higher place SMA200 or use IT in opposite manner when the price is below SMA200).

Opening trades supported one or several indicators is difficult. Most indicators show something after it happened (they lag). I have better results with trying to predict how other commercialise participants react in a destined situation. Then indicators can be precise useful as filters just like in your example.
Like take only a trade when price trades above the SMA 200 for representative.

Then you always have to be careful not be cost caught curve proper when adding indicators as filters to better carrying out in your backtests. Also selection bias is something you want to Google.

Another fashio to look at 50% success rate is that the indicator has no staunch relationship with prices. If there's no constant human relationship then pairing it with another indicator that has no constant relationship with price is not departure to help.

Probably, you'll find this site functional to transfer this discussion.

https://quant.stackexchange.com/

Entries are only one abuse of the strategy, some argue that the of success come in from the exits (and risk management)

There are gainful strategies successful less than 50% of time.

toggle quoted contentShow quoted text

El 8 dic 2022, a las 13:04, Desperation danlt;boris@...dangt; escribió:

Opening trades based on one or several indicators is difficult. Most indicators show something after it happened (they lag). I have better results with trying to predict how other market participants respond in a certain situation. Past indicators posterior be same useful as filters just like in your example.
Like take only a trade when damage trades above the SMA 200 for good example.

And then you always have to cost elaborate not beryllium be caught kink just when adding indicators As filters to improve public presentation in your backtests. Also extract bias is something you lack to google.

NickF

Hi Nick,
That is the interesting partly - I now disagree with that statement.
For example, if you amount alternate current with a District of Columbia voltmeter, it will show up zero volt. That is because the average voltage of an AC current is 0 volt. One could assume by this that there is no voltage on the wires. But if we strain out the negative set forth of the topical through rectification, we would suddenly discover thither is electric potential present.
The corresponding way, in the example I delineate early - having a fictional indicator that provides 100% right results during bull period and 0% accurate results during bearish periods and assuming these two periods are equal during the test stop, information technology may look that the indicator has no real correlativity with the price. It would be single when we combine that index number with some other indicator to filter out the bear period and leave just the Bull period that information technology would show the usefulness of the fictional index number. Of of course (and that is the near belik scenario with many indicators that take in 50% success ratio over long term) it is possible that so those indicators don't work and therefore are not useful, either by themselves OR in combination.

Yes, that is what I was nerve-racking to say. If you has an indicator that correctly told you optimistic and pessimistic periods then you wouldn't need anything else, would you?

toggle quoted messageShow quoted textual matter

Connected 12/8/2018 8:37 AM, NickF wrote:

Of course (and that is the most in all likelihood scenario with many indicators that have 50% success ratio over long term) information technology is possible that indeed those indicators don't puzzle out and therefore are not reclaimable, either by themselves or in combination.

Greetings --

I correspond with the comments that this may non be the best thread for discussion of this topic, merely the matter is remarkable.dannbsp; I have researched, applied, backhand, and posted extensively about development of trading systems.dannbsp; My internet site has several articles indirect to development of trading systems.dannbsp; Lead off at www.blueowlpress.com.dannbsp; Click the "System Evolution" tab, and read through the articles.dannbsp; Also read the "Papers" and "Posts" and watch the "Videos."dannbsp; All of that is completely aweigh.

I look to Ewald for guidance on where promote discussion could take place.

Best regards,dannbsp; Howard Unfit

toggle quoted messageShow quoted text

Happening 12/7/2018 9:42 Postmeridian, NickF wrote:

Hi Completely,
Sorry if I am asking a question that may not go to this group.
I am currently doing some testing, nerve-wracking to confirm if there is any truth in both common strategies.
For exemplar, I am checking to see if the next bar after a daily engulfing exclude is raising and the results are not whatsoever improved or worse than 50%.
I found the same for pivots. There is no edge in buying after a daily pivot is encountered so, as long Eastern Samoa I am concerned, seeing a pivot in a daily chart means utterly zipp.
Even if a pivot follows after many down bars, it can non predict better than random if the price trend testament change.
That is, if my code and thoughts are correct, at that place is no advantage to consider engulfing bars or pivots arsenic part ofanydannbsp;strategy, considering their success pace is very short to 50% when analysed independently.

Does anybody know any index number that would supply an border (even a small edge) when trading, that can be confirmed by running only that one indicator on a large number of trades?
My thoughts were that if we combine a some different indicators in a strategy, maybe we can increase the odds.

I am working to test a strategy which says that just about every day gaps fade.
I found a pretty small correlation about the day of the week, basically it looks the likes of a bad idea to try to fade gaps connected a Monday. So far I didn't study the Pdanamp;L of gap trading.
Is there any chance of finding a temporary strategy based on gap fading that could provide o'er 10% net income per year?

I am quite puzzled when you mix the auto trading and machine erudition, they are not supposed to choke hand in hand, I spent quite an few month implementing the several different auto learning technology into my app, and just found myself au fon imperfect (to me the Wisdom of Milliliter might glucinium relevant to a too large time redact, and auto trading is for much smaller clock systema skeletale to take vantage ofdannbsp; laborious repeatative tasks)

so I somebody still do expend time along ML, that's my advice, forget about IT!

Jeff --

While I somehow "understand" your view and "defeat" on using ML techniques to "predict" the markets, I believe that you harbor't selected the best words to express yourself. I hope that you don't feel attacked with my words, information technology hasn't been my purpose anytime.

Of line, there is no holy Holy Grail on victimization ML to trade the markets; the approach has some advantages and disadvantages equally everything other in earth. ML approach, together with the subject field implementations that are involved, will not give anybody "great" results on the first try. Still, recounting in a generalized speech that ML research (I understand research from your actor's line "spending time") should be forgotten by a spick-and-span comer as NickF because of not acquiring good results in your personal investigation, seems a second non sense for me. Just search the vane about all the papers and fields that are undergoing great research from big and little institutions; sometimes with better outcomes than others.

YES !dannbsp; merely i keep my work for Maine, sorry ...

On Sat, 8 Dec 2022, cross off collins wrote:

Difficult questions indeed, and whilst people are unlikely to share their
personal secrets, I rear end secernate you that a decade ago I was prepaid to research
the robustness of traditional technical trading. Information technology did not fare healed.

The definitive arrangement survey work therein area is probably:

Evidence-Based Technical Analysis: Applying the
Scientific Method acting and Applied mathematics Inference to Trading Signals
(Wiley Trading Book 274) Feb 2, 2007
by David R. Aronson

The work defines, then runs several systems, each without
compliments to 'market regime' as to applicability of a arrangement, and
essentially concludes that TA systems are not, standing away
themself, able to emit internet returns along a sustained, and
hands-off basis

Yet, clearly, by adding supervision as to when to run a given
scheme, one rear end pull out 'excess' returns. Attend John Carter
and Hubert Senters' work with their TTM indicator maybe tenner
Beaver State 15 years ago. I recognize I personally traded well with that
counsel, but had difficultness removing the 'artistic'
supervision, and distilling it down to a 'State machine' of
natural philosophy rules

Looks as though John has a new edition coming come out next week:

Mastering the Trade, Thirdly Edition: Proven Techniques for
Profiting from Intraday and Swing Trading Setups
Dec 18, 2022
by John F. Carter

Trading the Senters Crush Play Strategy
Multimedia CD – 1999
by Hubert Senters

The text of the Carter endflap seems to be:

Carter provides:

•Proven set-ups, with optimal markets and non-negotiable
trading rules
•Exact entry, exit, and stop consonant loss levels for swing ou and
intraday trading
•Seven key internals, from Skew to VIX
•Pre-food market checklist for analyzing recent market demeanour
•Scanning techniques for pinpointing high-probability setups
•Effective risk control techniques
•Methods for ensuring your data processor runs at max speed
•Techniques for predicting market corrections

and completely I force out aver is about how:
'no generalization is worth a damn, including this
one'

If one needs ideas to implement and trial run, once could act a lot
mold than that list ... Just don't expect to be able to number
IT on, ignore it, and have it net emit a positive return

-- Russ herrold

On Mon, 10 Dec 2022, R P Herrold wrote:

and all I can say is about how:
'no generalization is worth a damn, including this
one'

I wanted to quotatio in a distinguish objet d'art a clear exception to
that generalization. It is Sir Thomas More for a Systematised (although
perfectly capable of existence reduced to an Algorithm, Eastern Samoa the
source has done at GitHub) but on an Investing, kind of than a
'short term' trade basis

EOD data, the Saturday copy of IBD, and a pencil and paper
_can_ do the requisite computation, although he implements in
Python

Systematic Trading: A unique new method for design trading
and investing systems
1st Variant
by Carver Henry Martyn Robert

and here, clearly, able to algorithmically harvest returns on
a largely manpower away basis (EOD data so, ...)

boring, grinding it out, and no entertainment value at all
watching tickers jitter

-- Russ herrold

Greetings --

The references mentioned in this thread will cost useful, but they are not state of the art.

... All of them mentioned assume the information being processed is stationary.dannbsp; Stationarity is a feature film of distributions.dannbsp; Many of the model developing techniques delineate in the literature work with information that is unmoving -- loan repayment or epitome sorting, for examples.dannbsp; But unclothed fiscal data is not stationary.dannbsp; In order for a model to be serviceable for prediction, the future must resemble the preceding.dannbsp; That is, the information that bequeath be processed as the trading signal is generated using fresh information must have the same statistical distribution as the data that was put-upon to develop the model.dannbsp; One of the first stairs in developing a tradable model is applying transformations that result in the data being fairly stationary.dannbsp; Later, when trading, and losings come, one of the direct reasons is 'distribution drift' -- loss of stationarity between the predictor variables and the target.dannbsp; A well designed trading organisation must constitute able to discern distributional drift and adjust put back size to subjugate risk.

... All of them mentioned are explaining.dannbsp; The modeling summons used is fitting the manakin to the information and assuming that the backtest is adequate.dannbsp; It never is.dannbsp; A triple-crown backtest is necessary, but is not comfortable.dannbsp; Our primary need is for accurate prediction.dannbsp;dannbsp; In set up to assess the risk and profit potential of a trading system, we need validation.dannbsp; Substantiation is the most important aspect of the scientific process in mock up development.dannbsp; That is, gather data, define a model, fit the pose to the data, trial run the model on data that has not been used in the fitting process.dannbsp; Only the validation results have value in estimating future performance.dannbsp; Backtesting results give no value (otherwise draft our attention to a possibly profitable proficiency).

... All of them use a single decision tree as the model.dannbsp; Single decision trees have a provident history of use in trading systems.dannbsp; The traditional trading system development platforms (TradeStation, AmiBroker, NinjaTrader, etc) all use the single decision tree diagram as the model.dannbsp; It is the only mannikin available for most traditionalistic growth platforms.dannbsp; Until the 1990s, decision tree models were adequate.dannbsp; That has changed.dannbsp; Successful trading depends on accuracy of the model --
better accuracy than other traders are using.dannbsp; Powerful, low-priced computers, together with urbane model techniques, now enable much many accurate trading systems than can be developed for single decision trees. In model accuracy competitions, such as Kaggle, individual decisiveness trees are never among the best models.dannbsp; Current state of the art uses ensembles (often consisting of thousands of decision trees), boosted models, and deep learning.dannbsp; All but large trading companies are basing their trades on these more accurate models.

... They may idea profit potential, but they do not properly assess risk.dannbsp; Adventure is risk of drawdown in the trading account. There are individual aspects of risk that we must take into account -- our personal risk tolerance, the risk inherent in the data stream itself, the risk of exposure of losing trades, and the adventure of changes in the relationships that define the trading organization.dannbsp; Position size is a precise important aspect of trading and a critical component of report growth.dannbsp; Position size is settled and moderate by the risks, including recent trading performance.dannbsp; Risk assessment and position size determination is a function of the management of a trading system and should not be a component of the sign coevals pose.

Unexcelled regards,dannbsp; Howard

toggle quoted messageShow quoted text

Along 12/10/2018 1:40 PM, R P Herrold wrote:

On Mon, 10 Dec 2022, R P Herrold wrote:
and all I can say is about how:
'no inductive reasoning is worth a damn, including this
one'
I wanted to mention in a offprint piece a clear exception to
that generalization. It is more for a Systematic (although
dead capable of being reduced to an Algorithm, as the
source has done at GitHub) but on an Investing, preferably than a
'short term' trade basis
EOD data, the Sat copy of IBD, and a pencil and paper
_can_ do the needed computation, although helium implements in
Python

Systematic Trading: A unique new method for scheming trading
and investing systems
1st Edition
by Woodcarver Robert

and here, clearly, able to algorithmically harvest returns on
a largely workforce off basis (EOD data so, ...)

boring, detrition IT verboten, and No entertainment value at complete
watching tickers jitter

-- Russ herrold

THANK YOU to R P Herrold and Howard Bandy, for perhaps the best summary I've seen in a while.
100% tall on all points.
LOTS of fluff danamp; religion out there, but big-shouldered facts churn down to those posts.

I would recommend version some of Dr. Bandy's books, in caseful you didn't already know he's a well noteddannbsp;quantitative monger and author

https://www.amazon.com/Howard-B.-Bandy/e/B00MBDSZ3S

As utmost as my own quantitative systems, you can tap existing edges and use technicals to signal you when to use these edges. For example, here are some fit-known "edges" you can exploit, with lots of risk management added in

- Market characteristics, like "U.S.A equities generally burn down over the long term", so maybe only consider long-lasting-sole systems on US stocks in a bull market, like when SPX dangt; its 200-Day moving medium, and look for technical analysis buy signals during these bullish phases

- Mathematical peculiarities, look-alike
dannbsp; dannbsp;- Options price disintegration over time, entirely things being equal, so look for options strategies that sell options patc managing the risks at that place. In person, I algo-trade in SPX straddles connected expiration day because in 6.5 hours the extrinsic value of the straddle at the open is going to be valueless, so you have to manage the gamma risk of the underlying's movements by non material possession on to the side of the span that's going against you for too long, and come back in if the market turns around.
dannbsp; dannbsp;- Short VIX ETNs/ETFs like VXX or UVXY due "roll-yield", only lone when the VIX term structure is in contango, or sell calls/buy puts. This trade broke very much of multitude hypnoid at the wheel in Jan/Feb and IT hasn't been as profitable since the ETN leverage ratios have down, but there's still meat along the bone
dannbsp; dannbsp;- Stubby 3x leveraged ETFs due to "daily performance" tracking mathematical flaws ("beta slippage"), merely beware of/hedge against risen-trends, some try exploitation pairs, on the other hand there's this... https://seekingalpha.com/article/3070136-shorting-leveraged-etf-pairs-easier-aforesaid-than-through with

- Well-known seasonalities, like
dannbsp; dannbsp;- USdannbsp;stocks: "Corrupt after Halloween, Sell in May", trade long only during November 1-Apr 30, and use technical school analysis to look for buy signals in this typically bullish period
dannbsp; dannbsp;- There are other known seasonal worker patterns you can exploit, like in petroleum (bullish spring into the summertime unremarkably pessimistic into fall and winter)

...but seasonal patterns are not always repeated so use cautiously.

You can combine these, like only take bullish T.A. signals Nov-Apr when the SPX dangt; 200-day MA, the Sir Thomas More filters you summate, though, volition limit the amoun of trades you take, if that's OK with you, depends connected your personality and how frequently you wish to be in and out of the market.

Or s things to chew on...

______________________________________________
Scott Hodson / Founder, Probable Trades

NickF

Thank you to everybody who contributed with advice, opinions and ideas to my question. I will try to do more research in these directions.

Nick